sneak/routewatch
1
0
Fork 0
Code Issues 2 Pull Requests 1 Actions 25 Packages Projects Releases Wiki Activity

Compare commits

base: sneak/routewatch:next
Branches Tags
sneak/routewatch:main sneak/routewatch:next sneak/routewatch:issue-27-cached-stats sneak/routewatch:prod
..
compare: sneak/routewatch:4c2882c19c3159f9b3f770988c8e0c1762b31065
Branches Tags
sneak/routewatch:next sneak/routewatch:main sneak/routewatch:issue-27-cached-stats sneak/routewatch:prod
2 Commits
next .. 4c2882c19c
Author SHA1 Message Date
sneak 4c2882c19c Write batches through one connection with a 1 GiB cache, in key order (closes #96)
check / check (push) Canceled after 0s
Details 
Every write now uses one connection taken from the pool when the
database opens, under the database lock. Its 1 GiB page cache keeps
the pages the route writes touch from one batch to the next. A reading
connection's cache is emptied whenever another connection has written,
so the 9 reading connections drop to 16 MiB each. The soft and hard
heap limits rise to 1.5 and 2 GiB, above the caches; README "Memory"
adds up the new budget. Vacuum, Analyze and Checkpoint check their
context after the lock, which the kept connection does not. Batch
writes sort their rows by peer, then prefix. The benchmark also times
the largest prefix handler batch, 250,000 updates.

Model: opus-5-5
 2026-10-19 12:53:52 +00:00
clawbot 1f612642e3 Make live route writes cheaper so the prefix handler keeps up (closes #96)
check / check (push) Canceled after 0s
Details 
The live route tables lose their random UUID key: the row id, which the
prefix-length sample draws, is now declared as the integer key. The prefix
tables are keyed by the prefix, without the UUID or the duplicate index.
A route update sets only the columns that change, so SQLite no longer
rewrites the mask length indexes on every update, nor the origin index
while the origin AS stays the same. Each kept index names the queries
that use it. BenchmarkPrefixBatchOnFiveMillionRoutes, run by hand, times
one 25,000-update batch on 5 million routes.

Model: opus-5-5
 2026-10-19 11:58:03 +00:00
53 changed files with 779 additions and 5593 deletions
Expand all files Collapse all files
.dockerignore
+2 -12
View File
@@ -73,8 +73,7 @@
# This repository's host-built artifacts: `make build` writes bin/, `go
# build` in the root the binaries, `make run` log.txt, `make asupdate` its
# .tmp file; logs, test binaries, coverage profiles, source archives, and
# the binaries and libraries a build for another platform writes.
# .tmp file; test binaries, coverage profiles and source archives.
/bin
/routewatch
/asinfo-gen
@@ -83,24 +82,15 @@
/out
/pkg/asinfo/asdata.json
**/*.tar.zst
**/*.log
**/*.test
**/*.out
**/*.tmp
**/*.exe
**/*.dll
**/*.so
**/*.dylib
# Local databases. The image carries a source archive of the whole build
# context, so these would otherwise ship inside it. The daemon's database
# is db.sqlite, with its -wal and -shm files while it runs.
# context, so these would otherwise ship inside it.
**/*.db
**/*.db-journal
**/*.db-wal
**/*.sqlite
**/*.sqlite-wal
**/*.sqlite-shm
# A local Go workspace points at directories outside the build context.
/go.work
.gitignore
+1 -5
View File
@@ -67,14 +67,10 @@ node_modules/
go.work
go.work.sum
# Local databases, temporary files and debug output. The daemon's
# database is db.sqlite, with its -wal and -shm files while it runs.
# Local databases, temporary files and debug output.
*.db
*.db-journal
*.db-wal
*.sqlite
*.sqlite-wal
*.sqlite-shm
*.tmp
log.txt
out
Dockerfile
+4 -4
View File
@@ -24,8 +24,8 @@ RUN --mount=type=cache,target=/root/.cache/go-build \
# image ships and the alpine one does not. -count=1 because the mounted
# cache also holds test results, which go test would otherwise replay.
# -short skips the live RIS Live feed test, so the suite needs no network.
# golang:1.27.2-bookworm, 2026-10-19
FROM golang@sha256:5cf287a799e6b94384bad13d16b14904c531f51ba65792237e122ce42b392f61 AS test
# golang:1.24-bookworm, 2026-09-21
FROM golang@sha256:1a6d4452c65dea36aac2e2d606b01b4a029ec90cc1ae53890540ce6173ea77ac AS test
WORKDIR /src
COPY go.mod go.sum ./
RUN go mod download
@@ -39,8 +39,8 @@ RUN --mount=type=cache,target=/root/.cache/go-build \
# Build stage. Nothing is wanted from either phase above; the copies
# are what make BuildKit build them first, so nothing after them can
# run unless lint and test passed.
# golang:1.27.2-bookworm, 2026-10-19
FROM golang@sha256:5cf287a799e6b94384bad13d16b14904c531f51ba65792237e122ce42b392f61 AS builder
# golang:1.24-bookworm, 2026-09-21
FROM golang@sha256:1a6d4452c65dea36aac2e2d606b01b4a029ec90cc1ae53890540ce6173ea77ac AS builder
# Install build dependencies (zstd for archive, gcc for CGO/sqlite3)
RUN apt-get update && apt-get install -y --no-install-recommends \
README.md
+37 -168
View File
@@ -5,22 +5,6 @@ table in real time: it streams BGP UPDATE messages from the RIPE RIS Live
service, maintains a live routing table in SQLite, and provides HTTP APIs for
querying routing information.
## Rationale
RIPE RIS Live streams the BGP updates that RIPE's route collectors receive, as
they happen, from the moment a client connects. It reports changes, not the
routing table they add up to, so it cannot say which AS an IP address is routed
to now or which routes a prefix has. RouteWatch applies the feed's updates to
live routing tables in its own SQLite database and answers IP-to-AS lookups and
prefix pages from them.
The feed sends no table when a client connects, so these tables hold only the
routes the feed has announced since RouteWatch first started: an IP-to-AS lookup
finds nothing for an address whose routes have not been announced since then.
The tables also miss the updates RouteWatch drops when it falls behind under
load and those sent while it is stopped, so a lookup or a prefix page can show a
route the feed has withdrawn.
## Features
- Real-time streaming of BGP updates from RIPE RIS Live
@@ -52,76 +36,23 @@ DEBUG=routewatch ./routewatch
## HTTP Endpoints
Every route answers `HEAD` as it answers `GET`, without the body.
### Web Interface
- `GET /` - Home page (HTML): headline figures, and lookups by AS number and by
IP address
- `GET /status` - HTML status dashboard, which reads `/api/v1/stats`
- `GET /status.json` - JSON statistics: feed, memory, database and route figures
- `GET /.well-known/healthcheck.json` - Health check (JSON): 200 when the
database answers and the RIS Live feed is connected, otherwise 503;
`data.status` is `ok` or `error` and `data.checks` says which check failed
- `GET /as/{asn}` - AS detail page (HTML): the prefixes the AS originates and
its peers
- `GET /prefix/{prefix}/{length}` - Prefix detail page (HTML), such as
`/prefix/192.0.2.0/24`: every live route for the prefix, one per peer, and its
origin ASes
- `GET /prefixlength/{length}` - Up to 500 IPv4 prefixes with that mask length,
picked at random, each with its newest route (HTML)
- `GET /prefixlength6/{length}` - The same for IPv6
- `GET /ip/{addr}` - The same JSON answer as `/api/v1/ip/{ip}`
- `GET /ip/` - The same, for the caller's own address: the address the request
came from or, when that is a trusted proxy (`TRUSTED_PROXIES`, see
Configuration), the client address in its `X-Forwarded-For`
- `GET /` - Redirects to /status
- `GET /status` - HTML status dashboard
- `GET /status.json` - JSON statistics
- `GET /as/{asn}` - AS detail page (HTML)
- `GET /prefix/{prefix}` - Prefix detail page (HTML)
- `GET /prefixlength/{length}` - IPv4 prefixes by mask length
- `GET /prefixlength6/{length}` - IPv6 prefixes by mask length
- `GET /ip/{ip}` - Redirects to prefix containing the IP
### API v1
- `GET /api/v1/stats` - The `/status.json` figures plus connection, garbage
collector, stream and handler queue statistics
- `GET /api/v1/ip/{ip}` - Looks up each address: its most specific live route
(`netblock`), that route's origin AS with its WHOIS details, and the address's
PTR names. `{ip}` is an IP address, a host name, or a comma-separated list of
at most 10 of these. Of the addresses a host name resolves to, only public
ones are looked up: a name with none, such as one that resolves to a private
address, is reported as `host name has no public address`, without its
addresses. `results` has one entry per address with a live route and `errors`
one per address or name that failed. The answer is 400 when `{ip}` lists more
than 10, when it comes to more than 10 addresses once its names are resolved,
or when nothing in it is an IP address or a host name with a public address;
and 404 when none of the addresses has a live route. A PTR lookup gets 2
seconds; an address whose lookup takes longer is answered without PTR names.
- `GET /api/v1/as/{asn}` - The AS and the prefixes it originates, IPv4 and IPv6
apart, each with when its newest route was updated
- `GET /api/v1/prefix/{prefix}/{length}` - Every live route for the prefix, one
per peer, and the same routes grouped by origin AS in `origins`
A JSON answer is `{"status": "ok", "data": ...}`, or for an error
`{"status": "error", "error": {"msg": ..., "code": ...}}`, and carries an
`@meta` object with the time zone, the API version and the time the request
took. The health check is the exception: its 503 answer is also
`{"status": "ok", "data": ...}`, with `error` in `data.status`.
A request that runs past its time limit answers 504: the limit is 30 seconds,
and 4 seconds for the database reads behind `/status.json` and `/api/v1/stats`.
An AS number, prefix or mask length that does not parse answers 400, as does a
mask length out of range for its IP version. An AS the database does not hold,
or a prefix with no live route, answers 404. The HTML pages give these errors as
plain text. Under `/api/v1` and `/ip/`, a path no route matches answers 404 with
the message `Not found`, and `POST`, `PUT`, `PATCH`, `DELETE`, `OPTIONS`,
`TRACE` or `CONNECT` answers 405 with `Method not allowed`. A 500 answer says
only `Internal server error`, and a host name that does not resolve only
`could not resolve host name`; the detail goes to the log.
Every answer carries `Strict-Transport-Security`, `Content-Security-Policy`,
`X-Frame-Options`, `X-Content-Type-Options`, `Referrer-Policy` and
`Permissions-Policy`. The Content-Security-Policy lets a page load content only
from the server, and does not allow inline content with `'unsafe-inline'`: the
pages keep their CSS and JavaScript in inline `<style>` and `<script>` blocks,
so each answer's policy names a random nonce made for that answer, and the
browser applies or runs only the blocks that carry it. The pages therefore use
no `style` attributes, which a nonce cannot cover.
- `GET /api/v1/stats` - Detailed statistics with handler metrics
- `GET /api/v1/ip/{ip}` - Look up AS information for an IP address
- `GET /api/v1/as/{asn}` - Get prefixes announced by an AS
- `GET /api/v1/prefix/{prefix}` - Get routes for a specific prefix
An origin AS of 0 (`origin_asn` in a route, `asn` in an IP lookup, the key `0`
in a prefix's `origins`) means the route has no known origin: its AS path ends
@@ -145,9 +76,7 @@ routewatch/
│ ├── logger/ # Structured logging wrapper
│ ├── metrics/ # Performance metrics tracking
│ ├── config/ # Configuration management
│ ├── templates/ # HTML templates
│ ├── version/ # Build version information
│ └── whois/ # WHOIS lookups of AS details
│ └── templates/ # HTML templates
└── pkg/
└── asinfo/ # AS information lookup (public API)
```
@@ -166,7 +95,7 @@ routewatch/
│ │ Streamer │───▶│ Handlers │───▶│ Database │ │
│ │ │ │ │ │ │ │
│ │ RIS Live │ │ - ASHandler │ │ SQLite with │ │
│ │ HTTP stream │ │ - PeerHandler│ │ WAL mode │ │
│ │ WebSocket │ │ - PeerHandler│ │ WAL mode │ │
│ │ client │ │ - PrefixHdlr │ │ │ │
│ │ │ │ - PeeringHdlr│ │ Tables: │ │
│ └──────────────┘ └──────────────┘ │ - asns │ │
@@ -202,17 +131,13 @@ routewatch/
- Parses BGP UPDATE messages from JSON stream
- Dispatches to registered handlers based on message type
- Each handler has its own queue with backpressure handling
- Handlers batch writes for efficiency: the AS, peer and prefix handlers
write 30,000, 10,000 and 25,000 updates at a time, or after 2, 2 and 1
seconds; the peering handler writes the peerings it has collected every 30
seconds
- Handlers batch writes for efficiency (25K-30K ops, 1-2s timeout)
4. **Handler Details**
- **ASHandler**: Tracks all ASNs seen in AS paths
- **PeerHandler**: Records BGP peer information
- **PrefixHandler**: Maintains live routing table (upserts on announcement,
deletes on withdrawal or on an announcement whose AS path is empty or
holds AS 0 anywhere, deletes all of a peer's routes when its BGP session
deletes on withdrawal, deletes all of a peer's routes when its BGP session
goes down). Routes are kept per peer IP, so when one peer IP has sessions
on two collectors, a session going down on one also removes the routes the
other still carries, until they are announced again.
@@ -225,15 +150,10 @@ routewatch/
### Key Design Patterns
- **Batched Writes**: The handlers write the feed's updates to the database in
batches for performance
- **Batched Writes**: All database operations are batched for performance
- **Backpressure**: Probabilistic message dropping when queues exceed 50%
capacity
- **Graceful Shutdown**: on SIGTERM or SIGINT the streamer stops, the handlers
write the batches they hold, the HTTP server stops and the database is
checkpointed and closed. The daemon gives this 8 seconds and exits with status
0; "Running under upaas" says what a stop loses, logs and leaves undone when
the 8 seconds run out.
- **Graceful Shutdown**: 60-second timeout, flushes all pending batches
- **Reconnection**: Any end of the stream other than shutdown, including RIS
Live ending the response normally, marks the feed disconnected and reconnects
with exponential backoff (5s-320s). The backoff resets after a stream that
@@ -249,35 +169,21 @@ routewatch/
```sql
-- Core tables
asns(asn PRIMARY KEY, handle, description,
as_name, org_name, org_id, address, country_code, abuse_email,
abuse_phone, tech_email, tech_phone, rir, rir_registration_date,
rir_last_modified, whois_raw,
first_seen, last_seen, whois_updated_at)
prefixes_v4(prefix PRIMARY KEY, first_seen) WITHOUT ROWID
prefixes_v6(prefix PRIMARY KEY, first_seen) WITHOUT ROWID
asns(id, number, handle, description, first_seen, last_seen)
prefixes_v4(id, prefix, mask_length, first_seen, last_seen)
prefixes_v6(id, prefix, mask_length, first_seen, last_seen)
-- Live routing tables (one per IP version), one route per peer and prefix
live_routes_v4(id INTEGER PRIMARY KEY, prefix, mask_length, origin_asn,
peer_ip, as_path, next_hop, last_updated,
UNIQUE(peer_ip, prefix))
live_routes_v6(id INTEGER PRIMARY KEY, prefix, mask_length, origin_asn,
peer_ip, as_path, next_hop, last_updated,
UNIQUE(peer_ip, prefix))
-- Live routing tables (one per IP version)
live_routes_v4(id, prefix, mask_length, origin_asn, peer_ip, as_path,
next_hop, last_updated)
live_routes_v6(id, prefix, mask_length, origin_asn, peer_ip, as_path,
next_hop, last_updated)
-- Relationship tracking
peerings(id PRIMARY KEY, as_a, as_b, first_seen, last_seen,
UNIQUE(as_a, as_b))
bgp_peers(id PRIMARY KEY, peer_ip UNIQUE, peer_asn, first_seen, last_seen,
last_message_type)
-- Nothing in the daemon writes this table
announcements(id PRIMARY KEY, prefix_id, peer_asn, origin_asn, path,
next_hop, timestamp, is_withdrawal)
peerings(id, as_a, as_b, first_seen, last_seen)
bgp_peers(id, peer_ip, peer_asn, last_message_type, last_seen)
```
The indexes are in `internal/database/schema.sql`.
## Configuration
Configuration is handled via environment variables and OS-specific paths:
@@ -289,23 +195,13 @@ Configuration is handled via environment variables and OS-specific paths:
| `XDG_DATA_HOME` | `/var/lib` (in the Docker image) | Base of the state directory; must be an absolute path |
| `GOMEMLIMIT` | `1536MiB` (in the Docker image) | Go soft memory limit; see Memory |
| `MALLOC_ARENA_MAX` | `2` (in the Docker image) | glibc malloc arena cap, a positive whole number; see Memory |
| `TRUSTED_PROXIES` | (empty) | Addresses and CIDR ranges of trusted proxies; see below |
A variable that is set to an invalid value stops the start with an error and a
non-zero exit. An empty variable counts as unset.
`TRUSTED_PROXIES` lists the reverse proxies whose `X-Forwarded-For` the server
believes, comma-separated, such as `10.0.0.0/8,192.0.2.1`. For a request from
one of them, the request log and `/ip/` use the last address in
`X-Forwarded-For` that is not a trusted proxy: each proxy adds the address it
received the request from to the end, so the entries before it are whatever the
client sent. For a request from any other address they use the address the
request came from, whatever forwarding headers it carries. `X-Real-IP` and
`True-Client-IP` are not read.
State directory (database location):
- macOS: `~/Library/Application Support/berlin.sneak.app.routewatch/`
- macOS: `~/Library/Application Support/routewatch/`
- Linux: `/var/lib/berlin.sneak.app.routewatch/` when running as root, otherwise
`$XDG_DATA_HOME/berlin.sneak.app.routewatch/` (with `XDG_DATA_HOME` unset,
`~/.local/share/berlin.sneak.app.routewatch/`). In the Docker image this is
@@ -370,14 +266,8 @@ What happens at each limit:
(`peerBatchMaxSize`, about 10 MiB) and 250,000 for the prefix handler
(`prefixBatchMaxSize`, about 60 MiB). At that size the handler takes nothing
more off its queue until the write ends, so the queue fills and the streamer
drops messages as above. When the delete of a peer's live routes after its
session went down fails, the prefix handler keeps the updates the peer sent
after the down for the next write, which retries the delete first; on top of
its 250,000 it keeps at most the last 25,000 of them (`prefixBatchSize`, about
6 MiB) and drops and logs the older ones, so while the delete keeps failing
the other peers' updates still come off the queue. With the batches being
written, the three handlers' batches take at most about 0.25 GiB of the Go
heap's 1.5 GiB.
drops messages as above. With the batches being written, the three handlers'
batches take at most about 0.25 GiB of the Go heap's 1.5 GiB.
With `DEBUG=routewatch` the daemon logs a `System stats` line every 60 seconds
with the goroutine count and Go memory figures.
@@ -388,36 +278,15 @@ What the [upaas](https://git.eeqj.de/sneak/upaas) app needs:
- Container port: `8080`.
- Volume: one, at container path `/var/lib/berlin.sneak.app.routewatch`.
- Environment: `TRUSTED_PROXIES`, set to the address the reverse proxy in front
of the app connects from, as the container sees it; without it the request log
and `/ip/` give the proxy's address for every request. Nothing else is needed:
leave `XDG_DATA_HOME`, `GOMEMLIMIT` and `MALLOC_ARENA_MAX` at the image's
values. `DEBUG=routewatch` is optional and adds the `System stats` memory line
to the log.
- Environment: nothing is required. Leave `XDG_DATA_HOME`, `GOMEMLIMIT` and
`MALLOC_ARENA_MAX` at the image's values. `DEBUG=routewatch` is optional and
adds the `System stats` memory line to the log.
- Memory Limit: `5g`, the 5 GiB limit from Memory above. upaas sets no swap
limit, so on a host with swap Docker allows the same amount of swap again.
- Health check: the image's `HEALTHCHECK` requests
`/.well-known/healthcheck.json` on the container port. upaas reads the
container's health 60 seconds after a deploy and fails the deploy unless it is
`healthy`.
- Stop grace: the container needs 9 seconds between SIGTERM and SIGKILL. upaas
gives 10 (`stopTimeoutSeconds` in its `internal/docker/client.go`), as
`docker stop` does by default. The daemon allows 8 seconds for the handlers to
write what they hold and for the requests in progress to finish, then stops
the HTTP server, closes the database and exits with status 0. A longer grace
changes nothing. What is not written after 8 seconds is lost, and the streamer
and each handler log how much. When the prefix handler has fallen behind the
feed, a stop loses far more: its batch is full at 250,000 updates and the
streamer's worker for it waits for room, so the streamer's stop takes the
whole 8 seconds and every handler then skips its last write. Lost then are the
messages left on the four handlers' queues, what the AS, peer and peering
handlers hold, the prefix handler's batch, and the write it has in progress,
up to another 250,000 updates, which take about 20 seconds and are cut off by
the exit; withdrawals and peer downs are among them. A route withdrawn in them
stays in the live route table until its peer announces or withdraws the prefix
again or its session goes down. When the 8 seconds run out the database is not
closed, and SQLite rolls back a write the exit cut off when the database next
opens.
## Development
@@ -432,7 +301,7 @@ make fmt
make lint
# Build
make build
make
```
## Entrypoints
@@ -461,8 +330,8 @@ provide:
- `script/docker` — build the Docker image tagged via `script/projectname`
- `script/cibuild` — CI entrypoint: runs `script/bootstrap` and `script/check`,
then builds the image
- `script/precommit` — pre-commit gate: `go mod tidy` (must not change `go.mod`
or `go.sum`) and `go fmt` (writes), then `script/check`
- `script/precommit` — pre-commit gate: `go mod tidy` + `go fmt` (must not
change files), then `script/check`
- `script/install-precommit` — install the git pre-commit hook that runs
`script/precommit`
TODO.md
+14 -75
View File
@@ -10,91 +10,28 @@
# Status
pre-1.0. No git tags. `main` last took `next` on 2026-10-06, through
https://git.eeqj.de/sneak/routewatch/pulls/6. The linter and the tests are
phases of the Docker build, and the Gitea workflow runs `script/cibuild` (the
format check of the Go code and the markdown, both phases, then the image) on
every push. The image sets memory ceilings for a 5 GiB container (README
"Memory") and the README says how to run it under upaas (README "Running under
upaas"). A 25-hour run of `a9b6860` on the live feed peaked at about 0.6 GiB
(https://git.eeqj.de/sneak/routewatch/issues/3). A run of `825afa2` on the live
feed found the prefix handler dropping about a quarter of its updates because
its route writes were slower than the feed
(https://git.eeqj.de/sneak/routewatch/issues/96); the cheaper writes since then
are measured by a benchmark, not yet on the live feed.
pre-1.0. No git tags. The linter and the tests are phases of the Docker build,
and the Gitea workflow runs `script/cibuild` (the format check of the Go code
and the markdown, both phases, then the image) on every push. The image sets
memory ceilings for a 5 GiB container (README "Memory") and the README says how
to run it under upaas (README "Running under upaas"). A 35-hour run of `3898daa`
on the live feed peaked at about 1 GiB, without a container memory limit.
# Next Step
Run `next` on the live feed, as README "Running under upaas" describes, for the
final check https://git.eeqj.de/sneak/routewatch/issues/96 names: whether the
prefix handler still drops updates after the first hour. If it does, a new issue
follows with the numbers. Then an audit of the code at the commit that run
tested.
`next` waits for sneak to merge it to `main` through
https://git.eeqj.de/sneak/routewatch/pulls/6. After that, setting routewatch up
under upaas on fsn1app1 and deploying it are his
(https://git.eeqj.de/sneak/routewatch/issues/31), and so is the run under a real
5 GiB limit (https://git.eeqj.de/sneak/routewatch/issues/3).
# Completed Steps
- 2026-10-19: README "HTTP Endpoints" lists every route with what it answers,
"Database Schema" matches `schema.sql`, and the rest of the README matches the
code; this file brought up to date (closes #72)
- 2026-10-19: route writes are cheaper: the live route and prefix tables lose
their random UUID keys and the prefix tables their duplicate index, a route
update sets only the columns that change, and every write goes through one
connection with a page cache of 1 GiB (README "Memory"). A benchmark writes
the prefix handler's largest batch faster than the feed sends it (closes #96)
- 2026-10-19: the WAL checkpoint, the incremental vacuum and ANALYZE take the
database lock every batch write holds, so a vacuum no longer fails with
"database is locked" and an ANALYZE no longer makes a batch write lose its
updates; ANALYZE runs as `PRAGMA optimize` with an analysis limit (closes #94)
- 2026-10-19: shutdown stops the streamer first, and each handler writes
everything left on its queue before it stops, so the updates waiting there are
no longer lost (closes #69)
- 2026-10-19: a path that ends in an AS set of several members has origin AS 0,
no known origin, instead of the set's last member, and the set adds no
peerings; the pages show such a route as having no known origin (closes #75)
- 2026-10-19: `bgp_peer_count` drops to 0 when the last peer goes down (closes
#92)
- 2026-10-19: `bgp_peer_count` counts every peer the feed has heard from, not
only those whose session opened since startup, and drops a peer whose session
goes down (closes #74)
- 2026-10-19: the shared files are re-vendored from `sneak/prompts` at
`0653579`: golangci-lint v2.14.0 and its configuration, lint and test as
`Dockerfile` phases that keep Go's and golangci-lint's caches, markdown
formatted with prettier, agent guidance in `AGENTS.md`; the code is fixed for
what the new linter configuration finds (closes #52)
- 2026-10-19: the streamer reads feed lines up to 1 MiB, so a line over 64 KiB
no longer ends the connection (closes #76)
- 2026-10-19: a peer's live routes are removed when its BGP session goes down;
`STATE` messages are never dropped for backpressure, and one that finds its
queue full waits for room ahead of the handler's other messages (closes #56)
- 2026-10-19: the prefix-length pages pick their sample with a bounded number of
index lookups instead of reading every route of that length, and list each
prefix once (closes #66)
- 2026-10-19: a request past its time limit answers 504 instead of 408, and
`/api/v1/stats` answers 504 instead of 500 when its read runs out of time
(closes #70)
- 2026-10-19: the AS, peer and prefix handlers write each batch outside the lock
their message handling needs, so messages keep coming off their queues during
a write; meanwhile the next batch grows to at most ten batch sizes (closes
#65)
- 2026-10-19: a live route is identified by prefix and peer, so an announcement
with a new origin AS replaces the peer's route and a withdrawal removes it
whatever its origin AS (closes #55)
- 2026-10-14: the streamer reconnects when RIS Live ends the stream or sends
nothing for 60 seconds, and the health check reports the feed disconnected
meanwhile (closes #53)
- 2026-10-14: the prefix page takes the names of the origin ASes from the prefix
query instead of reading every route of each origin AS (closes #67)
- 2026-10-14: the daemon exits with status 1 when its web server cannot listen
on its port, instead of running on without it (closes #73)
- 2026-10-14: the prefix pages parse the prefix before looking it up: one that
does not parse answers 400, and any valid way of writing a prefix finds its
routes (closes #71)
- 2026-10-14: the IP lookup ignores routes shorter than /8 for IPv4 and /16 for
IPv6, so a default route is no longer named as an address's owner (closes #57)
- 2026-10-14: a prefix batch keeps each peer's last update for a prefix, instead
of one update per prefix across all peers (closes #54)
- 2026-10-14: `reconnect_count` in `/api/v1/stats` counts every reconnect
(closes #68)
- 2026-10-03: `/api/v1/stats` serves the prefix distribution from memory, seeded
at startup and adjusted on every live-route write, so a request no longer
reads every live route (closes #30)
@@ -179,4 +116,6 @@ tested.
# Future Steps
None open.
- Production memory under 5 GiB: whether to test under a real 5 GiB container
limit on fsn1app1 is open for sneak
(https://git.eeqj.de/sneak/routewatch/issues/3)
cmd/streamdumper/main.go
+1 -1
View File
@@ -35,7 +35,7 @@ func main() {
if err != nil {
log.Fatal("Failed to start streamer:", err)
}
defer s.Stop(context.Background())
defer s.Stop()
// Set up context with cancellation
ctx, cancel := context.WithCancel(context.Background())
go.mod
+1 -1
View File
@@ -1,6 +1,6 @@
module git.eeqj.de/sneak/routewatch
go 1.27.1
go 1.24.4
require (
github.com/dustin/go-humanize v1.0.1
internal/config/config.go
+2 -48
View File
@@ -4,12 +4,10 @@ package config
import (
"errors"
"fmt"
"net/netip"
"os"
"path/filepath"
"runtime"
"strconv"
"strings"
"time"
)
@@ -34,8 +32,6 @@ var (
errRelativeXDGDataHome = errors.New("XDG_DATA_HOME must be an absolute path")
errUnsupportedOS = errors.New("unsupported operating system")
errInvalidPort = errors.New("PORT must be a whole number")
errInvalidProxy = errors.New(
"TRUSTED_PROXIES must list IP addresses and CIDR ranges, comma-separated")
)
// Config holds configuration for the entire application
@@ -46,10 +42,6 @@ type Config struct {
// Port is the TCP port the HTTP server listens on
Port int
// TrustedProxies are the reverse proxies whose X-Forwarded-For header
// the HTTP server believes
TrustedProxies []netip.Prefix
// MaxRuntime is the maximum runtime (0 = run forever)
MaxRuntime time.Duration
@@ -76,15 +68,9 @@ func New() (*Config, error) {
return nil, err
}
trustedProxies, err := getTrustedProxies()
if err != nil {
return nil, err
}
return &Config{
StateDir: stateDir,
Port: port,
TrustedProxies: trustedProxies,
StateDir: stateDir,
Port: port,
// Run forever by default
MaxRuntime: 0,
// Enable batching by default
@@ -159,38 +145,6 @@ func getPort() (int, error) {
return int(port), nil
}
// getTrustedProxies returns the addresses and CIDR ranges listed in
// TRUSTED_PROXIES, such as "10.0.0.0/8,192.0.2.1", or none when it is not set.
// An address is returned as a range holding only that address.
func getTrustedProxies() ([]netip.Prefix, error) {
value := os.Getenv("TRUSTED_PROXIES")
if value == "" {
return nil, nil
}
var proxies []netip.Prefix
for entry := range strings.SplitSeq(value, ",") {
entry = strings.TrimSpace(entry)
addr, err := netip.ParseAddr(entry)
if err == nil {
proxies = append(proxies, netip.PrefixFrom(addr, addr.BitLen()))
continue
}
prefix, err := netip.ParsePrefix(entry)
if err != nil {
return nil, fmt.Errorf("%w, got %q", errInvalidProxy, entry)
}
proxies = append(proxies, prefix)
}
return proxies, nil
}
// EnsureDirectories creates all necessary directories if they don't exist
func (c *Config) EnsureDirectories() error {
// Ensure state directory exists
internal/config/config_internal_test.go
-51
View File
@@ -1,9 +1,7 @@
package config
import (
"net/netip"
"runtime"
"slices"
"testing"
)
@@ -46,55 +44,6 @@ func TestNewRefusesInvalidPort(t *testing.T) {
}
}
func TestNewReadsTrustedProxies(t *testing.T) {
t.Setenv("PORT", "")
t.Setenv("XDG_DATA_HOME", "")
t.Setenv("TRUSTED_PROXIES", "")
cfg, err := New()
if err != nil {
t.Fatalf("New() without TRUSTED_PROXIES: %v", err)
}
if len(cfg.TrustedProxies) != 0 {
t.Errorf("New() without TRUSTED_PROXIES: TrustedProxies = %v, want none",
cfg.TrustedProxies)
}
t.Setenv("TRUSTED_PROXIES", "10.0.0.0/8, 192.0.2.1,2001:db8::/32")
cfg, err = New()
if err != nil {
t.Fatalf("New() with TRUSTED_PROXIES: %v", err)
}
want := []netip.Prefix{
netip.MustParsePrefix("10.0.0.0/8"),
netip.MustParsePrefix("192.0.2.1/32"),
netip.MustParsePrefix("2001:db8::/32"),
}
if !slices.Equal(cfg.TrustedProxies, want) {
t.Errorf("TrustedProxies = %v, want %v", cfg.TrustedProxies, want)
}
}
func TestNewRefusesInvalidTrustedProxies(t *testing.T) {
invalid := []string{"proxy.example", "10.0.0.0/33", "192.0.2.1:80", "192.0.2.1,"}
for _, value := range invalid {
t.Run(value, func(t *testing.T) {
t.Setenv("PORT", "")
t.Setenv("XDG_DATA_HOME", "")
t.Setenv("TRUSTED_PROXIES", value)
_, err := New()
if err == nil {
t.Errorf("New() with TRUSTED_PROXIES=%q returned no error", value)
}
})
}
}
func TestNewRefusesRelativeXDGDataHome(t *testing.T) {
if runtime.GOOS == "darwin" {
t.Skip("macOS does not read XDG_DATA_HOME")
internal/database/counts_internal_test.go
+4 -128
View File
@@ -2,7 +2,6 @@ package database
import (
"context"
"fmt"
"slices"
"sync"
"testing"
@@ -148,9 +147,7 @@ func TestLiveCountsTrackWritesInRealtime(t *testing.T) {
func writeASNsPrefixesAndPeers(t *testing.T, db *Database, ts time.Time) {
t.Helper()
seen := SeenTimes{First: ts, Last: ts}
err := db.GetOrCreateASNBatch(map[int]SeenTimes{64500: seen, 64501: seen})
err := db.GetOrCreateASNBatch(map[int]time.Time{64500: ts, 64501: ts})
if err != nil {
t.Fatalf("GetOrCreateASNBatch: %v", err)
}
@@ -166,14 +163,14 @@ func writeASNsPrefixesAndPeers(t *testing.T, db *Database, ts time.Time) {
err = db.UpdatePeerBatch(map[string]PeerUpdate{
testV4PeerIP: {
PeerIP: testV4PeerIP, PeerASN: 64500,
MessageType: "UPDATE", Seen: seen,
MessageType: "UPDATE", Timestamp: ts,
},
})
if err != nil {
t.Fatalf("UpdatePeerBatch: %v", err)
}
err = db.RecordPeering(64500, 64501, seen)
err = db.RecordPeering(64500, 64501, ts)
if err != nil {
t.Fatalf("RecordPeering: %v", err)
}
@@ -194,10 +191,7 @@ func TestLiveCountsSeededFromDatabaseAtStartup(t *testing.T) {
ts := time.Date(2026, 1, 2, 3, 4, 5, 0, time.UTC)
seen := SeenTimes{First: ts, Last: ts}
err = db.GetOrCreateASNBatch(
map[int]SeenTimes{64500: seen, 64501: seen, 64502: seen})
err = db.GetOrCreateASNBatch(map[int]time.Time{64500: ts, 64501: ts, 64502: ts})
if err != nil {
t.Fatalf("GetOrCreateASNBatch: %v", err)
}
@@ -500,124 +494,6 @@ func TestDeleteLiveRoutesOfPeer(t *testing.T) {
checkOneLiveRoutePerIPVersion(t, db)
}
// TestDeleteLiveRoutesOfPeerDeletesInChunks gives a peer one IPv4 route more
// than a chunk holds, two of them for prefixes another peer also has. One chunk
// deletes peerRouteChunkSize of the peer's routes and commits, leaving the last
// one. The routes are written again, and DeleteLiveRoutesOfPeer deletes all of
// them, a chunk and then the last one. After each step the route count and
// prefix distribution held in memory and those read from the route table are
// the ones wanted.
func TestDeleteLiveRoutesOfPeerDeletesInChunks(t *testing.T) {
t.Parallel()
const (
peerIP = "192.0.2.1"
otherPeerIP = "192.0.2.2"
// How many of the peer's prefixes the other peer also has, the first
// ones in prefix order.
sharedPrefixes = 2
)
db, err := New(&config.Config{StateDir: t.TempDir()}, logger.New())
if err != nil {
t.Fatalf("failed to create database: %v", err)
}
defer func() { _ = db.Close() }()
ts := time.Date(2026, 1, 2, 3, 4, 5, 0, time.UTC)
peerRoutes := make([]*LiveRoute, 0, peerRouteChunkSize+1)
otherPeerRoutes := make([]*LiveRoute, 0, sharedPrefixes)
for i := range peerRouteChunkSize + 1 {
prefix := fmt.Sprintf("10.%d.%d.0/24", i/256, i%256)
route := mkV4Route(t, prefix, 64500, ts)
route.PeerIP = peerIP
peerRoutes = append(peerRoutes, route)
if i < sharedPrefixes {
route := mkV4Route(t, prefix, 64500, ts)
route.PeerIP = otherPeerIP
otherPeerRoutes = append(otherPeerRoutes, route)
}
}
err = db.UpsertLiveRouteBatch(slices.Concat(peerRoutes, otherPeerRoutes))
if err != nil {
t.Fatalf("UpsertLiveRouteBatch: %v", err)
}
deletedV4, deletedV6, err := db.deletePeerRouteChunk(peerIP)
if err != nil {
t.Fatalf("deletePeerRouteChunk: %v", err)
}
if deletedV4 != peerRouteChunkSize || deletedV6 != 0 {
t.Errorf("deletePeerRouteChunk deleted (%d, %d) routes, want (%d, 0)",
deletedV4, deletedV6, peerRouteChunkSize)
}
// The shared prefixes went with the chunk, so each route left has a
// prefix of its own.
checkV4RoutesOfOwnPrefixes(t, db, "after one chunk", 1+sharedPrefixes)
err = db.UpsertLiveRouteBatch(peerRoutes)
if err != nil {
t.Fatalf("UpsertLiveRouteBatch: %v", err)
}
deleted, err := db.DeleteLiveRoutesOfPeer(peerIP)
if err != nil {
t.Fatalf("DeleteLiveRoutesOfPeer: %v", err)
}
if deleted != len(peerRoutes) {
t.Errorf("DeleteLiveRoutesOfPeer deleted %d routes, want %d",
deleted, len(peerRoutes))
}
checkV4RoutesOfOwnPrefixes(t, db, "after DeleteLiveRoutesOfPeer",
sharedPrefixes)
}
// checkV4RoutesOfOwnPrefixes checks that the database has wantRoutes IPv4 live
// routes, each for a /24 no other live route has, both as counted in memory and
// as read from the route table. when says which step of the test checks.
func checkV4RoutesOfOwnPrefixes(
t *testing.T, db *Database, when string, wantRoutes int,
) {
t.Helper()
ctx := context.Background()
stats, err := db.GetStatsContext(ctx)
if err != nil {
t.Fatalf("GetStatsContext: %v", err)
}
tableRoutes, _, err := db.GetLiveRouteCountsContext(ctx)
if err != nil {
t.Fatalf("GetLiveRouteCountsContext: %v", err)
}
if stats.IPv4Routes != wantRoutes || tableRoutes != wantRoutes {
t.Errorf("%s: IPv4 live routes: %d in memory, %d in the table, want %d",
when, stats.IPv4Routes, tableRoutes, wantRoutes)
}
queryDistribution, _, err := db.GetPrefixDistributionContext(ctx)
if err != nil {
t.Fatalf("GetPrefixDistributionContext: %v", err)
}
want := []PrefixDistribution{{MaskLength: 24, Count: wantRoutes}}
assertDistribution(t, when+": IPv4 distribution",
stats.IPv4PrefixDistribution, want)
assertDistribution(t, when+": IPv4 distribution query",
queryDistribution, want)
}
// checkOneLiveRoutePerIPVersion checks that the database has one IPv4 live
// route, for a /24, and one IPv6 live route, for a /32, both as counted in
// memory and as read from the route tables.
internal/database/database.go
+228 -388
View File
@@ -81,23 +81,6 @@ const (
sqliteHardHeapLimitBytes = 2147483648
)
// Settings that bound the WAL file's size once no read is open (see
// Checkpoint).
const (
// sqliteWALSizeLimitBytes (64 MiB) is the WAL file size past which
// Checkpoint runs a TRUNCATE checkpoint. It is also the
// journal_size_limit of the connection kept for writing: when SQLite
// starts the WAL again from the beginning, it cuts a larger WAL file down
// to this size.
sqliteWALSizeLimitBytes = 67108864
// truncateBusyTimeoutMs is the busy timeout of the TRUNCATE checkpoint:
// the longest it waits for open reads, and so the longest it keeps batch
// writes waiting beyond the copy itself.
truncateBusyTimeoutMs = 500
// walFileSuffix ends the name of the WAL file, next to the database file.
walFileSuffix = "-wal"
)
// Settings of the ANALYZE that Analyze runs.
const (
// analysisLimit is about how many rows of each index ANALYZE reads
@@ -129,12 +112,6 @@ var (
ErrNoRoute = errors.New("no route found")
// ErrNoStaleASN is returned when no ASN needs WHOIS refresh
ErrNoStaleASN = errors.New("no stale ASN found")
// ErrCheckpointIncomplete is returned by Checkpoint when it left part of
// the WAL uncopied, or left a WAL file over sqliteWALSizeLimitBytes
// untruncated, because reads still needed the WAL or SQLite reported the
// checkpoint busy. A later checkpoint that runs when no read is open
// finishes the work.
ErrCheckpointIncomplete = errors.New("WAL checkpoint incomplete")
errInvalidASN = errors.New("invalid ASN")
errSameASNInPeering = errors.New("cannot create peering with same ASN")
@@ -169,20 +146,6 @@ func New(cfg *config.Config, logger *logger.Logger) (*Database, error) {
return nil, fmt.Errorf("failed to create database directory: %w", err)
}
// The size of the WAL file the previous run left, which Initialize logs
// with the start-up checkpoint. It is read before the database is opened,
// because opening it writes a frame to the WAL. There is no WAL file when
// the database is new or the last run closed the database, which deletes
// it.
var walBytes int64
wal, err := os.Stat(dbPath + walFileSuffix)
if err == nil {
walBytes = wal.Size()
} else if !errors.Is(err, os.ErrNotExist) {
logger.Warn("Failed to read the WAL file's size on startup", "error", err)
}
// Per-connection SQLite settings go in the DSN so every pooled connection
// gets them, not just the one that runs the Initialize pragmas. _txlock=
// immediate makes every transaction take the write lock at BEGIN. Without it
@@ -231,19 +194,12 @@ func New(cfg *config.Config, logger *logger.Logger) (*Database, error) {
err)
}
_, err = writer.ExecContext(ctx,
fmt.Sprintf("PRAGMA journal_size_limit=%d", sqliteWALSizeLimitBytes))
if err != nil {
return nil, fmt.Errorf(
"failed to set the writing connection's WAL size limit: %w", err)
}
database := &Database{
db: db, writer: writer, logger: logger, path: dbPath,
counts: &liveCounts{},
}
err = database.Initialize(walBytes)
err = database.Initialize()
if err != nil {
return nil, fmt.Errorf("failed to initialize database: %w", err)
}
@@ -259,16 +215,8 @@ func New(cfg *config.Config, logger *logger.Logger) (*Database, error) {
return database, nil
}
// The messages Initialize logs the start-up WAL checkpoint with.
const (
startupCheckpointMessage = "WAL checkpoint on startup"
startupCheckpointIncompleteMessage = "WAL checkpoint on startup incomplete"
)
// Initialize creates the database schema if it doesn't exist. walBytes is the
// size of the WAL file the previous run left, which New reads before it opens
// the database.
func (d *Database) Initialize(walBytes int64) error {
// Initialize creates the database schema if it doesn't exist.
func (d *Database) Initialize() error {
// Set SQLite pragmas for performance. Per-connection settings (cache_size,
// synchronous, busy_timeout, journal_mode) live in the DSN, and New sets the
// writing connection's cache_size; temp_store is left at its default so
@@ -288,33 +236,19 @@ func (d *Database) Initialize(walBytes int64) error {
}
}
// The start-up checkpoint copies the WAL a previous run left into the
// database file and empties the WAL file; its frame counts then read 0,
// so walBytes is logged with them.
var busy, walFrames, checkpointedFrames int
// Run WAL checkpoint on startup to consolidate any existing WAL
var walPages, checkpointedPages, movedPages int
err := d.writer.QueryRowContext(context.Background(),
"PRAGMA wal_checkpoint(TRUNCATE)",
).Scan(&busy, &walFrames, &checkpointedFrames)
switch {
case err != nil:
).Scan(&walPages, &checkpointedPages, &movedPages)
if err != nil {
d.logger.Warn("Failed to checkpoint WAL on startup", "error", err)
case busy != 0:
// Another process using the database kept the checkpoint from
// copying or emptying all of the WAL.
d.logger.Warn(startupCheckpointIncompleteMessage,
"wal_bytes", walBytes,
"busy", busy,
"wal_frames", walFrames,
"checkpointed_frames", checkpointedFrames,
)
default:
d.logger.Info(startupCheckpointMessage,
"wal_bytes", walBytes,
"busy", busy,
"wal_frames", walFrames,
"checkpointed_frames", checkpointedFrames,
} else {
d.logger.Info("WAL checkpoint on startup",
"wal_pages", walPages,
"checkpointed", checkpointedPages,
"moved", movedPages,
)
}
@@ -724,71 +658,48 @@ func (d *Database) DeleteLiveRouteBatch(deletions []LiveRouteDeletion) error {
return nil
}
// peerRouteChunkSize is the most routes of a peer that one transaction of
// DeleteLiveRoutesOfPeer deletes from each route table. A peer can have a
// million routes; deleted in one transaction, they would hold the database lock
// for most of a minute.
const peerRouteChunkSize = 10000
// Deleting a chunk of a peer's live routes. The first query of each table
// selects the peer's first routes in prefix order, through the
// UNIQUE(peer_ip, prefix) key, each with its mask length and whether another
// peer has a live route for its prefix: the delete takes the prefixes no other
// peer has out of the prefix distribution. The second query deletes the peer's
// routes up to the last prefix the first one selected.
// Deleting all live routes of a peer. The first query of each table selects the
// mask length of each of the peer's routes whose prefix no other peer has a
// live route for: the delete takes those prefixes out of the prefix
// distribution.
const (
peerRouteChunkV4SQL = `SELECT r.prefix, r.mask_length,
EXISTS (SELECT 1 FROM live_routes_v4 AS o
WHERE o.prefix = r.prefix AND o.peer_ip != r.peer_ip)
peerOnlyRouteMaskLengthsV4SQL = `SELECT r.mask_length
FROM live_routes_v4 AS r WHERE r.peer_ip = ?
ORDER BY r.prefix LIMIT ?`
deletePeerRouteChunkV4SQL = `DELETE FROM live_routes_v4
WHERE peer_ip = ? AND prefix <= ?`
AND NOT EXISTS (SELECT 1 FROM live_routes_v4 AS o
WHERE o.prefix = r.prefix AND o.peer_ip != r.peer_ip)`
deletePeerLiveRoutesV4SQL = `DELETE FROM live_routes_v4 WHERE peer_ip = ?`
peerRouteChunkV6SQL = `SELECT r.prefix, r.mask_length,
EXISTS (SELECT 1 FROM live_routes_v6 AS o
WHERE o.prefix = r.prefix AND o.peer_ip != r.peer_ip)
peerOnlyRouteMaskLengthsV6SQL = `SELECT r.mask_length
FROM live_routes_v6 AS r WHERE r.peer_ip = ?
ORDER BY r.prefix LIMIT ?`
deletePeerRouteChunkV6SQL = `DELETE FROM live_routes_v6
WHERE peer_ip = ? AND prefix <= ?`
AND NOT EXISTS (SELECT 1 FROM live_routes_v6 AS o
WHERE o.prefix = r.prefix AND o.peer_ip != r.peer_ip)`
deletePeerLiveRoutesV6SQL = `DELETE FROM live_routes_v6 WHERE peer_ip = ?`
)
// deletePeerRoutes deletes the first peerRouteChunkSize live routes of peerIP,
// or as many as it has, from one table, given that table's two statements
// above. It returns how many routes it deleted and the mask length of each
// prefix it left with no live route.
// deletePeerRoutes deletes all live routes of peerIP from one table, given that
// table's two statements above. It returns how many routes it deleted and the
// mask length of each prefix it left with no live route.
func deletePeerRoutes(
ctx context.Context, tx *loggingTx, chunkSQL, deleteSQL, peerIP string,
ctx context.Context, tx *loggingTx,
onlyRouteMaskLengthsSQL, deleteSQL, peerIP string,
) (int64, []int, error) {
rows, err := tx.QueryContext(ctx, chunkSQL, peerIP, peerRouteChunkSize)
rows, err := tx.QueryContext(ctx, onlyRouteMaskLengthsSQL, peerIP)
if err != nil {
return 0, nil, err
}
defer func() { _ = rows.Close() }()
var (
selected int
lastPrefix string
goneMaskLengths []int
)
var goneMaskLengths []int
for rows.Next() {
var (
maskLength int
otherPeerHasARoute bool
)
var maskLength int
err := rows.Scan(&lastPrefix, &maskLength, &otherPeerHasARoute)
err := rows.Scan(&maskLength)
if err != nil {
return 0, nil, err
}
selected++
if !otherPeerHasARoute {
goneMaskLengths = append(goneMaskLengths, maskLength)
}
goneMaskLengths = append(goneMaskLengths, maskLength)
}
err = rows.Err()
@@ -796,11 +707,7 @@ func deletePeerRoutes(
return 0, nil, err
}
if selected == 0 {
return 0, nil, nil
}
res, err := tx.ExecContext(ctx, deleteSQL, peerIP, lastPrefix)
res, err := tx.ExecContext(ctx, deleteSQL, peerIP)
if err != nil {
return 0, nil, err
}
@@ -814,33 +721,79 @@ func deletePeerRoutes(
}
// DeleteLiveRoutesOfPeer deletes every live route of the peer, IPv4 and IPv6,
// as when the peer's BGP session goes down. It deletes them a chunk at a time
// with deletePeerRouteChunk, each chunk in a transaction of its own, and
// releases the database lock between chunks so other writes go on. It returns
// how many routes it deleted; after an error, the chunks it deleted before stay
// deleted.
// in one transaction, as when the peer's BGP session goes down. It returns how
// many routes it deleted.
func (d *Database) DeleteLiveRoutesOfPeer(peerIP string) (int, error) {
total := 0
d.lock("DeleteLiveRoutesOfPeer")
defer d.unlock()
for {
deletedV4, deletedV6, err := d.deletePeerRouteChunk(peerIP)
if err != nil {
return total, err
}
total += deletedV4 + deletedV6
if deletedV4 < peerRouteChunkSize && deletedV6 < peerRouteChunkSize {
return total, nil
}
tx, err := d.beginTx()
if err != nil {
return 0, fmt.Errorf("failed to begin transaction: %w", err)
}
defer d.rollback(tx)
ctx := context.Background()
deletedV4, goneMaskLengthsV4, err := deletePeerRoutes(ctx, tx,
peerOnlyRouteMaskLengthsV4SQL, deletePeerLiveRoutesV4SQL, peerIP)
if err != nil {
return 0, fmt.Errorf("failed to delete the IPv4 routes of peer %s: %w",
peerIP, err)
}
deletedV6, goneMaskLengthsV6, err := deletePeerRoutes(ctx, tx,
peerOnlyRouteMaskLengthsV6SQL, deletePeerLiveRoutesV6SQL, peerIP)
if err != nil {
return 0, fmt.Errorf("failed to delete the IPv6 routes of peer %s: %w",
peerIP, err)
}
err = tx.Commit()
if err != nil {
return 0, fmt.Errorf("failed to commit transaction: %w", err)
}
d.counts.addRoutes(-int(deletedV4), -int(deletedV6))
d.counts.addToDistribution(goneMaskLengthsV4, goneMaskLengthsV6, -1)
return int(deletedV4 + deletedV6), nil
}
// insertPrefixesInTable inserts the prefixes, all of one IP version, that the
// prefix table of that version does not hold yet, each first seen at its time
// in prefixes. A prefix already stored is left as it is. It returns how many it
// inserted.
func insertPrefixesInTable(
// upsertPrefix updates the last_seen time of prefix with the statements of its
// IP version's table, inserting the prefix when the update matched no row. It
// reports whether it inserted it.
func upsertPrefix(
ctx context.Context, updateStmt, insertStmt *sql.Stmt,
prefix string, timestamp time.Time,
) (bool, error) {
res, err := updateStmt.ExecContext(ctx, timestamp, prefix)
if err != nil {
return false, fmt.Errorf("failed to update prefix %s: %w", prefix, err)
}
affected, err := res.RowsAffected()
if err != nil {
return false, fmt.Errorf("failed to count updated prefix %s: %w",
prefix, err)
}
if affected > 0 {
return false, nil
}
_, err = insertStmt.ExecContext(ctx, prefix, timestamp, timestamp)
if err != nil {
return false, fmt.Errorf("failed to insert prefix %s: %w", prefix, err)
}
return true, nil
}
// upsertPrefixesInTable updates the last_seen time of prefixes, all of one IP
// version, in the prefix table of that version, inserting the ones that are
// new. It returns how many it inserted.
func upsertPrefixesInTable(
ctx context.Context, tx *loggingTx, prefixes map[string]time.Time,
ipVersion int,
) (int, error) {
@@ -849,15 +802,24 @@ func insertPrefixesInTable(
}
family := ipv4Family
insertSQL := "INSERT INTO prefixes_v4 (prefix, first_seen) VALUES (?, ?) " +
"ON CONFLICT (prefix) DO NOTHING"
updateSQL := "UPDATE prefixes_v4 SET last_seen = ? WHERE prefix = ?"
insertSQL := "INSERT INTO prefixes_v4 (prefix, first_seen, last_seen) " +
"VALUES (?, ?, ?)"
if ipVersion == ipVersionV6 {
family = ipv6Family
insertSQL = "INSERT INTO prefixes_v6 (prefix, first_seen) VALUES (?, ?) " +
"ON CONFLICT (prefix) DO NOTHING"
updateSQL = "UPDATE prefixes_v6 SET last_seen = ? WHERE prefix = ?"
insertSQL = "INSERT INTO prefixes_v6 (prefix, first_seen, last_seen) " +
"VALUES (?, ?, ?)"
}
updateStmt, err := tx.PrepareContext(ctx, updateSQL)
if err != nil {
return 0, fmt.Errorf("failed to prepare %s update statement: %w",
family, err)
}
defer func() { _ = updateStmt.Close() }()
insertStmt, err := tx.PrepareContext(ctx, insertSQL)
if err != nil {
return 0, fmt.Errorf("failed to prepare %s insert statement: %w",
@@ -869,19 +831,13 @@ func insertPrefixesInTable(
// The rows are written in key order, as UpsertLiveRouteBatch writes routes.
for _, prefix := range slices.Sorted(maps.Keys(prefixes)) {
res, err := insertStmt.ExecContext(ctx, prefix, prefixes[prefix])
inserted, err := upsertPrefix(ctx, updateStmt, insertStmt, prefix,
prefixes[prefix])
if err != nil {
return 0, fmt.Errorf("failed to insert prefix %s: %w", prefix, err)
return 0, err
}
// One row for a new prefix, none for one already stored.
inserted, err := res.RowsAffected()
if err != nil {
return 0, fmt.Errorf("failed to count inserted prefix %s: %w",
prefix, err)
}
if inserted > 0 {
if inserted {
newPrefixes++
}
}
@@ -889,9 +845,8 @@ func insertPrefixesInTable(
return newPrefixes, nil
}
// UpdatePrefixesBatch stores, in a single transaction, the prefixes that are
// not stored yet, each first seen at its time in prefixes: the IPv4 prefixes
// first and then the IPv6 prefixes. A prefix that does not parse is not stored.
// UpdatePrefixesBatch updates the last_seen time for multiple prefixes in a
// single transaction, the IPv4 prefixes first and then the IPv6 prefixes.
func (d *Database) UpdatePrefixesBatch(prefixes map[string]time.Time) error {
if len(prefixes) == 0 {
return nil
@@ -910,11 +865,6 @@ func (d *Database) UpdatePrefixesBatch(prefixes map[string]time.Time) error {
prefixesV6 := make(map[string]time.Time)
for prefix, timestamp := range prefixes {
_, err := netip.ParsePrefix(prefix)
if err != nil {
continue
}
if detectIPVersion(prefix) == ipVersionV4 {
prefixesV4[prefix] = timestamp
} else {
@@ -924,12 +874,12 @@ func (d *Database) UpdatePrefixesBatch(prefixes map[string]time.Time) error {
ctx := context.Background()
newV4, err := insertPrefixesInTable(ctx, tx, prefixesV4, ipVersionV4)
newV4, err := upsertPrefixesInTable(ctx, tx, prefixesV4, ipVersionV4)
if err != nil {
return err
}
newV6, err := insertPrefixesInTable(ctx, tx, prefixesV6, ipVersionV6)
newV6, err := upsertPrefixesInTable(ctx, tx, prefixesV6, ipVersionV6)
if err != nil {
return err
}
@@ -944,14 +894,13 @@ func (d *Database) UpdatePrefixesBatch(prefixes map[string]time.Time) error {
return nil
}
// newASN returns a new ASN record first seen at seen.First and last seen at
// seen.Last, with the handle and description from the AS info lookup when it
// knows the number.
func newASN(number int, seen SeenTimes) ASN {
// newASN returns a new ASN record first and last seen at timestamp, with the
// handle and description from the AS info lookup when it knows the number.
func newASN(number int, timestamp time.Time) ASN {
asn := ASN{
ASN: number,
FirstSeen: seen.First,
LastSeen: seen.Last,
FirstSeen: timestamp,
LastSeen: timestamp,
}
// Look up ASN info
@@ -963,12 +912,12 @@ func newASN(number int, seen SeenTimes) ASN {
return asn
}
// upsertASN sets the last_seen time of ASN number to seen.Last with the given
// upsertASN updates the last_seen time of ASN number with the given
// statements, inserting the ASN when it is new. It reports whether it inserted
// it.
func upsertASN(
ctx context.Context, selectStmt, updateStmt, insertStmt *sql.Stmt,
number int, seen SeenTimes,
number int, timestamp time.Time,
) (bool, error) {
var (
asn ASN
@@ -979,7 +928,7 @@ func upsertASN(
&description, &asn.FirstSeen, &asn.LastSeen)
if err == nil {
// ASN exists, update last_seen
_, err = updateStmt.ExecContext(ctx, seen.Last, number)
_, err = updateStmt.ExecContext(ctx, timestamp, number)
if err != nil {
return false, fmt.Errorf("failed to update ASN %d: %w", number, err)
}
@@ -992,7 +941,7 @@ func upsertASN(
}
// ASN doesn't exist, create it
asn = newASN(number, seen)
asn = newASN(number, timestamp)
_, err = insertStmt.ExecContext(ctx, asn.ASN, asn.Handle,
asn.Description, asn.FirstSeen, asn.LastSeen)
@@ -1004,7 +953,7 @@ func upsertASN(
}
// GetOrCreateASNBatch creates or updates multiple ASNs in a single transaction
func (d *Database) GetOrCreateASNBatch(asns map[int]SeenTimes) error {
func (d *Database) GetOrCreateASNBatch(asns map[int]time.Time) error {
if len(asns) == 0 {
return nil
}
@@ -1046,9 +995,9 @@ func (d *Database) GetOrCreateASNBatch(asns map[int]SeenTimes) error {
var newASNs int
for number, seen := range asns {
for number, timestamp := range asns {
inserted, err := upsertASN(ctx, selectStmt, updateStmt, insertStmt,
number, seen)
number, timestamp)
if err != nil {
return err
}
@@ -1117,7 +1066,7 @@ func (d *Database) GetOrCreateASN(number int, timestamp time.Time) (*ASN, error)
}
// ASN doesn't exist, create it with ASN info lookup
asn = newASN(number, SeenTimes{First: timestamp, Last: timestamp})
asn = newASN(number, timestamp)
_, err = tx.Exec("INSERT INTO asns "+
"(asn, handle, description, first_seen, last_seen) "+
@@ -1163,13 +1112,30 @@ func (d *Database) GetOrCreatePrefix(
var p Prefix
query := fmt.Sprintf("SELECT prefix, first_seen FROM %s WHERE prefix = ?",
tableName)
query := fmt.Sprintf("SELECT prefix, first_seen, last_seen "+
"FROM %s WHERE prefix = ?", tableName)
err = tx.QueryRow(query, prefix).Scan(&p.Prefix, &p.FirstSeen)
err = tx.QueryRow(query, prefix).Scan(&p.Prefix, &p.FirstSeen, &p.LastSeen)
if err == nil {
// Prefix exists; the deferred rollback ends the read.
// Prefix exists, update last_seen
p.IPVersion = ipVersion
updateQuery := fmt.Sprintf("UPDATE %s SET last_seen = ? WHERE prefix = ?",
tableName)
_, err = tx.Exec(updateQuery, timestamp, prefix)
if err != nil {
return nil, err
}
p.LastSeen = timestamp
err = tx.Commit()
if err != nil {
d.logger.Error("Failed to commit transaction for prefix update",
"prefix", prefix, "error", err)
return nil, err
}
return &p, nil
}
@@ -1179,11 +1145,12 @@ func (d *Database) GetOrCreatePrefix(
}
// Prefix doesn't exist, create it
p = Prefix{Prefix: prefix, IPVersion: ipVersion, FirstSeen: timestamp}
p = Prefix{Prefix: prefix, IPVersion: ipVersion,
FirstSeen: timestamp, LastSeen: timestamp}
insertQuery := fmt.Sprintf("INSERT INTO %s "+
"(prefix, first_seen) VALUES (?, ?)", tableName)
"(prefix, first_seen, last_seen) VALUES (?, ?, ?)", tableName)
_, err = tx.Exec(insertQuery, p.Prefix, p.FirstSeen)
_, err = tx.Exec(insertQuery, p.Prefix, p.FirstSeen, p.LastSeen)
if err != nil {
return nil, err
}
@@ -1224,7 +1191,7 @@ func (d *Database) RecordAnnouncement(announcement *Announcement) error {
}
// RecordPeering records a peering relationship between two ASNs.
func (d *Database) RecordPeering(asA, asB int, seen SeenTimes) error {
func (d *Database) RecordPeering(asA, asB int, timestamp time.Time) error {
// Validate ASNs
if asA <= 0 || asB <= 0 {
return fmt.Errorf("%w: asA=%d, asB=%d", errInvalidASN, asA, asB)
@@ -1260,12 +1227,12 @@ func (d *Database) RecordPeering(asA, asB int, seen SeenTimes) error {
if exists {
_, err = tx.Exec(
"UPDATE peerings SET last_seen = ? WHERE as_a = ? AND as_b = ?",
seen.Last, asA, asB)
timestamp, asA, asB)
} else {
_, err = tx.Exec(`
INSERT INTO peerings (id, as_a, as_b, first_seen, last_seen)
VALUES (?, ?, ?, ?, ?)`,
generateUUID().String(), asA, asB, seen.First, seen.Last)
generateUUID().String(), asA, asB, timestamp, timestamp)
}
if err != nil {
@@ -1343,11 +1310,11 @@ func (d *Database) UpdatePeerBatch(peers map[string]PeerUpdate) error {
if exists {
_, err = updateStmt.ExecContext(ctx, update.PeerASN,
update.Seen.Last, update.MessageType, update.PeerIP)
update.Timestamp, update.MessageType, update.PeerIP)
} else {
_, err = insertStmt.ExecContext(ctx,
generateUUID().String(), update.PeerIP, update.PeerASN,
update.Seen.First, update.Seen.Last, update.MessageType)
update.Timestamp, update.Timestamp, update.MessageType)
}
if err != nil {
@@ -1745,9 +1712,8 @@ func (d *Database) GetASDetails(asn int) (*ASN, []LiveRoute, error) {
return d.GetASDetailsContext(context.Background(), asn)
}
// GetASDetailsContext returns the AS and the prefixes it originates, the IPv4
// ones first, each once with the time of its newest route. It makes two index
// searches per prefix, however many peers carry the prefix.
// GetASDetailsContext returns detailed information about an AS including
// prefixes with context support
func (d *Database) GetASDetailsContext(
ctx context.Context, asn int,
) (*ASN, []LiveRoute, error) {
@@ -1774,12 +1740,27 @@ func (d *Database) GetASDetailsContext(
asnInfo.Handle = handle.String
asnInfo.Description = description.String
allPrefixes, err := d.originPrefixes(ctx, asn, ipVersionV4)
// Get prefixes announced by this AS from both tables
queryV4 := `
SELECT prefix, mask_length, MAX(last_updated) as last_updated
FROM live_routes_v4
WHERE origin_asn = ?
GROUP BY prefix, mask_length
`
allPrefixes, err := d.originPrefixes(ctx, queryV4, asn, ipVersionV4)
if err != nil {
return &asnInfo, nil, err
}
prefixesV6, err := d.originPrefixes(ctx, asn, ipVersionV6)
queryV6 := `
SELECT prefix, mask_length, MAX(last_updated) as last_updated
FROM live_routes_v6
WHERE origin_asn = ?
GROUP BY prefix, mask_length
`
prefixesV6, err := d.originPrefixes(ctx, queryV6, asn, ipVersionV6)
if err != nil {
return &asnInfo, allPrefixes, err
}
@@ -2313,24 +2294,9 @@ func (d *Database) Analyze(ctx context.Context) error {
return nil
}
// Checkpoint copies the WAL into the database file. It holds the database
// lock, so batch writes wait while it runs.
//
// It runs a PASSIVE checkpoint, which never waits: it copies what no open read
// still needs, and leaves the file as large as it is. When the file is over
// sqliteWALSizeLimitBytes, Checkpoint runs one TRUNCATE checkpoint instead,
// which waits at most truncateBusyTimeoutMs for the open reads to end, then
// copies all of the WAL and empties the file, or gives up. It is logged with
// how long it took and whether it emptied the file.
//
// So a file over the limit is emptied by the first Checkpoint that finds no
// read open, at its start or at some moment of its wait. While reads follow
// one another with no pause between them, every TRUNCATE checkpoint can give
// up after its full wait, and the file can keep growing for as long as they
// do.
//
// It returns ErrCheckpointIncomplete when it leaves part of the WAL uncopied,
// or a file over the limit not emptied.
// Checkpoint runs a WAL checkpoint to transfer data from the WAL to the main database.
// Uses TRUNCATE mode which blocks writers briefly but ensures complete checkpoint,
// keeping the WAL small for fast read performance. It holds the database lock.
func (d *Database) Checkpoint(ctx context.Context) error {
d.lock("Checkpoint")
defer d.unlock()
@@ -2340,32 +2306,11 @@ func (d *Database) Checkpoint(ctx context.Context) error {
return fmt.Errorf("failed to checkpoint WAL: %w", err)
}
wal, err := os.Stat(d.path + walFileSuffix)
if err != nil {
return fmt.Errorf("failed to read the WAL file's size: %w", err)
}
if wal.Size() > sqliteWALSizeLimitBytes {
return d.truncateWAL(ctx, wal.Size())
}
var busy, walPages, copiedPages int
err = d.writer.QueryRowContext(ctx, "PRAGMA wal_checkpoint(PASSIVE)").
Scan(&busy, &walPages, &copiedPages)
_, err = d.writer.ExecContext(ctx, "PRAGMA wal_checkpoint(TRUNCATE)")
if err != nil {
return fmt.Errorf("failed to checkpoint WAL: %w", err)
}
if busy != 0 {
return fmt.Errorf("%w: SQLite reported it busy", ErrCheckpointIncomplete)
}
if copiedPages < walPages {
return fmt.Errorf("%w: copied %d of %d WAL pages",
ErrCheckpointIncomplete, copiedPages, walPages)
}
return nil
}
@@ -2381,50 +2326,6 @@ func (d *Database) Ping(ctx context.Context) error {
return nil
}
// truncateWAL runs the TRUNCATE checkpoint of Checkpoint on a WAL file of
// walBytes, with the writing connection's busy timeout lowered to
// truncateBusyTimeoutMs for it, and logs how long it took. The caller holds the
// database lock.
func (d *Database) truncateWAL(ctx context.Context, walBytes int64) error {
_, err := d.writer.ExecContext(ctx,
fmt.Sprintf("PRAGMA busy_timeout=%d", truncateBusyTimeoutMs))
if err != nil {
return fmt.Errorf("failed to lower the busy timeout to truncate WAL: %w",
err)
}
started := time.Now()
var busy, walPages, copiedPages int
checkpointErr := d.writer.QueryRowContext(ctx,
"PRAGMA wal_checkpoint(TRUNCATE)").Scan(&busy, &walPages, &copiedPages)
took := time.Since(started)
// Restored even when ctx has ended, so that no batch write runs with the
// short timeout.
_, restoreErr := d.writer.ExecContext(context.WithoutCancel(ctx),
fmt.Sprintf("PRAGMA busy_timeout=%d", sqliteBusyTimeoutMs))
if checkpointErr != nil || restoreErr != nil {
return fmt.Errorf("failed to truncate WAL: %w",
errors.Join(checkpointErr, restoreErr))
}
if busy != 0 {
d.logger.Warn("WAL file over its size limit not truncated",
"wal_bytes", walBytes, "wal_pages", walPages,
"copied_pages", copiedPages, "duration", took)
return fmt.Errorf("%w: WAL file of %d bytes not truncated",
ErrCheckpointIncomplete, walBytes)
}
d.logger.Info("WAL file over its size limit truncated",
"wal_bytes", walBytes, "duration", took)
return nil
}
// The debug messages lock logs, with the operation's name, when it starts to
// wait for the database mutex and when it holds it.
const (
@@ -2480,46 +2381,6 @@ func (d *Database) rollback(tx *loggingTx) {
}
}
// deletePeerRouteChunk deletes, in one transaction, the first
// peerRouteChunkSize IPv4 and the first peerRouteChunkSize IPv6 live routes of
// the peer, or as many as it has. It returns how many of each it deleted.
func (d *Database) deletePeerRouteChunk(peerIP string) (int, int, error) {
d.lock("DeleteLiveRoutesOfPeer")
defer d.unlock()
tx, err := d.beginTx()
if err != nil {
return 0, 0, fmt.Errorf("failed to begin transaction: %w", err)
}
defer d.rollback(tx)
ctx := context.Background()
deletedV4, goneMaskLengthsV4, err := deletePeerRoutes(ctx, tx,
peerRouteChunkV4SQL, deletePeerRouteChunkV4SQL, peerIP)
if err != nil {
return 0, 0, fmt.Errorf("failed to delete the IPv4 routes of peer %s: %w",
peerIP, err)
}
deletedV6, goneMaskLengthsV6, err := deletePeerRoutes(ctx, tx,
peerRouteChunkV6SQL, deletePeerRouteChunkV6SQL, peerIP)
if err != nil {
return 0, 0, fmt.Errorf("failed to delete the IPv6 routes of peer %s: %w",
peerIP, err)
}
err = tx.Commit()
if err != nil {
return 0, 0, fmt.Errorf("failed to commit transaction: %w", err)
}
d.counts.addRoutes(-int(deletedV4), -int(deletedV6))
d.counts.addToDistribution(goneMaskLengthsV4, goneMaskLengthsV6, -1)
return int(deletedV4), int(deletedV6), nil
}
// routeTimestampRange returns the earliest and latest last_updated across both
// live route tables, or nil values when both tables are empty. Each query reads
// one row from an end of the last_updated index rather than scanning the tables.
@@ -2590,87 +2451,66 @@ func (d *Database) scanRouteTimestamp(
}
}
// The AS page lists the prefixes an AS originates one at a time, each in two
// index searches that read no table row: one of the origin_asn and prefix
// index for the AS's first prefix after the previous one, and one of the
// prefix and last_updated index for the time of its newest route. Neither
// reads more for a prefix that more peers carry.
const (
nextOriginPrefixV4SQL = `SELECT prefix FROM live_routes_v4
WHERE origin_asn = ? AND prefix > ? ORDER BY prefix LIMIT 1`
newestRouteTimeV4SQL = `SELECT last_updated FROM live_routes_v4
WHERE prefix = ? ORDER BY last_updated DESC LIMIT 1`
nextOriginPrefixV6SQL = `SELECT prefix FROM live_routes_v6
WHERE origin_asn = ? AND prefix > ? ORDER BY prefix LIMIT 1`
newestRouteTimeV6SQL = `SELECT last_updated FROM live_routes_v6
WHERE prefix = ? ORDER BY last_updated DESC LIMIT 1`
)
// originPrefixes returns every prefix asn originates in the live route table
// of ipVersion, in text order, each with its mask length and the time of the
// prefix's newest route, whichever AS that route names as origin. A prefix
// whose routes were all withdrawn between the two searches is left out.
// originPrefixes runs query, which selects the prefix, mask length and latest
// last_updated of every prefix asn originates in the live route table of
// ipVersion. Rows that fail to scan or whose timestamp does not parse are
// logged and skipped.
func (d *Database) originPrefixes(
ctx context.Context, asn, ipVersion int,
ctx context.Context, query string, asn, ipVersion int,
) ([]LiveRoute, error) {
family, nextSQL, newestSQL := ipv4Family, nextOriginPrefixV4SQL,
newestRouteTimeV4SQL
family := ipv4Family
if ipVersion == ipVersionV6 {
family, nextSQL, newestSQL = ipv6Family, nextOriginPrefixV6SQL,
newestRouteTimeV6SQL
family = ipv6Family
}
nextStmt, err := d.db.PrepareContext(ctx, nextSQL)
rows, err := d.db.QueryContext(ctx, query, asn)
if err != nil {
return nil, fmt.Errorf("failed to prepare %s prefix query: %w",
family, err)
return nil, fmt.Errorf("failed to query %s prefixes: %w", family, err)
}
defer func() { _ = nextStmt.Close() }()
newestStmt, err := d.db.PrepareContext(ctx, newestSQL)
if err != nil {
return nil, fmt.Errorf("failed to prepare %s route time query: %w",
family, err)
}
defer func() { _ = newestStmt.Close() }()
defer func() { _ = rows.Close() }()
var routes []LiveRoute
previous := "" // every prefix sorts after the empty string
for {
var prefix string
err := nextStmt.QueryRowContext(ctx, asn, previous).Scan(&prefix)
if errors.Is(err, sql.ErrNoRows) {
return routes, nil
}
for rows.Next() {
var (
route LiveRoute
lastUpdatedStr string
)
err := rows.Scan(&route.Prefix, &route.MaskLength, &lastUpdatedStr)
if err != nil {
return nil, fmt.Errorf("failed to list %s prefixes: %w", family, err)
}
d.logger.Error("Failed to scan "+family+" prefix row",
"error", err, "asn", asn)
previous = prefix
route := LiveRoute{Prefix: prefix, IPVersion: ipVersion, OriginASN: asn}
err = newestStmt.QueryRowContext(ctx, prefix).Scan(&route.LastUpdated)
if errors.Is(err, sql.ErrNoRows) {
continue
}
// Parse the timestamp string
route.LastUpdated, err = time.Parse("2006-01-02 15:04:05-07:00",
lastUpdatedStr)
if err != nil {
return nil, fmt.Errorf("failed to query route time of %s: %w",
prefix, err)
}
// Try without timezone
route.LastUpdated, err = time.Parse("2006-01-02 15:04:05",
lastUpdatedStr)
if err != nil {
d.logger.Error("Failed to parse timestamp",
"error", err, "timestamp", lastUpdatedStr)
route.MaskLength, err = prefixMaskLength(prefix)
if err != nil {
return nil, err
continue
}
}
route.OriginASN = asn
route.IPVersion = ipVersion
routes = append(routes, route)
}
err = rows.Err()
if err != nil {
return nil, fmt.Errorf("failed to read %s prefixes: %w", family, err)
}
return routes, nil
}
// firstPrefixesByLength returns the first n prefixes, in text order, with the
internal/database/database_internal_test.go
+24 -683
View File
@@ -1,19 +1,15 @@
package database
import (
"bytes"
"context"
"database/sql"
"encoding/json"
"errors"
"fmt"
"log/slog"
"math/rand/v2"
"net/netip"
"os"
"slices"
"strconv"
"strings"
"sync"
"testing"
"time"
@@ -71,15 +67,6 @@ const (
lockWaitTimeout = time.Millisecond
// lockTestASN is the AS the batch write in the lock tests stores.
lockTestASN = 64500
// readerCheckpointLimit is how long a checkpoint, and a batch write
// waiting for it, may take while a reader is open: well under the busy
// timeout, which a checkpoint that waits for the reader takes.
readerCheckpointLimit = time.Second
// truncateCheckpointLimit is how long a checkpoint of a WAL file over
// sqliteWALSizeLimitBytes may take to give up while a reader is open: its
// wait for the reader plus a second, well under the busy timeout.
truncateCheckpointLimit = truncateBusyTimeoutMs*time.Millisecond +
time.Second
)
// Parameters for the prefix-length sample tests.
@@ -470,33 +457,6 @@ func TestWritingConnectionCache(t *testing.T) {
}
}
// TestWritingConnectionWALSizeLimit checks that the connection kept for
// writing, the one that writes the WAL, has the WAL size limit, which cuts the
// WAL file back when SQLite starts the WAL again from the beginning.
func TestWritingConnectionWALSizeLimit(t *testing.T) {
t.Parallel()
db, err := New(&config.Config{StateDir: t.TempDir()}, logger.New())
if err != nil {
t.Fatalf("failed to create database: %v", err)
}
defer func() { _ = db.Close() }()
var limit int64
err = db.writer.QueryRowContext(t.Context(), "PRAGMA journal_size_limit").
Scan(&limit)
if err != nil {
t.Fatalf("failed to read the writing connection's journal_size_limit: %v",
err)
}
if limit != sqliteWALSizeLimitBytes {
t.Errorf("writing connection: journal_size_limit = %d, want %d",
limit, sqliteWALSizeLimitBytes)
}
}
// readPragma reads the integer value of PRAGMA name on connection i, c.
func readPragma(
ctx context.Context, t *testing.T, c *sql.Conn, i int, name string,
@@ -604,8 +564,11 @@ func TestBatchWriteDuringCheckpoint(t *testing.T) {
ctx, cancel := context.WithCancel(context.Background())
var wg sync.WaitGroup
wg.Add(1)
go func() {
defer wg.Done()
wg.Go(func() {
for {
select {
case <-ctx.Done():
@@ -614,15 +577,13 @@ func TestBatchWriteDuringCheckpoint(t *testing.T) {
_ = db.Checkpoint(ctx) // errors are the checkpoint's own to absorb
}
}
})
}()
ts := time.Now().UTC()
seen := SeenTimes{First: ts, Last: ts}
for i := range contentionIterations {
asns := map[int]SeenTimes{
i % contendedASNCount: seen,
(i % contendedASNCount) + asnSecondBand: seen,
asns := map[int]time.Time{
i % contendedASNCount: ts,
(i % contendedASNCount) + asnSecondBand: ts,
}
err := db.GetOrCreateASNBatch(asns)
@@ -751,8 +712,7 @@ func TestBatchWriteWaitsForAnalyze(t *testing.T) {
db, messages := newDatabaseWithLockMessages(t)
ctx := context.Background()
now := time.Now().UTC()
asns := map[int]SeenTimes{lockTestASN: {First: now, Last: now}}
asns := map[int]time.Time{lockTestASN: time.Now().UTC()}
// A stored AS gives ANALYZE a table to analyze and statistics to write.
err := db.GetOrCreateASNBatch(asns)
@@ -842,388 +802,6 @@ func TestCheckpointLockWaitEndsWithContext(t *testing.T) {
db.mu.Unlock()
}
// TestCheckpointDoesNotWaitForReader keeps a read transaction on
// live_routes_v4 open while a route is committed, then starts Checkpoint and a
// batch write together. It checks that both end within readerCheckpointLimit,
// whichever of them takes the database lock first, and that Checkpoint reports
// the part of the WAL the reader still needs as left uncopied. A TRUNCATE
// checkpoint waited the busy timeout for the reader while holding the lock
// (https://git.eeqj.de/sneak/routewatch/issues/102).
func TestCheckpointDoesNotWaitForReader(t *testing.T) {
t.Parallel()
db, err := New(&config.Config{StateDir: t.TempDir()}, logger.New())
if err != nil {
t.Fatalf("failed to create database: %v", err)
}
defer func() { _ = db.Close() }()
ctx := context.Background()
ts := time.Now().UTC()
err = db.UpsertLiveRouteBatch(
[]*LiveRoute{mkV4Route(t, "198.51.100.0/24", lockTestASN, ts)})
if err != nil {
t.Fatalf("failed to store the route the reader reads: %v", err)
}
reader, err := db.db.Conn(ctx)
if err != nil {
t.Fatalf("failed to open the reading connection: %v", err)
}
defer func() { _ = reader.Close() }()
// A plain BEGIN starts a transaction that only reads, unlike the BEGIN
// IMMEDIATE of _txlock. It keeps the snapshot its first read takes until
// it ends.
_, err = reader.ExecContext(ctx, "BEGIN")
if err != nil {
t.Fatalf("failed to begin the read transaction: %v", err)
}
defer func() { _, _ = reader.ExecContext(ctx, "ROLLBACK") }()
var routes int
err = reader.QueryRowContext(ctx,
"SELECT count(*) FROM live_routes_v4").Scan(&routes)
if err != nil {
t.Fatalf("failed to read the live routes: %v", err)
}
err = db.UpsertLiveRouteBatch(
[]*LiveRoute{mkV4Route(t, "203.0.113.0/24", lockTestASN, ts)})
if err != nil {
t.Fatalf("failed to store a route after the read: %v", err)
}
started := time.Now()
checkpointed := make(chan error, 1)
go func() { checkpointed <- db.Checkpoint(ctx) }()
// A route write, not an AS write: the first new AS loads the embedded AS
// names, which takes seconds under -race.
route := mkV4Route(t, "192.0.2.0/24", lockTestASN, ts)
written := make(chan error, 1)
go func() { written <- db.UpsertLiveRouteBatch([]*LiveRoute{route}) }()
err = <-checkpointed
if !errors.Is(err, ErrCheckpointIncomplete) {
t.Errorf("Checkpoint = %v with a reader open, want %v",
err, ErrCheckpointIncomplete)
}
took := time.Since(started)
if took >= readerCheckpointLimit {
t.Errorf("Checkpoint took %v with a reader open, want under %v",
took, readerCheckpointLimit)
}
err = <-written
if err != nil {
t.Errorf("batch write during Checkpoint: %v", err)
}
took = time.Since(started)
if took >= readerCheckpointLimit {
t.Errorf("batch write during Checkpoint ended after %v, want under %v",
took, readerCheckpointLimit)
}
}
// TestCheckpointEmptiesWALOnceNoReadIsOpen fills the WAL file past
// sqliteWALSizeLimitBytes while the test holds a read open. With the read
// open, it checks that Checkpoint gives up within truncateCheckpointLimit and
// reports the checkpoint incomplete, and that a batch write still succeeds.
// After the test closes the read, it checks that the next Checkpoint brings
// the WAL file under the limit and gives the writing connection its busy
// timeout back (https://git.eeqj.de/sneak/routewatch/issues/102).
func TestCheckpointEmptiesWALOnceNoReadIsOpen(t *testing.T) {
t.Parallel()
db, err := New(&config.Config{StateDir: t.TempDir()}, logger.New())
if err != nil {
t.Fatalf("failed to create database: %v", err)
}
defer func() { _ = db.Close() }()
ctx := t.Context()
reader, err := db.db.Conn(ctx)
if err != nil {
t.Fatalf("failed to open the reading connection: %v", err)
}
defer func() { _ = reader.Close() }()
// A plain BEGIN starts a transaction that only reads, unlike the BEGIN
// IMMEDIATE of _txlock. It keeps the snapshot its first read takes until
// it ends, so no checkpoint can copy what is written after that read.
_, err = reader.ExecContext(ctx, "BEGIN")
if err != nil {
t.Fatalf("failed to begin the read transaction: %v", err)
}
var routes int
err = reader.QueryRowContext(ctx,
"SELECT count(*) FROM live_routes_v4").Scan(&routes)
if err != nil {
t.Fatalf("failed to read the live routes: %v", err)
}
fillWALPastLimit(ctx, t, db)
started := time.Now()
err = db.Checkpoint(ctx)
if !errors.Is(err, ErrCheckpointIncomplete) {
t.Errorf("Checkpoint = %v with a read open, want %v",
err, ErrCheckpointIncomplete)
}
took := time.Since(started)
if took >= truncateCheckpointLimit {
t.Errorf("Checkpoint took %v with a read open, want under %v",
took, truncateCheckpointLimit)
}
err = db.UpsertLiveRouteBatch([]*LiveRoute{
mkV4Route(t, "192.0.2.0/24", lockTestASN, time.Now().UTC()),
})
if err != nil {
t.Errorf("batch write after Checkpoint gave up: %v", err)
}
_, err = reader.ExecContext(ctx, "ROLLBACK")
if err != nil {
t.Fatalf("failed to end the read transaction: %v", err)
}
err = db.Checkpoint(ctx)
if err != nil {
t.Errorf("Checkpoint with no read open: %v", err)
}
size := walFileSize(t, db)
if size >= sqliteWALSizeLimitBytes {
t.Errorf("WAL file of %d bytes after Checkpoint with no read open, "+
"want under %d", size, sqliteWALSizeLimitBytes)
}
busyTimeout := readPragma(ctx, t, db.writer, 0, "busy_timeout")
if busyTimeout != sqliteBusyTimeoutMs {
t.Errorf("writing connection: busy_timeout = %d after the checkpoints, "+
"want %d", busyTimeout, sqliteBusyTimeoutMs)
}
}
// fillWALPastLimit writes sqliteWALSizeLimitBytes of random bytes to a table
// of the test's own, which with SQLite's page and frame headers leaves the WAL
// file of db over the limit, and checks that it does.
func fillWALPastLimit(ctx context.Context, t *testing.T, db *Database) {
t.Helper()
_, err := db.writer.ExecContext(ctx, "CREATE TABLE wal_fill (data BLOB)")
if err != nil {
t.Fatalf("failed to create the table the test writes to: %v", err)
}
_, err = db.writer.ExecContext(ctx,
"INSERT INTO wal_fill VALUES (randomblob(?))", sqliteWALSizeLimitBytes)
if err != nil {
t.Fatalf("failed to fill the WAL: %v", err)
}
size := walFileSize(t, db)
if size <= sqliteWALSizeLimitBytes {
t.Fatalf("WAL file of %d bytes after the test's write, want over %d",
size, sqliteWALSizeLimitBytes)
}
}
// walFileSize returns the size of the WAL file of db.
func walFileSize(t *testing.T, db *Database) int64 {
t.Helper()
wal, err := os.Stat(db.path + walFileSuffix)
if err != nil {
t.Fatalf("failed to read the WAL file's size: %v", err)
}
return wal.Size()
}
// TestStartupCheckpointLogsItsFigures opens a new database, writes a route,
// and opens the same file again while the first stays open, so that its WAL
// file stays. The first open finds no WAL file: it logs a size of 0 and no
// warning. The second logs the size of the WAL file the first left, and the
// checkpoint's busy flag and frame counts, each under its own name; the
// counts read 0 once the checkpoint has emptied the file.
// Initialize logged the busy flag as wal_pages and the counts under each
// other's names (https://git.eeqj.de/sneak/routewatch/issues/117).
func TestStartupCheckpointLogsItsFigures(t *testing.T) {
t.Parallel()
cfg := &config.Config{StateDir: t.TempDir()}
db, lines := openLogged(t, cfg)
for _, line := range lines {
if line[slog.LevelKey] == slog.LevelWarn.String() {
t.Errorf("opening a new database logged a warning: %v", line)
}
}
line := findLine(t, lines, startupCheckpointMessage)
if line["wal_bytes"] != 0.0 {
t.Errorf("new database: start-up checkpoint logged wal_bytes = %v, "+
"want 0", line["wal_bytes"])
}
err := db.UpsertLiveRouteBatch([]*LiveRoute{
mkV4Route(t, "192.0.2.0/24", lockTestASN, time.Now().UTC()),
})
if err != nil {
t.Fatalf("failed to store a route: %v", err)
}
walBytes := walFileSize(t, db)
if walBytes == 0 {
t.Fatal("WAL file empty after the route write, want frames in it")
}
_, lines = openLogged(t, cfg)
line = findLine(t, lines, startupCheckpointMessage)
for _, key := range []string{"busy", "wal_frames", "checkpointed_frames"} {
if line[key] != 0.0 {
t.Errorf("start-up checkpoint logged %s = %v, want 0", key, line[key])
}
}
if line["wal_bytes"] != float64(walBytes) {
t.Errorf("start-up checkpoint logged wal_bytes = %v, want the %d "+
"bytes the first database left", line["wal_bytes"], walBytes)
}
}
// TestStartupCheckpointWarnsWhenIncomplete keeps a read open on a database
// while a route is written after the read began, then opens the same file
// again. The second open's start-up checkpoint cannot copy the frames written
// after the read began, and gives up once the busy timeout has passed. It
// checks that the checkpoint is logged as a warning with busy 1 and fewer
// frames copied than the WAL holds.
func TestStartupCheckpointWarnsWhenIncomplete(t *testing.T) {
t.Parallel()
cfg := &config.Config{StateDir: t.TempDir()}
db, _ := openLogged(t, cfg)
ctx := context.Background()
reader, err := db.db.Conn(ctx)
if err != nil {
t.Fatalf("failed to open the reading connection: %v", err)
}
defer func() { _ = reader.Close() }()
// A plain BEGIN starts a transaction that only reads, unlike the BEGIN
// IMMEDIATE of _txlock. It keeps the snapshot its first read takes until
// it ends.
_, err = reader.ExecContext(ctx, "BEGIN")
if err != nil {
t.Fatalf("failed to begin the read transaction: %v", err)
}
defer func() { _, _ = reader.ExecContext(ctx, "ROLLBACK") }()
var routes int
err = reader.QueryRowContext(ctx,
"SELECT count(*) FROM live_routes_v4").Scan(&routes)
if err != nil {
t.Fatalf("failed to read the live routes: %v", err)
}
err = db.UpsertLiveRouteBatch([]*LiveRoute{
mkV4Route(t, "192.0.2.0/24", lockTestASN, time.Now().UTC()),
})
if err != nil {
t.Fatalf("failed to store a route after the read: %v", err)
}
_, lines := openLogged(t, cfg)
line := findLine(t, lines, startupCheckpointIncompleteMessage)
if line[slog.LevelKey] != slog.LevelWarn.String() {
t.Errorf("start-up checkpoint logged at level %v, want %s",
line[slog.LevelKey], slog.LevelWarn)
}
if line["busy"] != 1.0 {
t.Errorf("start-up checkpoint logged busy = %v, want 1", line["busy"])
}
walFrames, _ := line["wal_frames"].(float64)
checkpointedFrames, _ := line["checkpointed_frames"].(float64)
if checkpointedFrames >= walFrames {
t.Errorf("start-up checkpoint logged checkpointed_frames = %v and "+
"wal_frames = %v with a read open, want fewer copied than the WAL "+
"holds", line["checkpointed_frames"], line["wal_frames"])
}
}
// openLogged opens the database of cfg with a logger that writes JSON, and
// returns it with the lines it logged while it opened, decoded.
func openLogged(t *testing.T, cfg *config.Config) (*Database, []map[string]any) {
t.Helper()
var logged bytes.Buffer
db, err := New(cfg,
&logger.Logger{Logger: slog.New(slog.NewJSONHandler(&logged, nil))})
if err != nil {
t.Fatalf("failed to open the database: %v", err)
}
t.Cleanup(func() { _ = db.Close() })
var lines []map[string]any
decoder := json.NewDecoder(&logged)
for decoder.More() {
var line map[string]any
err := decoder.Decode(&line)
if err != nil {
t.Fatalf("failed to decode the log: %v", err)
}
lines = append(lines, line)
}
return db, lines
}
// findLine returns the first of lines logged with message.
func findLine(
t *testing.T,
lines []map[string]any,
message string,
) map[string]any {
t.Helper()
for _, line := range lines {
if line[slog.MessageKey] == message {
return line
}
}
t.Fatalf("no line %q in the log", message)
return nil
}
// TestGetRandomPrefixesByLengthListsPrefixOnce stores one prefix of each
// family learned from two peers in the same second and checks that the
// prefix-length sample lists each of them once.
@@ -1639,141 +1217,9 @@ func TestLiveRouteUpdateSetsOriginWhenItChanges(t *testing.T) {
}
}
// TestGetASDetailsListsEachPrefixOnce stores two IPv4 prefixes and an IPv6
// prefix an AS originates, two of them from two peers at different times, and
// a prefix another AS originates. The AS details list each of the AS's
// prefixes once, IPv4 first and in text order, with its mask length and the
// time of its newest route.
func TestGetASDetailsListsEachPrefixOnce(t *testing.T) {
t.Parallel()
db, err := New(&config.Config{StateDir: t.TempDir()}, logger.New())
if err != nil {
t.Fatalf("failed to create database: %v", err)
}
defer func() { _ = db.Close() }()
const asn, otherASN = 64500, 64501
ts := time.Date(2026, 1, 2, 3, 4, 5, 0, time.UTC)
_, err = db.GetOrCreateASN(asn, ts)
if err != nil {
t.Fatalf("GetOrCreateASN: %v", err)
}
v4Older := mkV4Route(t, "192.0.2.0/24", asn, ts)
v4Newer := mkV4Route(t, "192.0.2.0/24", asn, ts.Add(time.Minute))
v4Newer.PeerIP = secondV4PeerIP
v4Single := mkV4Route(t, "198.51.100.128/25", asn, ts)
v6Newer := mkV6Route(testV6Prefix, asn, ts.Add(time.Minute))
v6Older := mkV6Route(testV6Prefix, asn, ts)
v6Older.PeerIP = secondV6PeerIP
other := mkV4Route(t, "203.0.113.0/24", otherASN, ts)
err = db.UpsertLiveRouteBatch([]*LiveRoute{
v4Older, v4Newer, v4Single, v6Newer, v6Older, other,
})
if err != nil {
t.Fatalf("UpsertLiveRouteBatch: %v", err)
}
_, prefixes, err := db.GetASDetails(asn)
if err != nil {
t.Fatalf("GetASDetails: %v", err)
}
want := []*LiveRoute{v4Newer, v4Single, v6Newer}
listedOnce := slices.EqualFunc(prefixes, want,
func(got LiveRoute, want *LiveRoute) bool {
return got.Prefix == want.Prefix &&
got.MaskLength == want.MaskLength &&
got.IPVersion == want.IPVersion &&
got.OriginASN == asn &&
got.LastUpdated.Equal(want.LastUpdated)
})
if !listedOnce {
t.Errorf("GetASDetails listed %+v, want the prefixes and times of %+v",
prefixes, want)
}
}
// TestASPrefixQueriesUseCoveringIndexes checks the plan of each query that
// lists the prefixes an AS originates: a search of a covering index and no
// temporary B-tree, so the list reads one index entry per prefix, not one
// table row per route. Grouping the AS's routes by prefix read every route's
// row and sorted them in a temporary B-tree
// (https://git.eeqj.de/sneak/routewatch/issues/104).
func TestASPrefixQueriesUseCoveringIndexes(t *testing.T) {
t.Parallel()
db, err := New(&config.Config{StateDir: t.TempDir()}, logger.New())
if err != nil {
t.Fatalf("failed to create database: %v", err)
}
defer func() { _ = db.Close() }()
tests := []struct {
query string
args []any
index string
}{
{nextOriginPrefixV4SQL, []any{0, ""}, "idx_live_routes_v4_origin_prefix"},
{newestRouteTimeV4SQL, []any{""}, "idx_live_routes_v4_prefix_last_updated"},
{nextOriginPrefixV6SQL, []any{0, ""}, "idx_live_routes_v6_origin_prefix"},
{newestRouteTimeV6SQL, []any{""}, "idx_live_routes_v6_prefix_last_updated"},
}
for _, tt := range tests {
plan := queryPlan(t, db, tt.query, tt.args...)
if !strings.Contains(plan, "USING COVERING INDEX "+tt.index+" ") ||
strings.Contains(plan, "TEMP B-TREE") {
t.Errorf("plan of %s:\n%s\nwant a search of covering index %s "+
"and no temporary B-tree", tt.query, plan, tt.index)
}
}
}
// queryPlan returns the steps of EXPLAIN QUERY PLAN for query, one per line.
func queryPlan(t *testing.T, db *Database, query string, args ...any) string {
t.Helper()
rows, err := db.db.QueryContext(t.Context(), "EXPLAIN QUERY PLAN "+query,
args...)
if err != nil {
t.Fatalf("EXPLAIN QUERY PLAN %s: %v", query, err)
}
defer func() { _ = rows.Close() }()
var steps []string
for rows.Next() {
var (
id, parent, unused int
detail string
)
err := rows.Scan(&id, &parent, &unused, &detail)
if err != nil {
t.Fatalf("reading the plan of %s: %v", query, err)
}
steps = append(steps, detail)
}
err = rows.Err()
if err != nil {
t.Fatalf("reading the plan of %s: %v", query, err)
}
return strings.Join(steps, "\n")
}
// TestUpdatePrefixesBatchKeepsFirstSeen writes a prefix, then a batch with it
// and a new prefix. Only the new prefix is inserted and counted, and the known
// one keeps its first_seen time.
// and a new prefix. Only the new prefix is counted, and the known one keeps its
// first_seen time while its last_seen time moves on.
func TestUpdatePrefixesBatchKeepsFirstSeen(t *testing.T) {
t.Parallel()
@@ -1811,75 +1257,33 @@ func TestUpdatePrefixesBatchKeepsFirstSeen(t *testing.T) {
}
for _, tt := range []struct {
query, prefix string
firstSeen time.Time
query, prefix string
firstSeen, lastSeen time.Time
}{
{
query: "SELECT first_seen FROM prefixes_v4 WHERE prefix = ?",
prefix: known, firstSeen: first,
query: "SELECT first_seen, last_seen FROM prefixes_v4 WHERE prefix = ?",
prefix: known, firstSeen: first, lastSeen: later,
},
{
query: "SELECT first_seen FROM prefixes_v6 WHERE prefix = ?",
prefix: testV6Prefix, firstSeen: later,
query: "SELECT first_seen, last_seen FROM prefixes_v6 WHERE prefix = ?",
prefix: testV6Prefix, firstSeen: later, lastSeen: later,
},
} {
var firstSeen time.Time
var firstSeen, lastSeen time.Time
err := db.db.QueryRowContext(t.Context(), tt.query, tt.prefix).
Scan(&firstSeen)
Scan(&firstSeen, &lastSeen)
if err != nil {
t.Fatalf("reading %s: %v", tt.prefix, err)
}
if !firstSeen.Equal(tt.firstSeen) {
t.Errorf("%s: first seen %s, want %s", tt.prefix, firstSeen,
tt.firstSeen)
if !firstSeen.Equal(tt.firstSeen) || !lastSeen.Equal(tt.lastSeen) {
t.Errorf("%s: first seen %s, last seen %s; want %s, %s", tt.prefix,
firstSeen, lastSeen, tt.firstSeen, tt.lastSeen)
}
}
}
// TestUpdatePeerBatchStoresTheEarliestAndLatestTimesOfANewPeer writes a new
// peer whose updates span 5 seconds. Its row is first seen at the earliest time
// and last seen at the latest.
func TestUpdatePeerBatchStoresTheEarliestAndLatestTimesOfANewPeer(
t *testing.T,
) {
t.Parallel()
db, err := New(&config.Config{StateDir: t.TempDir()}, logger.New())
if err != nil {
t.Fatalf("failed to create database: %v", err)
}
defer func() { _ = db.Close() }()
first := time.Date(2026, 1, 2, 3, 4, 5, 0, time.UTC)
last := first.Add(5 * time.Second)
err = db.UpdatePeerBatch(map[string]PeerUpdate{
testPeerIP: {
PeerIP: testPeerIP, PeerASN: 64500, MessageType: "UPDATE",
Seen: SeenTimes{First: first, Last: last},
},
})
if err != nil {
t.Fatalf("UpdatePeerBatch: %v", err)
}
var firstSeen, lastSeen time.Time
err = db.db.QueryRowContext(t.Context(),
"SELECT first_seen, last_seen FROM bgp_peers WHERE peer_ip = ?",
testPeerIP).Scan(&firstSeen, &lastSeen)
if err != nil {
t.Fatalf("reading peer %s: %v", testPeerIP, err)
}
if !firstSeen.Equal(first) || !lastSeen.Equal(last) {
t.Errorf("peer %s: first seen %s, last seen %s; want %s, %s",
testPeerIP, firstSeen, lastSeen, first, last)
}
}
// TestUpsertLiveRouteBatchWritesInKeyOrder writes new routes of two peers out
// of key order, the last one a second announcement of the first, and checks
// that the rows were inserted in key order, peer then prefix (each new row's id
@@ -2085,8 +1489,8 @@ func benchRoute(rng *rand.Rand, prefix, peer int, ts time.Time) *LiveRoute {
}
// writeBenchBatch writes what the prefix handler writes for a batch: the
// announced routes, the withdrawals, and every prefix the batch names, of
// which only the ones not stored yet are inserted.
// announced routes, the withdrawals, and the last-seen time of every prefix
// the batch names.
func writeBenchBatch(
b *testing.B, db *Database, routes []*LiveRoute,
withdrawals []LiveRouteDeletion, ts time.Time,
@@ -2117,66 +1521,3 @@ func writeBenchBatch(
b.Fatalf("UpdatePrefixesBatch: %v", err)
}
}
// Parameters for BenchmarkGetASDetailsWithFiveHundredPeers: the AS originates
// benchASPrefixes prefixes and every one of benchASPeers peers has a route for
// each, 2.5 million routes in all.
const (
benchASPrefixes = 5000
benchASPeers = 500
)
// BenchmarkGetASDetailsWithFiveHundredPeers times the read behind the AS page
// for an AS that originates 5,000 IPv4 prefixes, each from 500 peers. It
// builds that database first, one peer at a time, which takes some minutes and
// about a GB of disk under TMPDIR. It is run by hand, not by make test, for
// example in the test image:
//
// go test -run '^$' -bench GetASDetails -benchtime 10x ./internal/database/
func BenchmarkGetASDetailsWithFiveHundredPeers(b *testing.B) {
db, err := New(&config.Config{StateDir: b.TempDir()}, logger.New())
if err != nil {
b.Fatalf("failed to create database: %v", err)
}
defer func() { _ = db.Close() }()
ts := time.Date(2026, 1, 2, 3, 4, 5, 0, time.UTC)
_, err = db.GetOrCreateASN(benchFirstASN, ts)
if err != nil {
b.Fatalf("GetOrCreateASN: %v", err)
}
for peer := range benchASPeers {
peerIP := fmt.Sprintf("198.18.%d.%d", peer/256, peer%256)
routes := make([]*LiveRoute, 0, benchASPrefixes)
for prefix := range benchASPrefixes {
routes = append(routes, &LiveRoute{
Prefix: fmt.Sprintf("%d.%d.%d.0/%d",
1+prefix>>16, (prefix>>8)%256, prefix%256, benchMaskLength),
MaskLength: benchMaskLength,
IPVersion: ipVersionV4,
OriginASN: benchFirstASN,
PeerIP: peerIP,
ASPath: []int{benchFirstASN + 1 + peer, benchFirstASN},
NextHop: peerIP,
LastUpdated: ts,
})
}
writeBenchBatch(b, db, routes, nil, ts)
}
for b.Loop() {
_, prefixes, err := db.GetASDetails(benchFirstASN)
if err != nil {
b.Fatalf("GetASDetails: %v", err)
}
if len(prefixes) != benchASPrefixes {
b.Fatalf("GetASDetails listed %d prefixes, want %d",
len(prefixes), benchASPrefixes)
}
}
}
internal/database/dockerignore_internal_test.go
-61
View File
@@ -1,61 +0,0 @@
package database
import (
"os"
"path/filepath"
"strings"
"testing"
"git.eeqj.de/sneak/routewatch/internal/config"
"git.eeqj.de/sneak/routewatch/internal/logger"
)
// TestDockerignoreExcludesDatabaseFiles checks that .dockerignore keeps the
// database file, and the WAL and shared-memory files SQLite writes next to it,
// out of the build context wherever they sit in the clone. The image ships the
// whole build context as a source archive.
func TestDockerignoreExcludesDatabaseFiles(t *testing.T) {
t.Parallel()
db, err := New(&config.Config{StateDir: t.TempDir()}, logger.New())
if err != nil {
t.Fatalf("failed to create database: %v", err)
}
defer func() { _ = db.Close() }()
data, err := os.ReadFile("../../.dockerignore")
if err != nil {
t.Fatalf("failed to read .dockerignore: %v", err)
}
// Docker reads `**/` as any number of directories and the rest of the
// pattern as filepath.Match does, so `**/` followed by a pattern that
// matches a file's name excludes that file at any depth.
var anyDepth []string
for line := range strings.Lines(string(data)) {
pattern, ok := strings.CutPrefix(strings.TrimSpace(line), "**/")
if ok {
anyDepth = append(anyDepth, pattern)
}
}
name := filepath.Base(db.path)
for _, file := range []string{name, name + "-wal", name + "-shm"} {
excluded := false
for _, pattern := range anyDepth {
matched, err := filepath.Match(pattern, file)
if err != nil {
t.Fatalf("bad .dockerignore pattern **/%s: %v", pattern, err)
}
excluded = excluded || matched
}
if !excluded {
t.Errorf(".dockerignore has no **/ pattern that excludes %s", file)
}
}
}
internal/database/interface.go
+2 -2
View File
@@ -34,7 +34,7 @@ type Stats struct {
type Store interface {
// ASN operations
GetOrCreateASN(number int, timestamp time.Time) (*ASN, error)
GetOrCreateASNBatch(asns map[int]SeenTimes) error
GetOrCreateASNBatch(asns map[int]time.Time) error
// Prefix operations
GetOrCreatePrefix(prefix string, timestamp time.Time) (*Prefix, error)
@@ -44,7 +44,7 @@ type Store interface {
RecordAnnouncement(announcement *Announcement) error
// Peering operations
RecordPeering(asA, asB int, seen SeenTimes) error
RecordPeering(asA, asB int, timestamp time.Time) error
// Statistics
GetStats() (Stats, error)
internal/database/models.go
+4 -26
View File
@@ -35,11 +35,12 @@ type ASN struct {
}
// Prefix represents an IP prefix (CIDR block) with its IP version (4 or 6)
// and the time it was first seen.
// and first/last seen timestamps.
type Prefix struct {
Prefix string
IPVersion int // 4 or 6
FirstSeen time.Time
LastSeen time.Time
}
// Announcement represents a BGP announcement or withdrawal event,
@@ -122,35 +123,12 @@ type LiveRouteDeletion struct {
IPVersion int
}
// SeenTimes is the earliest and the latest time one batch of updates saw an AS,
// a peer or a peering. A row the batch adds is stored first seen at First and
// last seen at Last; a row already stored keeps its first_seen and has its
// last_seen set to Last.
type SeenTimes struct {
First time.Time
Last time.Time
}
// Add makes t the earliest time of s when it is earlier than First, and the
// latest when it is later than Last. The zero SeenTimes holds no time yet, so
// the first Add sets both.
func (s *SeenTimes) Add(t time.Time) {
if s.First.IsZero() || t.Before(s.First) {
s.First = t
}
if s.Last.IsZero() || t.After(s.Last) {
s.Last = t
}
}
// PeerUpdate represents parameters for updating a peer. PeerASN and
// MessageType are those of the peer's latest update.
// PeerUpdate represents parameters for updating a peer
type PeerUpdate struct {
PeerIP string
PeerASN int
MessageType string
Seen SeenTimes
Timestamp time.Time
}
// ASNWHOISUpdate contains WHOIS data for updating an ASN record.
internal/database/schema.sql
+8 -7
View File
@@ -31,13 +31,15 @@ CREATE TABLE IF NOT EXISTS asns (
-- rows are stored by prefix (WITHOUT ROWID) and need no index of their own.
CREATE TABLE IF NOT EXISTS prefixes_v4 (
prefix TEXT PRIMARY KEY,
first_seen DATETIME NOT NULL
first_seen DATETIME NOT NULL,
last_seen DATETIME NOT NULL
) WITHOUT ROWID;
-- IPv6 prefixes table, stored as prefixes_v4 is.
CREATE TABLE IF NOT EXISTS prefixes_v6 (
prefix TEXT PRIMARY KEY,
first_seen DATETIME NOT NULL
first_seen DATETIME NOT NULL,
last_seen DATETIME NOT NULL
) WITHOUT ROWID;
CREATE TABLE IF NOT EXISTS announcements (
@@ -128,7 +130,7 @@ CREATE TABLE IF NOT EXISTS live_routes_v6 (
-- prefix, last_updated: a prefix's routes, for the prefix page and its API
-- answer, the IP lookup and its peer count, and the check whether a prefix
-- still has a live route; and a prefix's newest route for the prefix-length
-- pages and the AS page, in one search whatever the number of peers.
-- pages, in one search whatever the number of peers.
CREATE INDEX IF NOT EXISTS idx_live_routes_v4_prefix_last_updated ON live_routes_v4(prefix, last_updated);
CREATE INDEX IF NOT EXISTS idx_live_routes_v6_prefix_last_updated ON live_routes_v6(prefix, last_updated);
-- mask_length: the prefix-length pages' random sample, the row id range of a
@@ -139,10 +141,9 @@ CREATE INDEX IF NOT EXISTS idx_live_routes_v6_mask_length ON live_routes_v6(mask
-- prefixes in text order, and the prefix distribution read at startup.
CREATE INDEX IF NOT EXISTS idx_live_routes_v4_mask_prefix ON live_routes_v4(mask_length, prefix);
CREATE INDEX IF NOT EXISTS idx_live_routes_v6_mask_prefix ON live_routes_v6(mask_length, prefix);
-- origin_asn, prefix: the prefixes an AS originates, each in one search
-- whatever the number of peers, for the AS page and its API answer.
CREATE INDEX IF NOT EXISTS idx_live_routes_v4_origin_prefix ON live_routes_v4(origin_asn, prefix);
CREATE INDEX IF NOT EXISTS idx_live_routes_v6_origin_prefix ON live_routes_v6(origin_asn, prefix);
-- origin_asn: the prefixes an AS originates, for the AS page and its API answer.
CREATE INDEX IF NOT EXISTS idx_live_routes_v4_origin_asn ON live_routes_v4(origin_asn);
CREATE INDEX IF NOT EXISTS idx_live_routes_v6_origin_asn ON live_routes_v6(origin_asn);
-- last_updated: the oldest and newest route times in /api/v1/stats and
-- /status.json.
CREATE INDEX IF NOT EXISTS idx_live_routes_v4_last_updated ON live_routes_v4(last_updated);
internal/ristypes/ris.go
+9 -27
View File
@@ -4,7 +4,6 @@ package ristypes
import (
"encoding/json"
"slices"
"time"
)
@@ -25,8 +24,7 @@ const (
)
// ASPath is a BGP AS path as the feed writes it: an array of AS numbers in
// which an AS set is an array of its members, such as [1, 2, [3, 4]]. A path
// that holds AS 0 decodes as an empty one (UnmarshalJSON).
// which an AS set is an array of its members, such as [1, 2, [3, 4]].
type ASPath struct {
// Sequence is the path's AS numbers in order. A set of one member that
// ends the path counts as that AS and is its last element. The members of
@@ -50,41 +48,23 @@ func (p *ASPath) Origin() int {
return p.Sequence[len(p.Sequence)-1]
}
// UnmarshalJSON implements the json.Unmarshaler interface for ASPath. A path
// that holds AS 0 anywhere, a member of an AS set included, decodes as an empty
// path: RFC 7607 makes such a path malformed and its announcements
// withdrawals. On an announcement with an empty path the prefix handler
// deletes the peer's route, and the AS and peering handlers record nothing.
// UnmarshalJSON implements the json.Unmarshaler interface for ASPath.
func (p *ASPath) UnmarshalJSON(data []byte) error {
path, err := decodeASPath(data)
if err != nil {
return err
}
if slices.Contains(path.Sequence, 0) || slices.Contains(path.Set, 0) {
path = ASPath{}
}
*p = path
return nil
}
// decodeASPath decodes an AS path as the feed writes it.
func decodeASPath(data []byte) (ASPath, error) {
// Most paths have no AS set and decode as a plain array of integers.
var sequence []int
err := json.Unmarshal(data, &sequence)
if err == nil {
return ASPath{Sequence: sequence}, nil
*p = ASPath{Sequence: sequence}
return nil
}
var raw []any
err = json.Unmarshal(data, &raw)
if err != nil {
return ASPath{}, err
return err
}
path := ASPath{Sequence: make([]int, 0, len(raw))}
@@ -110,7 +90,9 @@ func decodeASPath(data []byte) (ASPath, error) {
}
}
return path, nil
*p = path
return nil
}
// RISLiveMessage represents the outer wrapper message from the RIPE RIS Live stream.
internal/ristypes/ris_test.go
-30
View File
@@ -145,33 +145,3 @@ func TestASPathKeepsATrailingSetApart(t *testing.T) {
})
}
}
// TestASPathWithASZeroDecodesAsAnEmptyPath decodes AS paths that hold AS 0
// before their end, at their end, or in an AS set. RFC 7607 makes such a path
// malformed, so each decodes as an empty path.
func TestASPathWithASZeroDecodesAsAnEmptyPath(t *testing.T) {
t.Parallel()
for _, path := range []string{
"[64496, 0, 64497]",
"[64496, 0]",
"[64496, [0, 64497], 64498]",
"[64496, [64497, 0]]",
} {
t.Run(path, func(t *testing.T) {
t.Parallel()
var got ristypes.ASPath
err := json.Unmarshal([]byte(path), &got)
if err != nil {
t.Fatalf("unmarshal %s: %v", path, err)
}
if len(got.Sequence) > 0 || len(got.Set) > 0 {
t.Errorf("%s decoded to sequence %v and set %v, want an empty path",
path, got.Sequence, got.Set)
}
})
}
}
internal/routewatch/app.go
+23 -47
View File
@@ -116,18 +116,17 @@ func (rw *RouteWatch) Run(ctx context.Context) error {
return err
}
// Start ASN WHOIS fetcher for background updates. The HTTP handlers read
// the server's fetcher without a lock, so it is set before they run.
rw.asnFetcher = NewASNFetcher(rw.db, rw.logger.Logger)
rw.asnFetcher.Start(ctx)
rw.server.SetASNFetcher(rw.asnFetcher)
// Start HTTP server
err = rw.server.Start(ctx)
if err != nil {
return err
}
// Start ASN WHOIS fetcher for background updates
rw.asnFetcher = NewASNFetcher(rw.db, rw.logger.Logger)
rw.asnFetcher.Start(ctx)
rw.server.SetASNFetcher(rw.asnFetcher)
// Start database maintenance goroutine
rw.dbMaintainer = NewDBMaintainer(rw.db, rw.logger.Logger)
rw.dbMaintainer.Start(ctx)
@@ -138,12 +137,7 @@ func (rw *RouteWatch) Run(ctx context.Context) error {
return nil
}
// Shutdown stops all services in order: the streamer, the handlers, which
// write what they hold, the WHOIS fetcher, the database maintainer and the
// HTTP server, and last the database, which it checkpoints and closes. Each
// stage waits at most until ctx ends, so Shutdown returns soon after: a
// handler logs what it could not write by then, and the HTTP server closes
// the connections of requests still running.
// Shutdown performs graceful shutdown of all services
func (rw *RouteWatch) Shutdown(ctx context.Context) {
rw.mu.Lock()
if rw.shutdown {
@@ -156,48 +150,51 @@ func (rw *RouteWatch) Shutdown(ctx context.Context) {
rw.mu.Unlock()
// Stop the streamer first: it returns once the handlers have been handed
// every message it queued, or when ctx ends. Then each handler writes
// what it holds.
rw.streamer.Stop(ctx)
// every message it queued. Then each handler writes what it holds.
rw.streamer.Stop()
if rw.dbHandler != nil {
rw.logger.Info("Flushing database handler")
rw.dbHandler.Stop(ctx)
rw.dbHandler.Stop()
}
if rw.peerHandler != nil {
rw.logger.Info("Flushing peer handler")
rw.peerHandler.Stop(ctx)
rw.peerHandler.Stop()
}
if rw.prefixHandler != nil {
rw.logger.Info("Flushing prefix handler")
rw.prefixHandler.Stop(ctx)
rw.prefixHandler.Stop()
}
if rw.peeringHandler != nil {
rw.logger.Info("Flushing peering handler")
rw.peeringHandler.Stop(ctx)
rw.peeringHandler.Stop()
}
// Stop ASN WHOIS fetcher
if rw.asnFetcher != nil {
rw.asnFetcher.Stop(ctx)
rw.asnFetcher.Stop()
}
// Stop database maintainer
if rw.dbMaintainer != nil {
rw.dbMaintainer.Stop(ctx)
rw.dbMaintainer.Stop()
}
err := rw.server.Stop(ctx)
// Stop HTTP server with a timeout of its own, however much of ctx is left
const serverStopTimeout = 5 * time.Second
stopCtx, cancel := context.WithTimeout(
context.WithoutCancel(ctx), serverStopTimeout)
defer cancel()
err := rw.server.Stop(stopCtx)
if err != nil {
rw.logger.Error("Stopped the HTTP server before every request finished",
"error", err)
rw.logger.Error("Failed to stop HTTP server gracefully", "error", err)
}
rw.closeDatabase(ctx)
// Log final metrics
metrics := rw.streamer.GetMetrics()
rw.logger.Info("Final metrics",
@@ -209,27 +206,6 @@ func (rw *RouteWatch) Shutdown(ctx context.Context) {
)
}
// closeDatabase checkpoints and closes the database, unless ctx has ended: a
// write may then still be running, and both would wait for it. SQLite rolls
// back a write cut off by the exit when the database next opens.
func (rw *RouteWatch) closeDatabase(ctx context.Context) {
if ctx.Err() != nil {
rw.logger.Error("Stop deadline passed; the database is not closed")
return
}
err := rw.db.Checkpoint(ctx)
if err != nil {
rw.logger.Error("Failed to checkpoint the database", "error", err)
}
err = rw.db.Close()
if err != nil {
rw.logger.Error("Failed to close the database", "error", err)
}
}
// getModule provides all fx dependencies
func getModule() []fx.Option {
return []fx.Option{
internal/routewatch/app_integration_internal_test.go
+13 -195
View File
@@ -4,11 +4,8 @@ import (
"context"
"fmt"
"io"
"net"
"net/http"
"net/http/httptest"
"slices"
"strconv"
"strings"
"sync"
"testing"
@@ -85,7 +82,9 @@ func (m *mockStore) UpdatePrefixesBatch(prefixes map[string]time.Time) error {
defer m.mu.Unlock()
for prefix, timestamp := range prefixes {
if _, exists := m.Prefixes[prefix]; !exists {
if p, exists := m.Prefixes[prefix]; exists {
p.LastSeen = timestamp
} else {
const (
ipVersionV4 = 4
ipVersionV6 = 6
@@ -100,6 +99,7 @@ func (m *mockStore) UpdatePrefixesBatch(prefixes map[string]time.Time) error {
Prefix: prefix,
IPVersion: ipVersion,
FirstSeen: timestamp,
LastSeen: timestamp,
}
}
}
@@ -115,6 +115,8 @@ func (m *mockStore) GetOrCreatePrefix(
defer m.mu.Unlock()
if p, exists := m.Prefixes[prefix]; exists {
p.LastSeen = timestamp
return p, nil
}
@@ -132,6 +134,7 @@ func (m *mockStore) GetOrCreatePrefix(
Prefix: prefix,
IPVersion: ipVersion,
FirstSeen: timestamp,
LastSeen: timestamp,
}
m.Prefixes[prefix] = p
m.PrefixCount++
@@ -152,7 +155,7 @@ func (m *mockStore) RecordAnnouncement(_ *database.Announcement) error {
}
// RecordPeering mock implementation
func (m *mockStore) RecordPeering(asA, asB int, _ database.SeenTimes) error {
func (m *mockStore) RecordPeering(asA, asB int, _ time.Time) error {
m.mu.Lock()
defer m.mu.Unlock()
@@ -367,6 +370,7 @@ func (m *mockStore) UpsertLiveRouteBatch(routes []*database.LiveRoute) error {
Prefix: route.Prefix,
IPVersion: route.IPVersion,
FirstSeen: route.LastUpdated,
LastSeen: route.LastUpdated,
}
m.PrefixCount++
@@ -396,16 +400,16 @@ func (m *mockStore) DeleteLiveRoutesOfPeer(_ string) (int, error) {
}
// GetOrCreateASNBatch mock implementation
func (m *mockStore) GetOrCreateASNBatch(asns map[int]database.SeenTimes) error {
func (m *mockStore) GetOrCreateASNBatch(asns map[int]time.Time) error {
m.mu.Lock()
defer m.mu.Unlock()
for number, seen := range asns {
for number, timestamp := range asns {
if _, exists := m.ASNs[number]; !exists {
m.ASNs[number] = &database.ASN{
ASN: number,
FirstSeen: seen.First,
LastSeen: seen.Last,
FirstSeen: timestamp,
LastSeen: timestamp,
}
m.ASNCount++
}
@@ -646,189 +650,3 @@ func TestShutdownWritesTheUpdatesQueuedInTheStreamer(t *testing.T) {
}
}
}
// freePort returns a TCP port that was free a moment ago, for a server the
// test starts.
func freePort(t *testing.T) int {
t.Helper()
var listenConfig net.ListenConfig
listener, err := listenConfig.Listen(t.Context(), "tcp", ":0")
if err != nil {
t.Fatalf("failed to find a free port: %v", err)
}
defer func() { _ = listener.Close() }()
addr, ok := listener.Addr().(*net.TCPAddr)
if !ok {
t.Fatalf("listener address %v is not a TCP address", listener.Addr())
}
return addr.Port
}
// TestShutdownReturnsByItsDeadlineWhileAWriteIsBlocked starts the HTTP server
// and has the AS handler start a write that the store holds until the test
// ends, past Shutdown's deadline. Shutdown must return within its deadline
// plus a small margin, with the HTTP server stopped, and must neither
// checkpoint nor close the database: the real one would wait for the write.
// Before the fix Shutdown waited for the write.
func TestShutdownReturnsByItsDeadlineWhileAWriteIsBlocked(t *testing.T) {
t.Parallel()
const (
deadline = 100 * time.Millisecond
// margin is how long Shutdown may take after its deadline, to stop
// the HTTP server.
margin = 2 * time.Second
)
blocking := newBlockingStore()
defer blocking.releaseWrites()
store := &closeOrderStore{Store: blocking}
log := logger.New()
port := freePort(t)
cfg := &config.Config{Port: port}
s := streamer.New(log, metrics.New())
rw := New(Dependencies{
DB: store,
Streamer: s,
Server: server.New(store, s, log, cfg),
Logger: log,
Config: cfg,
})
err := rw.server.Start(t.Context())
if err != nil {
t.Fatalf("failed to start the HTTP server: %v", err)
}
rw.dbHandler = NewASHandler(store, log)
fill := ristypes.ASPath{Sequence: slices.Repeat([]int{testASNA}, asnBatchSize)}
startBlockedWrite(t, blocking, rw.dbHandler,
[]*ristypes.RISMessage{{Path: fill}})
ctx, cancel := context.WithTimeout(t.Context(), deadline)
defer cancel()
shutDown := make(chan struct{})
go func() {
rw.Shutdown(ctx)
close(shutDown)
}()
select {
case <-shutDown:
case <-time.After(deadline + margin):
t.Fatalf("Shutdown did not return within %s of its deadline while "+
"a write was blocked", margin)
}
var dialer net.Dialer
addr := net.JoinHostPort("127.0.0.1", strconv.Itoa(port))
conn, err := dialer.DialContext(t.Context(), "tcp", addr)
if err == nil {
_ = conn.Close()
t.Errorf("the HTTP server still accepts connections on %s after "+
"Shutdown returned", addr)
}
store.callsMu.Lock()
defer store.callsMu.Unlock()
if slices.Contains(store.calls, callCheckpoint) ||
slices.Contains(store.calls, callClose) {
t.Errorf("Shutdown made the calls %q past its deadline, want no %s "+
"and no %s", store.calls, callCheckpoint, callClose)
}
}
// closeOrderStore is a store that records, in order, the AS writes,
// checkpoints and closes it is given. It passes the AS writes on to Store.
type closeOrderStore struct {
database.Store
callsMu sync.Mutex
calls []string
}
// The calls closeOrderStore records.
const (
callWrite = "AS write"
callCheckpoint = "checkpoint"
callClose = "close"
)
func (s *closeOrderStore) GetOrCreateASNBatch(
asns map[int]database.SeenTimes,
) error {
s.record(callWrite)
return s.Store.GetOrCreateASNBatch(asns)
}
func (s *closeOrderStore) Checkpoint(_ context.Context) error {
s.record(callCheckpoint)
return nil
}
func (s *closeOrderStore) Close() error {
s.record(callClose)
return nil
}
func (s *closeOrderStore) record(call string) {
s.callsMu.Lock()
defer s.callsMu.Unlock()
s.calls = append(s.calls, call)
}
// TestShutdownCheckpointsAndClosesTheDatabaseLast hands the AS handler one
// message and shuts down with time to spare. The handler's last write must
// come first, then a checkpoint, then the close. Before the fix Shutdown
// neither checkpointed nor closed the database.
func TestShutdownCheckpointsAndClosesTheDatabaseLast(t *testing.T) {
t.Parallel()
const deadline = 10 * time.Second
store := &closeOrderStore{Store: newMockStore()}
log := logger.New()
cfg := &config.Config{}
s := streamer.New(log, metrics.New())
rw := New(Dependencies{
DB: store,
Streamer: s,
Server: server.New(store, s, log, cfg),
Logger: log,
Config: cfg,
})
rw.dbHandler = NewASHandler(store, log)
rw.dbHandler.HandleMessage(&ristypes.RISMessage{
Path: ristypes.ASPath{Sequence: []int{testASNA}},
})
ctx, cancel := context.WithTimeout(t.Context(), deadline)
defer cancel()
rw.Shutdown(ctx)
store.callsMu.Lock()
defer store.callsMu.Unlock()
want := []string{callWrite, callCheckpoint, callClose}
if !slices.Equal(store.calls, want) {
t.Errorf("Shutdown made the calls %q, want %q", store.calls, want)
}
}
internal/routewatch/ashandler.go
+27 -55
View File
@@ -1,7 +1,6 @@
package routewatch
import (
"context"
"slices"
"sync"
"time"
@@ -40,23 +39,18 @@ type ASHandler struct {
db database.Store
logger *logger.Logger
// Batching. mu guards batch and writing only: the flush goroutine takes
// the batch under mu and writes it after releasing mu, so HandleMessage
// waits for a database write only while batch is at asnBatchMaxSize.
// Batching. mu guards batch only: the flush goroutine takes the batch
// under mu and writes it after releasing mu, so HandleMessage waits for
// a database write only while batch is at asnBatchMaxSize.
mu sync.Mutex
batch []asnOp
// writing is how many operations the flush goroutine took from batch
// and has not finished writing.
writing int
// batchTaken is signalled when the flush goroutine takes batch; its L is
// &mu.
batchTaken sync.Cond
batchFullCh chan struct{} // wakes the flush goroutine when batch is full
lastFlush time.Time // used only by the flush goroutine
// stopCh hands Stop's context to the flush goroutine. It holds one, so
// Stop does not wait for a write in progress to hand it over.
stopCh chan context.Context
done chan struct{} // closed when the flush goroutine returns
stopCh chan struct{}
wg sync.WaitGroup
}
type asnOp struct {
@@ -76,12 +70,12 @@ func NewASHandler(db database.Store, logger *logger.Logger) *ASHandler {
batch: make([]asnOp, 0, asnBatchSize),
batchFullCh: make(chan struct{}, 1),
lastFlush: time.Now(),
stopCh: make(chan context.Context, 1),
done: make(chan struct{}),
stopCh: make(chan struct{}),
}
h.batchTaken.L = &h.mu
// Start the flush timer goroutine
h.wg.Add(1)
go h.flushLoop()
return h
@@ -153,32 +147,20 @@ func (h *ASHandler) HandleMessage(msg *ristypes.RISMessage) {
}
}
// Stop stops the flush goroutine, which writes the operations the handler holds
// before it returns, and waits for it until ctx ends. Once ctx has ended the
// flush goroutine skips that last write, and a write still running is lost
// when the process exits, so Stop logs how many operations are not written.
func (h *ASHandler) Stop(ctx context.Context) {
h.stopCh <- ctx
select {
case <-h.done:
case <-ctx.Done():
}
h.mu.Lock()
unwritten := h.writing + len(h.batch)
h.mu.Unlock()
if unwritten > 0 {
h.logger.Error("Stop deadline passed before the AS handler wrote "+
"its updates", "updates", unwritten)
}
// Stop gracefully shuts down the ASHandler by signaling the background flush
// goroutine to terminate and waiting for it to complete. Any pending ASN
// operations in the current batch are flushed to the database before Stop
// returns. This method should be called during application shutdown to ensure
// no data is lost.
func (h *ASHandler) Stop() {
close(h.stopCh)
h.wg.Wait()
}
// flushLoop runs in a goroutine, the only one that writes the batch: it
// flushes when the batch is full, periodically, and once more on Stop.
func (h *ASHandler) flushLoop() {
defer close(h.done)
defer h.wg.Done()
ticker := time.NewTicker(asnBatchTimeout)
defer ticker.Stop()
@@ -191,11 +173,9 @@ func (h *ASHandler) flushLoop() {
}
case <-h.batchFullCh:
h.flushBatch()
case ctx := <-h.stopCh:
// Final flush, unless the stop deadline has passed
if ctx.Err() == nil {
h.flushBatch()
}
case <-h.stopCh:
// Final flush
h.flushBatch()
return
}
@@ -203,8 +183,7 @@ func (h *ASHandler) flushLoop() {
}
// takeBatch returns the pending operations and leaves an empty batch in their
// place for HandleMessage to fill. They count in writing until flushBatch has
// written them.
// place for HandleMessage to fill.
func (h *ASHandler) takeBatch() []asnOp {
h.mu.Lock()
defer h.mu.Unlock()
@@ -214,7 +193,6 @@ func (h *ASHandler) takeBatch() []asnOp {
}
batch := h.batch
h.writing = len(batch)
h.batch = make([]asnOp, 0, asnBatchSize)
h.batchTaken.Broadcast()
@@ -229,26 +207,20 @@ func (h *ASHandler) flushBatch() {
return
}
// The earliest and the latest time the batch saw each AS. The batch holds
// several updates for an AS, from many peers, so a new AS is stored first
// seen at the earliest of them and last seen at the latest.
asns := make(map[int]database.SeenTimes)
// Process ASNs first (deduped)
asnMap := make(map[int]time.Time)
for _, op := range batch {
seen := asns[op.number]
seen.Add(op.timestamp)
asns[op.number] = seen
if existing, ok := asnMap[op.number]; !ok || op.timestamp.After(existing) {
asnMap[op.number] = op.timestamp
}
}
// Process all ASNs in a single batch transaction
err := h.db.GetOrCreateASNBatch(asns)
err := h.db.GetOrCreateASNBatch(asnMap)
if err != nil {
h.logger.Error("Failed to process ASN batch",
"error", err, "count", len(asns))
"error", err, "count", len(asnMap))
}
h.mu.Lock()
h.writing = 0
h.mu.Unlock()
h.lastFlush = time.Now()
}
internal/routewatch/ashandler_internal_test.go
-68
View File
@@ -1,68 +0,0 @@
package routewatch
import (
"maps"
"slices"
"testing"
"git.eeqj.de/sneak/routewatch/internal/logger"
)
// TestASHandlerStoresTheEarliestAndLatestTimesOfANewAS handles two updates
// whose path carries an AS the database does not hold yet, the second 5
// seconds after the first, in one batch into a real database. The AS is stored
// first seen at the time of the first update and last seen at the time of the
// second.
func TestASHandlerStoresTheEarliestAndLatestTimesOfANewAS(t *testing.T) {
t.Parallel()
db := newTestDatabase(t)
// Built without NewASHandler's flush goroutine, so the test flushes itself.
h := &ASHandler{db: db, logger: logger.New()}
first := peerAnnouncement(testPeerIP, testV4Prefix)
second := peerAnnouncement(testPeerIP, testV4Prefix)
second.ParsedTimestamp = first.ParsedTimestamp.Add(secondUpdateAfter)
h.HandleMessage(first)
h.HandleMessage(second)
h.flushBatch()
asn, _, err := db.GetASDetails(testASNA)
if err != nil {
t.Fatalf("GetASDetails(%d): %v", testASNA, err)
}
if !asn.FirstSeen.Equal(first.ParsedTimestamp) {
t.Errorf("first_seen of AS%d = %s, want %s", testASNA,
asn.FirstSeen, first.ParsedTimestamp)
}
if !asn.LastSeen.Equal(second.ParsedTimestamp) {
t.Errorf("last_seen of AS%d = %s, want %s", testASNA,
asn.LastSeen, second.ParsedTimestamp)
}
}
// TestASHandlerRecordsNothingFromAPathWithASZero handles an update whose AS
// path holds AS 0 between two other ASes. RFC 7607 makes such a path
// malformed, so neither AS is recorded.
func TestASHandlerRecordsNothingFromAPathWithASZero(t *testing.T) {
t.Parallel()
store := newMockStore()
// Built without NewASHandler's flush goroutine, so the test flushes itself.
h := &ASHandler{db: store, logger: logger.New()}
msg := peerAnnouncement(testPeerIP, testV4Prefix)
msg.Path = decodePath(t, testASNA, 0, testASNB)
h.HandleMessage(msg)
h.flushBatch()
if len(store.ASNs) > 0 {
t.Errorf("ASes recorded = %v, want none",
slices.Sorted(maps.Keys(store.ASNs)))
}
}
internal/routewatch/asnfetcher.go
+22 -48
View File
@@ -43,7 +43,7 @@ type ASNFetcher struct {
logger *slog.Logger
immediateQueue chan int
stopCh chan struct{}
done chan struct{} // closed when the fetcher goroutine returns
wg sync.WaitGroup
// fetchMu ensures only one fetch runs at a time
fetchMu sync.Mutex
@@ -53,8 +53,7 @@ type ASNFetcher struct {
currentInterval time.Duration
consecutiveFails int
// hourly stats tracking; now is the clock, time.Now except in tests
now func() time.Time
// hourly stats tracking
statsMu sync.Mutex
successTimes []time.Time
errorTimes []time.Time
@@ -68,9 +67,7 @@ func NewASNFetcher(db database.Store, logger *slog.Logger) *ASNFetcher {
logger: logger.With("component", "asn_fetcher"),
immediateQueue: make(chan int, immediateQueueSize),
stopCh: make(chan struct{}),
done: make(chan struct{}),
currentInterval: baseInterval,
now: time.Now,
successTimes: make([]time.Time, 0),
errorTimes: make([]time.Time, 0),
}
@@ -79,6 +76,8 @@ func NewASNFetcher(db database.Store, logger *slog.Logger) *ASNFetcher {
// Start begins the background ASN fetcher goroutine. It runs until Stop:
// the end of ctx does not stop it.
func (f *ASNFetcher) Start(ctx context.Context) {
f.wg.Add(1)
go f.run(context.WithoutCancel(ctx))
f.logger.Info("ASN fetcher started",
@@ -88,17 +87,11 @@ func (f *ASNFetcher) Start(ctx context.Context) {
)
}
// Stop stops the fetcher and waits, until ctx ends, for a lookup in progress to
// finish.
func (f *ASNFetcher) Stop(ctx context.Context) {
// Stop gracefully shuts down the fetcher.
func (f *ASNFetcher) Stop() {
close(f.stopCh)
select {
case <-f.done:
f.logger.Info("ASN fetcher stopped")
case <-ctx.Done():
f.logger.Warn("Stop deadline passed before the ASN fetcher stopped")
}
f.wg.Wait()
f.logger.Info("ASN fetcher stopped")
}
// QueueImmediate queues an ASN for immediate WHOIS lookup.
@@ -123,7 +116,7 @@ func (f *ASNFetcher) GetStats() server.ASNFetcherStats {
f.intervalMu.Unlock()
// Prune old entries and count
cutoff := f.now().Add(-statsWindow)
cutoff := time.Now().Add(-statsWindow)
f.successTimes = pruneOldTimes(f.successTimes, cutoff)
f.errorTimes = pruneOldTimes(f.errorTimes, cutoff)
@@ -135,15 +128,16 @@ func (f *ASNFetcher) GetStats() server.ASNFetcherStats {
}
}
// pruneOldTimes drops the times at or before cutoff from the front of times,
// which holds times in the order they were recorded, and returns the rest. It
// copies nothing, so it is cheap enough to run on every append.
// pruneOldTimes removes times older than cutoff and returns the pruned slice.
func pruneOldTimes(times []time.Time, cutoff time.Time) []time.Time {
for len(times) > 0 && !times[0].After(cutoff) {
times = times[1:]
result := make([]time.Time, 0, len(times))
for _, t := range times {
if t.After(cutoff) {
result = append(result, t)
}
}
return times
return result
}
// getInterval returns the current fetch interval.
@@ -171,11 +165,9 @@ func (f *ASNFetcher) recordSuccess() {
}
f.intervalMu.Unlock()
// Record success time for stats, dropping those older than statsWindow
// Record success time for stats
f.statsMu.Lock()
now := f.now()
f.successTimes = pruneOldTimes(f.successTimes, now.Add(-statsWindow))
f.successTimes = append(f.successTimes, now)
f.successTimes = append(f.successTimes, time.Now())
f.statsMu.Unlock()
}
@@ -197,17 +189,15 @@ func (f *ASNFetcher) recordFailure() {
}
f.intervalMu.Unlock()
// Record error time for stats, dropping those older than statsWindow
// Record error time for stats
f.statsMu.Lock()
now := f.now()
f.errorTimes = pruneOldTimes(f.errorTimes, now.Add(-statsWindow))
f.errorTimes = append(f.errorTimes, now)
f.errorTimes = append(f.errorTimes, time.Now())
f.statsMu.Unlock()
}
// run is the main background loop.
func (f *ASNFetcher) run(ctx context.Context) {
defer close(f.done)
defer f.wg.Done()
timer := time.NewTimer(f.getInterval())
defer timer.Stop()
@@ -274,10 +264,7 @@ func (f *ASNFetcher) fetchNextStale(ctx context.Context) {
}
// fetchAndUpdate performs a WHOIS lookup and updates the database.
// Returns true if successful. A failed lookup leaves the AS's row as it was,
// so the AS stays stale. An AS IANA refers to no RIR is stored with no
// WHOIS data, so it is not looked up again until it is stale; that counts
// as neither a success nor a failure.
// Returns true if successful.
func (f *ASNFetcher) fetchAndUpdate(ctx context.Context, asn int) bool {
ctx, cancel := context.WithTimeout(ctx, time.Minute)
defer cancel()
@@ -285,19 +272,6 @@ func (f *ASNFetcher) fetchAndUpdate(ctx context.Context, asn int) bool {
f.logger.Info("Fetching WHOIS data", "asn", asn)
info, err := f.whoisClient.LookupASN(ctx, asn)
if errors.Is(err, whois.ErrNoReferral) {
f.logger.Info("IANA refers the AS to no RIR", "asn", asn)
err = f.db.UpdateASNWHOIS(ctx, &database.ASNWHOISUpdate{ASN: asn})
if err != nil {
f.logger.Error("Failed to update ASN WHOIS data",
"asn", asn, "error", err)
f.recordFailure()
}
return false
}
if err != nil {
f.logger.Error("WHOIS lookup failed", "asn", asn, "error", err)
f.recordFailure()
internal/routewatch/asnfetcher_internal_test.go
-237
View File
@@ -1,237 +0,0 @@
package routewatch
import (
"bufio"
"context"
"errors"
"io"
"net"
"sync"
"syscall"
"testing"
"time"
"git.eeqj.de/sneak/routewatch/internal/database"
"git.eeqj.de/sneak/routewatch/internal/logger"
)
// The WHOIS servers a lookup asks: IANA first, then the RIR IANA refers it
// to.
const (
ianaWHOISServer = "whois.iana.org:43"
ripeWHOISServer = "whois.ripe.net:43"
)
// ianaRefersToRIPE is IANA's answer for an AS in a block it gave to RIPE.
const ianaRefersToRIPE = `refer: whois.ripe.net
as-block: 64496-64511
organisation: Assigned by RIPE NCC
changed: 2002-09-01
source: IANA
`
// ripeAccessDenied is RIPE's answer to a host it no longer serves.
const ripeAccessDenied = `% This is the RIPE Database query service.
%ERROR:201: access denied for 192.0.2.1
`
// ianaPrivateUseBlock is IANA's answer for a private-use AS: its block and
// no referral.
const ianaPrivateUseBlock = `as-block: 64512-65534
organisation: Reserved for Private Use
source: IANA
`
// newFetcherWithWHOISAnswers returns an ASN fetcher on a new database that
// holds testASNA, never looked up. The fetcher's WHOIS client connects to a
// fake server for each address in answers, which reads the query and sends
// answers[address]; connecting to any other address is refused.
func newFetcherWithWHOISAnswers(
t *testing.T, answers map[string]string,
) (*ASNFetcher, *database.Database) {
t.Helper()
db := newTestDatabase(t)
_, err := db.GetOrCreateASN(testASNA, time.Now())
if err != nil {
t.Fatalf("GetOrCreateASN: %v", err)
}
var servers sync.WaitGroup
t.Cleanup(servers.Wait)
f := NewASNFetcher(db, logger.New().Logger)
f.whoisClient.DialContext = func(
_ context.Context, _, address string,
) (net.Conn, error) {
answer, ok := answers[address]
if !ok {
return nil, syscall.ECONNREFUSED
}
client, server := net.Pipe()
servers.Go(func() {
defer func() { _ = server.Close() }()
_, _ = bufio.NewReader(server).ReadString('\n')
_, _ = io.WriteString(server, answer)
})
return client, nil
}
return f, db
}
// TestFetchAndUpdateKeepsTheRowWhenTheLookupFails looks up an AS IANA refers
// to RIPE when RIPE's server refuses the connection, and when it answers only
// with an error. The AS must stay never looked up, and the fetcher must back
// off.
func TestFetchAndUpdateKeepsTheRowWhenTheLookupFails(t *testing.T) {
t.Parallel()
for name, answers := range map[string]map[string]string{
"connection refused": {ianaWHOISServer: ianaRefersToRIPE},
"only an error": {
ianaWHOISServer: ianaRefersToRIPE,
ripeWHOISServer: ripeAccessDenied,
},
} {
t.Run(name, func(t *testing.T) {
t.Parallel()
f, db := newFetcherWithWHOISAnswers(t, answers)
if f.fetchAndUpdate(t.Context(), testASNA) {
t.Error("fetchAndUpdate reported success")
}
stats, err := db.GetWHOISStats(t.Context(), whoisStaleThreshold)
if err != nil {
t.Fatalf("GetWHOISStats: %v", err)
}
if stats.NeverFetched != 1 {
t.Errorf("never looked up = %d, want 1: the failed lookup "+
"was stored", stats.NeverFetched)
}
got := f.GetStats()
if got.CurrentInterval != baseInterval*backoffMultiplier ||
got.ConsecutiveFails != 1 {
t.Errorf("interval %v after %d failures, want %v after 1",
got.CurrentInterval, got.ConsecutiveFails,
baseInterval*backoffMultiplier)
}
})
}
}
// TestFetchAndUpdateMarksAnASWithoutReferral looks up a private-use AS, which
// IANA refers to no RIR. The AS must not be looked up again until it is
// stale, and the lookup counts as neither a success nor a failure.
func TestFetchAndUpdateMarksAnASWithoutReferral(t *testing.T) {
t.Parallel()
f, db := newFetcherWithWHOISAnswers(t,
map[string]string{ianaWHOISServer: ianaPrivateUseBlock})
f.fetchAndUpdate(t.Context(), testASNA)
asn, err := db.GetNextStaleASN(t.Context(), whoisStaleThreshold)
if !errors.Is(err, database.ErrNoStaleASN) {
t.Errorf("GetNextStaleASN = %d, %v; want ErrNoStaleASN", asn, err)
}
got := f.GetStats()
if got.CurrentInterval != baseInterval || got.SuccessesLastHour != 0 ||
got.ErrorsLastHour != 0 {
t.Errorf("interval %v, %d successes, %d errors; want %v, 0, 0",
got.CurrentInterval, got.SuccessesLastHour, got.ErrorsLastHour,
baseInterval)
}
}
// TestASNFetcherStopReturnsByItsDeadline starts a fetcher, has it look up an
// AS whose WHOIS server does not answer until the test ends, and stops it with
// a short deadline while that lookup is in progress. Stop must return all the
// same instead of waiting for the lookup.
func TestASNFetcherStopReturnsByItsDeadline(t *testing.T) {
t.Parallel()
const stopDeadline = 100 * time.Millisecond
lookupStarted := make(chan struct{})
release := make(chan struct{})
f := NewASNFetcher(newMockStore(), logger.New().Logger)
f.whoisClient.DialContext = func(
context.Context, string, string,
) (net.Conn, error) {
close(lookupStarted)
<-release
return nil, syscall.ECONNREFUSED
}
f.Start(t.Context())
f.QueueImmediate(testASNA)
select {
case <-lookupStarted:
case <-time.After(handleMessageTimeout):
t.Fatal("the fetcher did not start the lookup")
}
ctx, cancel := context.WithTimeout(t.Context(), stopDeadline)
defer cancel()
if !returnsInTime(func() { f.Stop(ctx) }) {
t.Fatalf("Stop with a deadline of %s did not return while a lookup "+
"was in progress", stopDeadline)
}
close(release)
select {
case <-f.done:
case <-time.After(handleMessageTimeout):
t.Fatal("the fetcher goroutine did not return after the lookup ended")
}
}
// TestRecordKeepsOnlyTheStatsWindow records a success and a failure every
// second, by a clock the test sets, far longer than statsWindow, and never
// asks for the stats. The fetcher must keep only the times within the last
// statsWindow.
func TestRecordKeepsOnlyTheStatsWindow(t *testing.T) {
t.Parallel()
const (
records = 100_000
step = time.Second
want = int(statsWindow / step)
)
clock := time.Now()
f := NewASNFetcher(newMockStore(), logger.New().Logger)
f.now = func() time.Time { return clock }
for range records {
clock = clock.Add(step)
f.recordSuccess()
f.recordFailure()
}
if len(f.successTimes) != want || len(f.errorTimes) != want {
t.Errorf("%d success and %d error times kept, want %d of each, "+
"those of the last %s", len(f.successTimes), len(f.errorTimes),
want, statsWindow)
}
}
internal/routewatch/batchwrite_internal_test.go
+7 -227
View File
@@ -1,10 +1,6 @@
package routewatch
import (
"bytes"
"context"
"encoding/json"
"log/slog"
"slices"
"sync"
"testing"
@@ -26,9 +22,8 @@ const (
)
// blockingStore is a mockStore whose batch writes for the AS, peer and prefix
// handlers, delete of a peer's routes and peering write wait until the test
// calls releaseWrites. writeStarted is closed when the first of them starts. It
// records what each batch write was given.
// handlers wait until the test calls releaseWrites. writeStarted is closed when
// the first of them starts. It records what each write was given.
type blockingStore struct {
*mockStore
@@ -66,7 +61,7 @@ func (s *blockingStore) UpsertLiveRouteBatch(routes []*database.LiveRoute) error
return nil
}
func (s *blockingStore) GetOrCreateASNBatch(asns map[int]database.SeenTimes) error {
func (s *blockingStore) GetOrCreateASNBatch(asns map[int]time.Time) error {
s.waitForRelease()
s.mu.Lock()
defer s.mu.Unlock()
@@ -90,20 +85,6 @@ func (s *blockingStore) UpdatePeerBatch(peers map[string]database.PeerUpdate) er
return nil
}
func (s *blockingStore) DeleteLiveRoutesOfPeer(peerIP string) (int, error) {
s.waitForRelease()
return s.mockStore.DeleteLiveRoutesOfPeer(peerIP)
}
func (s *blockingStore) RecordPeering(
asA, asB int, seen database.SeenTimes,
) error {
s.waitForRelease()
return s.mockStore.RecordPeering(asA, asB, seen)
}
func (s *blockingStore) waitForRelease() {
s.startOnce.Do(func() { close(s.writeStarted) })
<-s.release
@@ -115,40 +96,6 @@ func (s *blockingStore) releaseWrites() {
s.releaseOnce.Do(func() { close(s.release) })
}
// stopLog holds what a handler's Stop logs it did not write.
//
//nolint:tagliatelle // field names are the keys the handlers log
type stopLog struct {
Msg string `json:"msg"`
Updates int `json:"updates"`
PeersDown int `json:"peers_down"`
Paths int `json:"paths"`
}
// newErrorLogger returns a logger that writes only errors, as JSON lines, to
// logs.
func newErrorLogger(logs *bytes.Buffer) *logger.Logger {
return &logger.Logger{Logger: slog.New(slog.NewJSONHandler(logs,
&slog.HandlerOptions{Level: slog.LevelError}))}
}
// lastLogged decodes the last line in logs, which is the one Stop logs: a
// handler logs nothing after it.
func lastLogged(t *testing.T, logs *bytes.Buffer) stopLog {
t.Helper()
lines := bytes.Split(bytes.TrimSpace(logs.Bytes()), []byte("\n"))
var logged stopLog
err := json.Unmarshal(lines[len(lines)-1], &logged)
if err != nil {
t.Fatalf("want the error Stop logs, got %q: %v", logs.String(), err)
}
return logged
}
// returnsInTime reports whether f returns within handleMessageTimeout.
func returnsInTime(f func()) bool {
done := make(chan struct{})
@@ -204,14 +151,14 @@ func handleWhileAWriteIsBlocked(
store *blockingStore,
h interface {
HandleMessage(msg *ristypes.RISMessage)
Stop(ctx context.Context)
Stop()
},
fill []*ristypes.RISMessage,
extra *ristypes.RISMessage,
) {
t.Helper()
defer h.Stop(t.Context())
defer h.Stop()
defer store.releaseWrites()
startBlockedWrite(t, store, h, fill)
@@ -231,14 +178,14 @@ func handleAtMaxBatchSize(
store *blockingStore,
h interface {
HandleMessage(msg *ristypes.RISMessage)
Stop(ctx context.Context)
Stop()
},
fill, toMax []*ristypes.RISMessage,
extra *ristypes.RISMessage,
) {
t.Helper()
defer h.Stop(t.Context())
defer h.Stop()
defer store.releaseWrites()
startBlockedWrite(t, store, h, fill)
@@ -270,173 +217,6 @@ func handleAtMaxBatchSize(
}
}
// stopWhileAWriteIsBlocked starts a blocked write with the messages in fill,
// hands h the message extra, and stops h with a deadline that passes while
// store holds the write. Stop must return all the same. Then it lets the write
// finish and waits until done, closed when h's flush goroutine returns. Past
// the deadline that goroutine skips its last write, so extra is not written,
// and Stop logs as not written both extra and what the held write took.
func stopWhileAWriteIsBlocked(
t *testing.T,
store *blockingStore,
h interface {
HandleMessage(msg *ristypes.RISMessage)
Stop(ctx context.Context)
},
done <-chan struct{},
fill []*ristypes.RISMessage,
extra *ristypes.RISMessage,
) {
t.Helper()
const stopDeadline = 100 * time.Millisecond
defer store.releaseWrites()
startBlockedWrite(t, store, h, fill)
h.HandleMessage(extra)
ctx, cancel := context.WithTimeout(t.Context(), stopDeadline)
defer cancel()
if !returnsInTime(func() { h.Stop(ctx) }) {
t.Fatalf("Stop with a deadline of %s did not return while a write "+
"was blocked", stopDeadline)
}
store.releaseWrites()
select {
case <-done:
case <-time.After(handleMessageTimeout):
t.Fatal("the flush goroutine did not return after the write finished")
}
}
func TestPrefixHandlerSkipsItsLastWriteAfterTheStopDeadline(t *testing.T) {
t.Parallel()
const extraPrefix = "198.51.100.0/24"
var logs bytes.Buffer
store := newBlockingStore()
h := NewPrefixHandler(store, newErrorLogger(&logs))
fill := slices.Repeat([]string{testV4Prefix}, prefixBatchSize)
// The peer's session goes down first, so the write held is the delete of
// its routes, and Stop must log that peer down too.
stopWhileAWriteIsBlocked(t, store, h, h.done,
[]*ristypes.RISMessage{sessionDown(), prefixAnnouncement(fill)},
prefixAnnouncement([]string{extraPrefix}))
store.mu.Lock()
defer store.mu.Unlock()
if store.prefixes[extraPrefix] {
t.Errorf("%s was written after the stop deadline", extraPrefix)
}
logged := lastLogged(t, &logs)
if logged.Updates != prefixBatchSize+1 || logged.PeersDown != 1 {
t.Errorf("Stop logged %d updates and %d peer downs not written, "+
"want %d and 1", logged.Updates, logged.PeersDown, prefixBatchSize+1)
}
}
func TestASHandlerSkipsItsLastWriteAfterTheStopDeadline(t *testing.T) {
t.Parallel()
// Each message gives an update for its origin AS and one for each AS
// in its path: fill asnBatchSize+1, extra 2.
const wantUpdates = asnBatchSize + 1 + 2
var logs bytes.Buffer
store := newBlockingStore()
h := NewASHandler(store, newErrorLogger(&logs))
fill := ristypes.ASPath{Sequence: slices.Repeat([]int{testASNA}, asnBatchSize)}
stopWhileAWriteIsBlocked(t, store, h, h.done,
[]*ristypes.RISMessage{{Path: fill}},
&ristypes.RISMessage{Path: ristypes.ASPath{Sequence: []int{testASNB}}})
store.mu.Lock()
defer store.mu.Unlock()
if store.asns[testASNB] {
t.Errorf("AS%d was written after the stop deadline", testASNB)
}
if logged := lastLogged(t, &logs); logged.Updates != wantUpdates {
t.Errorf("Stop logged %d updates not written, want %d",
logged.Updates, wantUpdates)
}
}
func TestPeerHandlerSkipsItsLastWriteAfterTheStopDeadline(t *testing.T) {
t.Parallel()
const otherPeerIP = "2001:db8:ffff::2"
var logs bytes.Buffer
store := newBlockingStore()
h := NewPeerHandler(store, newErrorLogger(&logs))
stopWhileAWriteIsBlocked(t, store, h, h.done,
slices.Repeat([]*ristypes.RISMessage{{Peer: testPeerIP}}, peerBatchSize),
&ristypes.RISMessage{Peer: otherPeerIP})
store.mu.Lock()
defer store.mu.Unlock()
if store.peers[otherPeerIP] {
t.Errorf("peer %s was written after the stop deadline", otherPeerIP)
}
if logged := lastLogged(t, &logs); logged.Updates != peerBatchSize+1 {
t.Errorf("Stop logged %d updates not written, want %d",
logged.Updates, peerBatchSize+1)
}
}
// TestPeeringHandlerStopCountsTheWriteInProgress hands a running peering
// handler one AS path and stops it with a deadline that passes while store
// holds Stop's last write, of that path's peering. Stop must log that path as
// not written.
func TestPeeringHandlerStopCountsTheWriteInProgress(t *testing.T) {
t.Parallel()
const stopDeadline = 100 * time.Millisecond
var logs bytes.Buffer
store := newBlockingStore()
defer store.releaseWrites()
h := NewPeeringHandler(store, newErrorLogger(&logs))
handle(h, time.Now().UTC(), testASNA, testASNB)
ctx, cancel := context.WithTimeout(t.Context(), stopDeadline)
defer cancel()
if !returnsInTime(func() { h.Stop(ctx) }) {
t.Fatalf("Stop with a deadline of %s did not return while a write "+
"was blocked", stopDeadline)
}
store.releaseWrites()
select {
case <-h.done:
case <-time.After(handleMessageTimeout):
t.Fatal("the processing goroutine did not return after the write " +
"finished")
}
if logged := lastLogged(t, &logs); logged.Paths != 1 {
t.Errorf("Stop logged %d paths not written, want 1", logged.Paths)
}
}
// prefixAnnouncement returns a message in which testPeerIP announces prefixes
// with origin AS testASNA.
func prefixAnnouncement(prefixes []string) *ristypes.RISMessage {
internal/routewatch/cli.go
+23 -17
View File
@@ -3,8 +3,10 @@ package routewatch
import (
"context"
"os"
"os/signal"
"runtime"
"strings"
"syscall"
"time"
"git.eeqj.de/sneak/routewatch/internal/logger"
@@ -12,14 +14,8 @@ import (
)
const (
// shutdownTimeout is how long Shutdown has after a stop signal. upaas,
// like docker stop by default, kills the container 10 seconds after the
// stop signal (README "Running under upaas").
shutdownTimeout = 8 * time.Second
// stopTimeout is fx's limit on the stop hook, past which fx exits with
// status 1: shutdownTimeout and time for Shutdown to close the HTTP
// server after its deadline.
stopTimeout = 9 * time.Second
// shutdownTimeout is the maximum time allowed for graceful shutdown
shutdownTimeout = 60 * time.Second
// debugInterval is how often to log debug stats
debugInterval = 60 * time.Second
// bytesPerMB is bytes per megabyte
@@ -60,14 +56,14 @@ func logDebugStats(logger *logger.Logger) {
}
// CLIEntry is the main entry point for the routewatch command-line interface.
// It initializes the application using the fx dependency injection framework
// and starts the RouteWatch service. It blocks until fx receives SIGINT or
// SIGTERM or RouteWatch stops on an error; fx then runs the stop hook and sets
// the exit code.
// It initializes the application using the fx dependency injection framework,
// sets up signal handling for graceful shutdown, and starts the RouteWatch
// service. This function blocks until the application receives a shutdown
// signal or encounters a fatal error.
func CLIEntry() {
app := fx.New(
fx.Options(getModule()...),
fx.StopTimeout(stopTimeout),
fx.StopTimeout(shutdownTimeout), // Allow 60 seconds for graceful shutdown
fx.Invoke(func(
lc fx.Lifecycle,
rw *RouteWatch,
@@ -79,6 +75,20 @@ func CLIEntry() {
// Start debug stats logging
go logDebugStats(logger)
// Handle shutdown signals
sigCh := make(chan os.Signal, 1)
signal.Notify(sigCh, os.Interrupt, syscall.SIGTERM)
go func() {
<-sigCh
logger.Info("Received shutdown signal")
err := shutdowner.Shutdown()
if err != nil {
logger.Error("Failed to shutdown gracefully", "error", err)
}
}()
go func() {
err := rw.Run(ctx)
if err != nil {
@@ -96,10 +106,6 @@ func CLIEntry() {
},
OnStop: func(ctx context.Context) error {
logger.Info("Shutting down RouteWatch")
ctx, cancel := context.WithTimeout(ctx, shutdownTimeout)
defer cancel()
rw.Shutdown(ctx)
return nil
internal/routewatch/dbmaintainer.go
+10 -27
View File
@@ -2,7 +2,6 @@ package routewatch
import (
"context"
"errors"
"log/slog"
"sync"
"time"
@@ -35,18 +34,12 @@ const (
analyzeTimeout = 5 * time.Minute
)
// checkpointIncompleteMessage is logged, at debug level like a completed
// checkpoint, when a checkpoint leaves part of the WAL for a later one: while
// reads overlap, most do. The database warns when the WAL file stays over its
// size limit.
const checkpointIncompleteMessage = "WAL checkpoint incomplete"
// DBMaintainer handles background database maintenance tasks.
type DBMaintainer struct {
db database.Store
logger *slog.Logger
stopCh chan struct{}
done chan struct{} // closed when the maintenance goroutine returns
wg sync.WaitGroup
// Stats tracking
statsMu sync.Mutex
@@ -67,13 +60,14 @@ func NewDBMaintainer(db database.Store, logger *slog.Logger) *DBMaintainer {
db: db,
logger: logger.With("component", "db_maintainer"),
stopCh: make(chan struct{}),
done: make(chan struct{}),
}
}
// Start begins the background maintenance goroutine. It runs until Stop:
// the end of ctx does not stop it.
func (m *DBMaintainer) Start(ctx context.Context) {
m.wg.Add(1)
go m.run(context.WithoutCancel(ctx))
m.logger.Info("Database maintainer started",
@@ -83,23 +77,16 @@ func (m *DBMaintainer) Start(ctx context.Context) {
)
}
// Stop stops the maintainer and waits, until ctx ends, for a maintenance task
// in progress to finish.
func (m *DBMaintainer) Stop(ctx context.Context) {
// Stop gracefully shuts down the maintainer.
func (m *DBMaintainer) Stop() {
close(m.stopCh)
select {
case <-m.done:
m.logger.Info("Database maintainer stopped")
case <-ctx.Done():
m.logger.Warn("Stop deadline passed before the database maintainer " +
"stopped")
}
m.wg.Wait()
m.logger.Info("Database maintainer stopped")
}
// run is the main background loop.
func (m *DBMaintainer) run(ctx context.Context) {
defer close(m.done)
defer m.wg.Done()
// Use different timers for each task
checkpointTimer := time.NewTimer(checkpointInterval)
@@ -136,13 +123,9 @@ func (m *DBMaintainer) runCheckpoint(ctx context.Context) {
duration, err := runWithTimeout(ctx, checkpointTimeout, m.db.Checkpoint)
m.recordResult(&m.lastCheckpoint, &m.lastCheckpointError, &m.checkpointCount, err)
switch {
case errors.Is(err, database.ErrCheckpointIncomplete):
m.logger.Debug(checkpointIncompleteMessage,
"error", err, "duration", duration)
case err != nil:
if err != nil {
m.logger.Error("WAL checkpoint failed", "error", err, "duration", duration)
default:
} else {
m.logger.Debug("WAL checkpoint completed", "duration", duration)
}
}
internal/routewatch/dbmaintainer_internal_test.go
-120
View File
@@ -1,120 +0,0 @@
package routewatch
import (
"bytes"
"context"
"encoding/json"
"log/slog"
"testing"
"time"
"git.eeqj.de/sneak/routewatch/internal/database"
)
// incompleteCheckpointStore is a store whose every checkpoint leaves part of
// the WAL uncopied.
type incompleteCheckpointStore struct{ *mockStore }
// Checkpoint reports the checkpoint incomplete.
func (incompleteCheckpointStore) Checkpoint(context.Context) error {
return database.ErrCheckpointIncomplete
}
// TestIncompleteCheckpointNotLoggedAsCompleted runs a checkpoint that leaves
// part of the WAL uncopied and checks that the maintainer logs it as
// incomplete, not as completed, and does not count it as a checkpoint that
// completed.
func TestIncompleteCheckpointNotLoggedAsCompleted(t *testing.T) {
t.Parallel()
var logs bytes.Buffer
logger := slog.New(slog.NewJSONHandler(&logs,
&slog.HandlerOptions{Level: slog.LevelDebug}))
m := NewDBMaintainer(incompleteCheckpointStore{newMockStore()}, logger)
m.runCheckpoint(t.Context())
var logged struct {
Msg string `json:"msg"`
}
err := json.Unmarshal(logs.Bytes(), &logged)
if err != nil {
t.Fatalf("want one log message, got %q: %v", logs.String(), err)
}
if logged.Msg != checkpointIncompleteMessage {
t.Errorf("logged %q, want %q", logged.Msg, checkpointIncompleteMessage)
}
if m.checkpointCount != 0 {
t.Errorf("checkpointCount = %d after an incomplete checkpoint, want 0",
m.checkpointCount)
}
}
// heldCheckpointStore is a store whose checkpoint closes started and then
// waits until release is closed.
type heldCheckpointStore struct {
*mockStore
started chan struct{}
release chan struct{}
}
// Checkpoint waits until release is closed.
func (s heldCheckpointStore) Checkpoint(context.Context) error {
close(s.started)
<-s.release
return nil
}
// TestDBMaintainerStopReturnsByItsDeadline stops the maintainer with a short
// deadline while its goroutine runs a checkpoint that the store holds until
// the test ends. Stop must return all the same instead of waiting for the
// checkpoint.
func TestDBMaintainerStopReturnsByItsDeadline(t *testing.T) {
t.Parallel()
const stopDeadline = 100 * time.Millisecond
store := heldCheckpointStore{
mockStore: newMockStore(),
started: make(chan struct{}),
release: make(chan struct{}),
}
m := NewDBMaintainer(store, slog.Default())
// Start would run the first checkpoint only after checkpointInterval.
// This goroutine stands for the one Start runs and checkpoints at once.
go func() {
defer close(m.done)
m.runCheckpoint(t.Context())
}()
select {
case <-store.started:
case <-time.After(handleMessageTimeout):
t.Fatal("the maintainer did not start the checkpoint")
}
ctx, cancel := context.WithTimeout(t.Context(), stopDeadline)
defer cancel()
if !returnsInTime(func() { m.Stop(ctx) }) {
t.Fatalf("Stop with a deadline of %s did not return while a "+
"checkpoint was in progress", stopDeadline)
}
close(store.release)
select {
case <-m.done:
case <-time.After(handleMessageTimeout):
t.Fatal("the maintenance goroutine did not return after the " +
"checkpoint ended")
}
}
internal/routewatch/peerhandler.go
+35 -63
View File
@@ -5,7 +5,6 @@ package routewatch
// peer state in the database.
import (
"context"
"strconv"
"sync"
"time"
@@ -37,30 +36,24 @@ const (
// PeerHandler tracks BGP peers from UPDATE messages using batched operations.
// It maintains a queue of peer updates and periodically flushes them to the database
// in batches to improve performance. The handler writes one row per peer IP
// address for each batch, with the earliest and the latest time of its updates.
// in batches to improve performance. The handler deduplicates peer updates within
// each batch, keeping only the most recent update for each peer IP address.
type PeerHandler struct {
db database.Store
logger *logger.Logger
// Batching. mu guards peerBatch and writing only: the flush goroutine
// takes the batch under mu and writes it after releasing mu, so
// HandleMessage waits for a database write only while peerBatch is at
// peerBatchMaxSize.
// Batching. mu guards peerBatch only: the flush goroutine takes the batch
// under mu and writes it after releasing mu, so HandleMessage waits for
// a database write only while peerBatch is at peerBatchMaxSize.
mu sync.Mutex
peerBatch []peerUpdate
// writing is how many updates the flush goroutine took from peerBatch
// and has not finished writing.
writing int
// batchTaken is signalled when the flush goroutine takes peerBatch; its L
// is &mu.
batchTaken sync.Cond
batchFullCh chan struct{} // wakes the flush goroutine when peerBatch is full
lastFlush time.Time // used only by the flush goroutine
// stopCh hands Stop's context to the flush goroutine. It holds one, so
// Stop does not wait for a write in progress to hand it over.
stopCh chan context.Context
done chan struct{} // closed when the flush goroutine returns
stopCh chan struct{}
wg sync.WaitGroup
}
type peerUpdate struct {
@@ -81,12 +74,12 @@ func NewPeerHandler(db database.Store, logger *logger.Logger) *PeerHandler {
peerBatch: make([]peerUpdate, 0, peerBatchSize),
batchFullCh: make(chan struct{}, 1),
lastFlush: time.Now(),
stopCh: make(chan context.Context, 1),
done: make(chan struct{}),
stopCh: make(chan struct{}),
}
h.batchTaken.L = &h.mu
// Start the flush timer goroutine
h.wg.Add(1)
go h.flushLoop()
return h
@@ -151,32 +144,16 @@ func (h *PeerHandler) HandleMessage(msg *ristypes.RISMessage) {
}
}
// Stop stops the flush goroutine, which writes the updates the handler holds
// before it returns, and waits for it until ctx ends. Once ctx has ended the
// flush goroutine skips that last write, and a write still running is lost
// when the process exits, so Stop logs how many updates are not written.
func (h *PeerHandler) Stop(ctx context.Context) {
h.stopCh <- ctx
select {
case <-h.done:
case <-ctx.Done():
}
h.mu.Lock()
unwritten := h.writing + len(h.peerBatch)
h.mu.Unlock()
if unwritten > 0 {
h.logger.Error("Stop deadline passed before the peer handler wrote "+
"its updates", "updates", unwritten)
}
// Stop gracefully stops the handler and flushes remaining batches
func (h *PeerHandler) Stop() {
close(h.stopCh)
h.wg.Wait()
}
// flushLoop runs in a goroutine, the only one that writes the batch: it
// flushes when the batch is full, periodically, and once more on Stop.
func (h *PeerHandler) flushLoop() {
defer close(h.done)
defer h.wg.Done()
ticker := time.NewTicker(peerBatchTimeout)
defer ticker.Stop()
@@ -189,11 +166,9 @@ func (h *PeerHandler) flushLoop() {
}
case <-h.batchFullCh:
h.flushBatch()
case ctx := <-h.stopCh:
// Final flush, unless the stop deadline has passed
if ctx.Err() == nil {
h.flushBatch()
}
case <-h.stopCh:
// Final flush
h.flushBatch()
return
}
@@ -201,8 +176,7 @@ func (h *PeerHandler) flushLoop() {
}
// takeBatch returns the pending updates and leaves an empty batch in their
// place for HandleMessage to fill. They count in writing until flushBatch has
// written them.
// place for HandleMessage to fill.
func (h *PeerHandler) takeBatch() []peerUpdate {
h.mu.Lock()
defer h.mu.Unlock()
@@ -212,7 +186,6 @@ func (h *PeerHandler) takeBatch() []peerUpdate {
}
batch := h.peerBatch
h.writing = len(batch)
h.peerBatch = make([]peerUpdate, 0, peerBatchSize)
h.batchTaken.Broadcast()
@@ -227,33 +200,32 @@ func (h *PeerHandler) flushBatch() {
return
}
// The earliest and the latest time the batch saw each peer, with the peer
// AS and message type of its latest update. The batch holds several
// updates for a peer, so a new peer is stored first seen at the earliest of
// them and last seen at the latest.
peers := make(map[string]database.PeerUpdate)
// Deduplicate by peer IP, keeping the latest update for each peer
peerMap := make(map[string]peerUpdate)
for _, update := range batch {
peer, ok := peers[update.peerIP]
if !ok || update.timestamp.After(peer.Seen.Last) {
peer.PeerIP = update.peerIP
peer.PeerASN = update.peerASN
peer.MessageType = update.messageType
existing, ok := peerMap[update.peerIP]
if !ok || update.timestamp.After(existing.timestamp) {
peerMap[update.peerIP] = update
}
}
peer.Seen.Add(update.timestamp)
peers[update.peerIP] = peer
// Convert to database format
dbPeerMap := make(map[string]database.PeerUpdate)
for peerIP, update := range peerMap {
dbPeerMap[peerIP] = database.PeerUpdate{
PeerIP: update.peerIP,
PeerASN: update.peerASN,
MessageType: update.messageType,
Timestamp: update.timestamp,
}
}
// Process all peers in a single batch transaction
err := h.db.UpdatePeerBatch(peers)
err := h.db.UpdatePeerBatch(dbPeerMap)
if err != nil {
h.logger.Error("Failed to process peer batch",
"error", err, "count", len(peers))
"error", err, "count", len(dbPeerMap))
}
h.mu.Lock()
h.writing = 0
h.mu.Unlock()
h.lastFlush = time.Now()
}
internal/routewatch/peerhandler_internal_test.go
-59
View File
@@ -1,59 +0,0 @@
package routewatch
import (
"testing"
"git.eeqj.de/sneak/routewatch/internal/database"
"git.eeqj.de/sneak/routewatch/internal/logger"
)
// peerBatchStore keeps the peers of the last UpdatePeerBatch call. The
// database has no read of the peer table, so the peer handler's test checks
// what the handler writes.
type peerBatchStore struct {
*mockStore
peers map[string]database.PeerUpdate
}
func (s *peerBatchStore) UpdatePeerBatch(
peers map[string]database.PeerUpdate,
) error {
s.peers = peers
return nil
}
// TestPeerHandlerWritesTheEarliestAndLatestTimesOfANewPeer handles two updates
// from one peer, the second 5 seconds after the first, in one batch. The peer
// handler writes the peer first seen at the time of the first update and last
// seen at the time of the second.
func TestPeerHandlerWritesTheEarliestAndLatestTimesOfANewPeer(t *testing.T) {
t.Parallel()
store := &peerBatchStore{mockStore: newMockStore()}
// Built without NewPeerHandler's flush goroutine, so the test flushes
// itself.
h := &PeerHandler{db: store, logger: logger.New()}
first := peerAnnouncement(testPeerIP, testV4Prefix)
second := peerAnnouncement(testPeerIP, testV4Prefix)
second.ParsedTimestamp = first.ParsedTimestamp.Add(secondUpdateAfter)
h.HandleMessage(first)
h.HandleMessage(second)
h.flushBatch()
seen := store.peers[testPeerIP].Seen
if !seen.First.Equal(first.ParsedTimestamp) {
t.Errorf("first_seen of peer %s = %s, want %s", testPeerIP,
seen.First, first.ParsedTimestamp)
}
if !seen.Last.Equal(second.ParsedTimestamp) {
t.Errorf("last_seen of peer %s = %s, want %s", testPeerIP,
seen.Last, second.ParsedTimestamp)
}
}
internal/routewatch/peeringhandler.go
+36 -73
View File
@@ -1,7 +1,6 @@
package routewatch
import (
"context"
"encoding/json"
"sync"
"time"
@@ -42,19 +41,13 @@ type PeeringHandler struct {
db database.Store
logger *logger.Logger
// In-memory AS path tracking: each path, JSON-encoded, with the earliest
// and the latest time an update carried it.
// In-memory AS path tracking
mu sync.Mutex
asPaths map[string]database.SeenTimes
droppedPaths int // paths dropped since the last run because the map was full
// writing is how many paths processPeerings took from asPaths and has
// not finished writing the peerings of.
writing int
asPaths map[string]time.Time // key is JSON-encoded AS path
droppedPaths int // paths dropped because the map was full
// stopCh hands Stop's context to the processing goroutine. It holds one,
// so Stop does not wait for a write in progress to hand it over.
stopCh chan context.Context
done chan struct{} // closed when the processing goroutine returns
stopCh chan struct{}
wg sync.WaitGroup
}
// NewPeeringHandler creates and initializes a new PeeringHandler with the
@@ -65,12 +58,13 @@ func NewPeeringHandler(db database.Store, logger *logger.Logger) *PeeringHandler
h := &PeeringHandler{
db: db,
logger: logger,
asPaths: make(map[string]database.SeenTimes),
stopCh: make(chan context.Context, 1),
done: make(chan struct{}),
asPaths: make(map[string]time.Time),
stopCh: make(chan struct{}),
}
// Start the periodic processing goroutine
h.wg.Add(1)
go h.processLoop()
return h
@@ -119,18 +113,16 @@ func (h *PeeringHandler) HandleMessage(msg *ristypes.RISMessage) {
key := string(pathJSON)
h.mu.Lock()
defer h.mu.Unlock()
seen, exists := h.asPaths[key]
if !exists && len(h.asPaths) >= maxTrackedPaths {
if _, exists := h.asPaths[key]; exists {
// Already tracked: refresh its timestamp.
h.asPaths[key] = timestamp
} else if len(h.asPaths) >= maxTrackedPaths {
// Map is full; drop this new path and count it.
h.droppedPaths++
return
} else {
h.asPaths[key] = timestamp
}
seen.Add(timestamp)
h.asPaths[key] = seen
h.mu.Unlock()
}
// ProcessPeeringsNow triggers immediate processing of all accumulated AS
@@ -140,33 +132,18 @@ func (h *PeeringHandler) ProcessPeeringsNow() {
h.processPeerings()
}
// Stop stops the processing goroutine, which writes the peerings of the AS
// paths the handler holds before it returns, and waits for it until ctx ends.
// Once ctx has ended the processing goroutine skips that last write, and a
// write still running is lost when the process exits, so Stop logs how many
// paths' peerings are not written.
func (h *PeeringHandler) Stop(ctx context.Context) {
h.stopCh <- ctx
select {
case <-h.done:
case <-ctx.Done():
}
h.mu.Lock()
unwritten := h.writing + len(h.asPaths)
h.mu.Unlock()
if unwritten > 0 {
h.logger.Error("Stop deadline passed before the peering handler wrote "+
"its peerings", "paths", unwritten)
}
// Stop gracefully shuts down the handler by signaling the background
// goroutine to stop and waiting while it processes the remaining AS paths,
// so no peering data is lost during shutdown.
func (h *PeeringHandler) Stop() {
close(h.stopCh)
h.wg.Wait()
}
// processLoop runs periodically to process AS paths into peerings, and once
// more on Stop.
func (h *PeeringHandler) processLoop() {
defer close(h.done)
defer h.wg.Done()
ticker := time.NewTicker(peeringProcessInterval)
defer ticker.Stop()
@@ -175,11 +152,9 @@ func (h *PeeringHandler) processLoop() {
select {
case <-ticker.C:
h.processPeerings()
case ctx := <-h.stopCh:
// Final processing, unless the stop deadline has passed
if ctx.Err() == nil {
h.processPeerings()
}
case <-h.stopCh:
// Final processing
h.processPeerings()
return
}
@@ -190,14 +165,11 @@ func (h *PeeringHandler) processLoop() {
func (h *PeeringHandler) processPeerings() {
// Take the accumulated paths and replace the map with a fresh empty one
// under the lock. Each path is processed exactly once and the memory is
// released, so the map never grows past a single interval's traffic. The
// count of dropped paths starts again with the new map.
// released, so the map never grows past a single interval's traffic.
h.mu.Lock()
paths := h.asPaths
h.asPaths = make(map[string]database.SeenTimes)
h.writing = len(paths)
h.asPaths = make(map[string]time.Time)
dropped := h.droppedPaths
h.droppedPaths = 0
h.mu.Unlock()
if len(paths) == 0 {
@@ -209,13 +181,9 @@ func (h *PeeringHandler) processPeerings() {
low, high int
}
// The earliest and the latest time an update carried each peering. Several
// paths, each carried at several times, can hold a peering, so a new
// peering is stored first seen at the earliest of them and last seen at the
// latest.
peerings := make(map[peeringKey]database.SeenTimes)
peerings := make(map[peeringKey]time.Time)
for pathJSON, pathSeen := range paths {
for pathJSON, timestamp := range paths {
var path []int
err := json.Unmarshal([]byte(pathJSON), &path)
@@ -242,11 +210,10 @@ func (h *PeeringHandler) processPeerings() {
}
key := peeringKey{low: low, high: high}
seen := peerings[key]
seen.Add(pathSeen.First)
seen.Add(pathSeen.Last)
peerings[key] = seen
// Update timestamp if this is newer
if existing, ok := peerings[key]; !ok || timestamp.After(existing) {
peerings[key] = timestamp
}
}
}
@@ -254,8 +221,8 @@ func (h *PeeringHandler) processPeerings() {
start := time.Now()
successCount := 0
for key, seen := range peerings {
err := h.db.RecordPeering(key.low, key.high, seen)
for key, ts := range peerings {
err := h.db.RecordPeering(key.low, key.high, ts)
if err != nil {
h.logger.Error("Failed to record peering",
"as_a", key.low,
@@ -267,10 +234,6 @@ func (h *PeeringHandler) processPeerings() {
}
}
h.mu.Lock()
h.writing = 0
h.mu.Unlock()
h.logger.Info("Processed AS peerings",
"paths", len(paths),
"unique_peerings", len(peerings),
internal/routewatch/peeringhandler_internal_test.go
+11 -121
View File
@@ -31,14 +31,12 @@ type recordingStore struct {
calls int
}
func (r *recordingStore) RecordPeering(
asA, asB int, seen database.SeenTimes,
) error {
func (r *recordingStore) RecordPeering(asA, asB int, ts time.Time) error {
r.mu.Lock()
r.calls++
r.mu.Unlock()
return r.mockStore.RecordPeering(asA, asB, seen)
return r.mockStore.RecordPeering(asA, asB, ts)
}
func (r *recordingStore) callCount() int {
@@ -54,7 +52,8 @@ func newTestHandler(db database.Store) *PeeringHandler {
return &PeeringHandler{
db: db,
logger: logger.New(),
asPaths: make(map[string]database.SeenTimes),
asPaths: make(map[string]time.Time),
stopCh: make(chan struct{}),
}
}
@@ -153,43 +152,6 @@ func TestPeeringHandlerLeavesATrailingASSetOutOfThePeerings(t *testing.T) {
}
}
// TestPeeringHandlerRecordsNothingFromAPathWithASZero handles AS paths that
// hold AS 0. RFC 7607 makes such a path malformed, so no peering is recorded
// from it, not even one between two other ASes of the path.
func TestPeeringHandlerRecordsNothingFromAPathWithASZero(t *testing.T) {
t.Parallel()
tests := []struct {
name string
path []any
}{
{name: "AS 0 between two ASes", path: []any{testASNA, 0, testASNB}},
{
name: "AS 0 after two ASes",
path: []any{testASNA, testASNB, 0, testASNC},
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
t.Parallel()
store := newMockStore()
h := newTestHandler(store)
h.HandleMessage(&ristypes.RISMessage{
Path: decodePath(t, tt.path...),
ParsedTimestamp: time.Now().UTC(),
})
h.ProcessPeeringsNow()
if len(store.Peerings) > 0 {
t.Errorf("peerings recorded = %v, want none",
slices.Sorted(maps.Keys(store.Peerings)))
}
})
}
}
// TestPeeringHandlerProcessesEachRunAndEmpties verifies that a processing run
// empties the path map (the swap) and that a path seen again after a run is
// recorded in the next run too.
@@ -235,7 +197,7 @@ func TestPeeringHandlerStopWritesTheRemainingPaths(t *testing.T) {
h := NewPeeringHandler(store, logger.New())
handle(h, time.Now().UTC(), testASNA, testASNB)
h.Stop(t.Context())
h.Stop()
if got := store.callCount(); got != 1 {
t.Errorf("want 1 RecordPeering call when Stop returns, got %d", got)
@@ -259,10 +221,10 @@ func TestPeeringHandlerCapDropsAndCounts(t *testing.T) {
existing := pathKey(t, testASNA, testASNB)
h.mu.Lock()
h.asPaths[existing] = database.SeenTimes{First: now, Last: now}
h.asPaths[existing] = now
for i := 0; len(h.asPaths) < maxTrackedPaths; i++ {
h.asPaths[strconv.Itoa(i)] = database.SeenTimes{First: now, Last: now}
h.asPaths[strconv.Itoa(i)] = now
}
h.mu.Unlock()
@@ -292,83 +254,11 @@ func TestPeeringHandlerCapDropsAndCounts(t *testing.T) {
}
h.mu.Lock()
gotSeen := h.asPaths[existing]
gotTS := h.asPaths[existing]
h.mu.Unlock()
if !gotSeen.Last.Equal(refreshed) {
t.Fatalf("existing path last seen not refreshed: want %v, got %v",
refreshed, gotSeen.Last)
}
}
// TestPeeringHandlerRunResetsTheDroppedCount fills the map, has the handler
// drop one path, and runs the processing. The count of dropped paths is the
// run's own, so it is 0 again after the run.
func TestPeeringHandlerRunResetsTheDroppedCount(t *testing.T) {
t.Parallel()
h := newTestHandler(newMockStore())
now := time.Now().UTC()
// Paths of one AS carry no peering, so the run records nothing.
h.mu.Lock()
for asn := 1; len(h.asPaths) < maxTrackedPaths; asn++ {
h.asPaths[pathKey(t, asn)] = database.SeenTimes{First: now, Last: now}
}
h.mu.Unlock()
handle(h, now, testASNA, testASNB)
if _, dropped := snapshot(h); dropped != 1 {
t.Fatalf("want dropped count 1 with the map full, got %d", dropped)
}
h.ProcessPeeringsNow()
if _, dropped := snapshot(h); dropped != 0 {
t.Errorf("want dropped count 0 after a run, got %d", dropped)
}
}
// TestPeeringHandlerStoresTheEarliestAndLatestTimesOfANewPeering handles two
// updates whose path carries a peering the database does not hold yet, the
// second 5 seconds after the first, in one run into a real database. The
// peering is stored first seen at the time of the first update and last seen
// at the time of the second.
func TestPeeringHandlerStoresTheEarliestAndLatestTimesOfANewPeering(
t *testing.T,
) {
t.Parallel()
db := newTestDatabase(t)
h := newTestHandler(db)
first := peerAnnouncement(testPeerIP, testV4Prefix)
second := peerAnnouncement(testPeerIP, testV4Prefix)
second.ParsedTimestamp = first.ParsedTimestamp.Add(secondUpdateAfter)
h.HandleMessage(first)
h.HandleMessage(second)
h.ProcessPeeringsNow()
peers, err := db.GetASPeers(testASNA)
if err != nil {
t.Fatalf("GetASPeers(%d): %v", testASNA, err)
}
if len(peers) != 1 || peers[0].ASN != testASNB {
t.Fatalf("peers of AS%d = %+v, want AS%d alone", testASNA, peers,
testASNB)
}
if !peers[0].FirstSeen.Equal(first.ParsedTimestamp) {
t.Errorf("first_seen of peering AS%d-AS%d = %s, want %s", testASNA,
testASNB, peers[0].FirstSeen, first.ParsedTimestamp)
}
if !peers[0].LastSeen.Equal(second.ParsedTimestamp) {
t.Errorf("last_seen of peering AS%d-AS%d = %s, want %s", testASNA,
testASNB, peers[0].LastSeen, second.ParsedTimestamp)
if !gotTS.Equal(refreshed) {
t.Fatalf("existing path timestamp not refreshed: want %v, got %v",
refreshed, gotTS)
}
}
internal/routewatch/prefixhandler.go
+50 -200
View File
@@ -1,7 +1,6 @@
package routewatch
import (
"context"
"net"
"net/netip"
"slices"
@@ -29,9 +28,7 @@ const (
// prefixBatchMaxSize is the most prefix updates the batch holds while the
// flush goroutine writes the previous one: ten batches, about 60 MiB
// (README "Memory"). At this size HandleMessage waits until the flush
// goroutine takes the batch. On top of these, the handler holds at most
// prefixBatchSize updates, about 6 MiB, that a flush kept for the peers
// whose route delete failed (held).
// goroutine takes the batch.
prefixBatchMaxSize = 250000
// prefixBatchTimeout is the maximum time to wait before flushing a batch
@@ -44,45 +41,28 @@ const (
)
// PrefixHandler tracks BGP prefixes and maintains a live routing table in the
// database. Routes are added on announcement and deleted on withdrawal or on an
// announcement whose AS path is empty or holds AS 0 anywhere, and all routes of
// a peer are deleted when its BGP session goes down.
// database. Routes are added on announcement and deleted on withdrawal, and all
// routes of a peer are deleted when its BGP session goes down.
type PrefixHandler struct {
db database.Store
logger *logger.Logger
metrics *metrics.Tracker
// Batching. mu guards batch, held, peersDown, writing and deleting only:
// the flush goroutine takes them under mu and writes them after releasing
// mu, so HandleMessage waits for a database write only while batch is at
// prefixBatchMaxSize.
// Batching. mu guards batch and peersDown only: the flush goroutine takes
// both under mu and writes them after releasing mu, so HandleMessage
// waits for a database write only while batch is at prefixBatchMaxSize.
mu sync.Mutex
batch []prefixUpdate
// held holds the updates that the peers whose route delete failed in the
// last flush sent after their session went down, for the next flush to
// write after it retries the delete. They are not in batch, so they do
// not count toward a full batch: a delete that keeps failing is retried
// at the pace of the flush timer.
held []prefixUpdate
// peersDown holds, each once, the peers whose BGP session went down
// since the batch was taken, and those whose route delete failed in the
// last flush.
// peersDown holds the peers whose BGP session went down since the batch
// was taken.
peersDown []string
// writing is how many updates the flush goroutine took and has not
// finished writing.
writing int
// deleting is how many peers from peersDown the flush goroutine took and
// has not yet deleted the routes of.
deleting int
// batchTaken is signalled when the flush goroutine takes batch; its L is
// &mu.
batchTaken sync.Cond
batchFullCh chan struct{} // wakes the flush goroutine when batch is full
lastFlush time.Time // used only by the flush goroutine
// stopCh hands Stop's context to the flush goroutine. It holds one, so
// Stop does not wait for a write in progress to hand it over.
stopCh chan context.Context
done chan struct{} // closed when the flush goroutine returns
stopCh chan struct{}
wg sync.WaitGroup
}
type prefixUpdate struct {
@@ -105,12 +85,12 @@ func NewPrefixHandler(db database.Store, logger *logger.Logger) *PrefixHandler {
batch: make([]prefixUpdate, 0, prefixBatchSize),
batchFullCh: make(chan struct{}, 1),
lastFlush: time.Now(),
stopCh: make(chan context.Context, 1),
done: make(chan struct{}),
stopCh: make(chan struct{}),
}
h.batchTaken.L = &h.mu
// Start the flush timer goroutine
h.wg.Add(1)
go h.flushLoop()
return h
@@ -212,64 +192,31 @@ func (h *PrefixHandler) HandleMessage(msg *ristypes.RISMessage) {
}
}
// Stop stops the flush goroutine, which writes the updates and peer downs the
// handler holds before it returns, and waits for it until ctx ends. Once ctx
// has ended the flush goroutine skips that last write, and a write still
// running is lost when the process exits. A peer whose route delete fails in
// the last write is not retried either. Stop logs how many updates are not
// written and how many peer downs have not had their routes deleted.
func (h *PrefixHandler) Stop(ctx context.Context) {
h.stopCh <- ctx
select {
case <-h.done:
case <-ctx.Done():
}
h.mu.Lock()
unwritten := h.writing + len(h.batch) + len(h.held)
peersDown := h.deleting + len(h.peersDown)
h.mu.Unlock()
if unwritten == 0 && peersDown == 0 {
return
}
if ctx.Err() != nil {
h.logger.Error("Stop deadline passed before the prefix handler wrote "+
"its updates", "updates", unwritten, "peers_down", peersDown)
return
}
h.logger.Error("The prefix handler stopped without deleting the routes "+
"of peers whose session went down", "updates", unwritten,
"peers_down", peersDown)
// Stop gracefully stops the handler and flushes remaining batches
func (h *PrefixHandler) Stop() {
close(h.stopCh)
h.wg.Wait()
}
// peerSessionDown notes that the peer's BGP session went down, so that the next
// flush deletes all of the peer's live routes before it writes the batch. The
// peer's updates still in the batch or held came before the down, which
// replaces them, so they are dropped; updates the peer sends later stay in the
// batch and are written after the delete. A peer already in peersDown is not
// added again: one delete removes all of its routes.
// peer's updates still in the batch came before the down, which replaces them,
// so they are dropped; updates the peer sends later stay in the batch and are
// written after the delete.
func (h *PrefixHandler) peerSessionDown(peer string) {
h.mu.Lock()
defer h.mu.Unlock()
isPeers := func(update prefixUpdate) bool { return update.peer == peer }
h.batch = slices.DeleteFunc(h.batch, isPeers)
h.held = slices.DeleteFunc(h.held, isPeers)
if !slices.Contains(h.peersDown, peer) {
h.peersDown = append(h.peersDown, peer)
}
h.batch = slices.DeleteFunc(h.batch, func(update prefixUpdate) bool {
return update.peer == peer
})
h.peersDown = append(h.peersDown, peer)
}
// flushLoop runs in a goroutine, the only one that writes the batch: it
// flushes when the batch is full, periodically, and once more on Stop.
func (h *PrefixHandler) flushLoop() {
defer close(h.done)
defer h.wg.Done()
ticker := time.NewTicker(prefixBatchTimeout)
defer ticker.Stop()
@@ -282,38 +229,27 @@ func (h *PrefixHandler) flushLoop() {
}
case <-h.batchFullCh:
h.flushBatch()
case ctx := <-h.stopCh:
// Final flush, unless the stop deadline has passed
if ctx.Err() == nil {
h.flushBatch()
}
case <-h.stopCh:
// Final flush
h.flushBatch()
return
}
}
}
// takeBatch returns the pending updates, the held ones first, and the peers in
// peersDown, and leaves an empty batch, no held updates and no peers in their
// place for HandleMessage to fill. The updates count in writing until
// flushBatch has written them, and the peers in deleting until their routes
// are deleted.
// takeBatch returns the pending updates and the peers whose session went down
// since the last flush, and leaves an empty batch and no peers in their place
// for HandleMessage to fill.
func (h *PrefixHandler) takeBatch() ([]prefixUpdate, []string) {
h.mu.Lock()
defer h.mu.Unlock()
if len(h.held) == 0 && len(h.batch) == 0 && len(h.peersDown) == 0 {
if len(h.batch) == 0 && len(h.peersDown) == 0 {
return nil, nil
}
batch, peersDown := h.batch, h.peersDown
if len(h.held) > 0 {
batch = slices.Concat(h.held, h.batch)
}
h.writing = len(batch)
h.deleting = len(peersDown)
h.held = nil
h.batch = make([]prefixUpdate, 0, prefixBatchSize)
h.peersDown = nil
h.batchTaken.Broadcast()
@@ -333,12 +269,8 @@ func (h *PrefixHandler) flushBatch() {
batchSize := len(batch)
// The routes of a peer whose session went down are deleted first: the
// batch holds only the updates the peer sent after the down. A peer whose
// delete failed is kept, with those updates, for the next flush to retry.
failedPeers := h.deleteLiveRoutesOfPeers(peersDown)
if len(failedPeers) > 0 {
batch = h.keepForNextFlush(batch, failedPeers)
}
// batch holds only the updates the peer sent after the down.
h.deleteLiveRoutesOfPeers(peersDown)
// For each prefix and peer, keep only the last update the feed delivered.
// Updates from different peers for the same prefix are all kept. Arrival
@@ -348,19 +280,8 @@ func (h *PrefixHandler) flushBatch() {
lastUpdates := make(map[prefixAndPeer]prefixUpdate)
// Each prefix the batch names, announced or withdrawn, with the earliest
// timestamp among all of its updates, not only each peer's last one: the
// prefix tables store only first_seen, written when the prefix is new and
// never changed after.
prefixesToUpdate := make(map[string]time.Time)
for _, update := range batch {
lastUpdates[prefixAndPeer{update.prefix, update.peer}] = update
seen, exists := prefixesToUpdate[update.prefix]
if !exists || update.timestamp.Before(seen) {
prefixesToUpdate[update.prefix] = update.timestamp
}
}
// Collect routes to upsert and delete
@@ -369,14 +290,19 @@ func (h *PrefixHandler) flushBatch() {
routesToDelete []database.LiveRouteDeletion
)
// Collect unique prefixes to update
prefixesToUpdate := make(map[string]time.Time)
for _, update := range lastUpdates {
// Track prefix for both announcements and withdrawals
seen, exists := prefixesToUpdate[update.prefix]
if !exists || update.timestamp.After(seen) {
prefixesToUpdate[update.prefix] = update.timestamp
}
// An announcement is stored when its path gives an origin, or when it
// ends in an AS set of several members: that route is stored with
// origin 0 (none). Any other update deletes the peer's route for the
// prefix: a withdrawal, or an announcement whose path is empty or
// holds AS 0 anywhere (the decoder empties such a path). Such an
// announcement replaces the peer's route, as any announcement does,
// with none (RFC 7607 treats AS 0 in a path as a withdrawal).
// origin 0 (none). An announcement with an empty path is not stored.
if update.messageType == "announcement" &&
(update.originASN > 0 || update.endsInSet) {
// Create live route for batch upsert
@@ -384,11 +310,11 @@ func (h *PrefixHandler) flushBatch() {
if route != nil {
routesToUpsert = append(routesToUpsert, route)
}
} else {
} else if update.messageType == "withdrawal" {
// Parse CIDR to get IP version
_, ipVersion, err := parseCIDR(update.prefix)
if err != nil {
h.logger.Error("Failed to parse CIDR for route deletion",
h.logger.Error("Failed to parse CIDR for withdrawal",
"prefix", update.prefix, "error", err)
continue
@@ -405,10 +331,6 @@ func (h *PrefixHandler) flushBatch() {
successCount := h.writeBatch(routesToUpsert, routesToDelete, prefixesToUpdate)
h.mu.Lock()
h.writing = 0
h.mu.Unlock()
elapsed := time.Since(startTime)
h.logger.Debug("Flushed prefix batch",
"batch_size", batchSize,
@@ -422,97 +344,25 @@ func (h *PrefixHandler) flushBatch() {
}
// deleteLiveRoutesOfPeers deletes all live routes of each of peers, the peers
// whose BGP session went down, and returns those whose delete failed. Most
// downs come from peers that have no routes, so only a delete that removed
// routes is logged. A peer whose delete succeeded no longer counts in deleting;
// a failed one counts until keepForNextFlush puts it back in peersDown.
func (h *PrefixHandler) deleteLiveRoutesOfPeers(peers []string) []string {
var failed []string
// whose BGP session went down. Most downs come from peers that have no routes,
// so only a delete that removed routes is logged.
func (h *PrefixHandler) deleteLiveRoutesOfPeers(peers []string) {
for _, peer := range peers {
deleted, err := h.db.DeleteLiveRoutesOfPeer(peer)
if err != nil {
h.logger.Error(
"Failed to delete the live routes of a peer whose session "+
"went down; the next flush retries",
"peer", peer, "routes_deleted", deleted, "error", err)
failed = append(failed, peer)
"Failed to delete the live routes of a peer whose session went down",
"peer", peer, "error", err)
continue
}
h.mu.Lock()
h.deleting--
h.mu.Unlock()
if deleted > 0 {
h.logger.Info(
"Deleted the live routes of a peer whose session went down",
"peer", peer, "routes", deleted)
}
}
return failed
}
// keepForNextFlush puts the peers whose route delete failed back in peersDown,
// and their updates from batch in held, so that the next flush retries each
// delete before it writes the peer's updates; written now, they would be
// deleted by the retry. The updates of a peer whose session went down again
// since the batch was taken came before that down, so they are dropped, as
// peerSessionDown drops them, and the peer is already in peersDown. It returns
// the updates of the other peers, for this flush to write.
//
// While a delete keeps failing, the held updates would grow with every flush,
// so only the last prefixBatchSize of them are kept and the older ones are
// dropped. This never leaves in place a route the peer withdrew: a prefix with
// a kept update keeps its last one, and the retry deletes all of the peer's
// routes, so a prefix whose updates were all dropped has no route from the peer
// until the peer announces it again.
func (h *PrefixHandler) keepForNextFlush(
batch []prefixUpdate, failedPeers []string,
) []prefixUpdate {
var kept, toWrite []prefixUpdate
for _, update := range batch {
if slices.Contains(failedPeers, update.peer) {
kept = append(kept, update)
} else {
toWrite = append(toWrite, update)
}
}
h.mu.Lock()
defer h.mu.Unlock()
kept = slices.DeleteFunc(kept, func(update prefixUpdate) bool {
return slices.Contains(h.peersDown, update.peer)
})
if len(kept) > prefixBatchSize {
h.logger.Error(
"Dropped the oldest updates of peers whose route delete failed",
"peers", failedPeers, "dropped", len(kept)-prefixBatchSize)
// A copy, so that the dropped updates are freed.
kept = slices.Clone(kept[len(kept)-prefixBatchSize:])
}
// takeBatch left held empty, and only this flush fills it. The kept
// updates now count in held, not in writing, and the failed peers in
// peersDown, not in deleting.
h.held = kept
h.writing = len(toWrite)
h.deleting = 0
for _, peer := range failedPeers {
if !slices.Contains(h.peersDown, peer) {
h.peersDown = append(h.peersDown, peer)
}
}
return toWrite
}
// writeBatch writes the collected route upserts, route deletions and prefix
internal/routewatch/prefixhandler_internal_test.go
-429
View File
@@ -1,13 +1,10 @@
package routewatch
import (
"bytes"
"context"
"errors"
"fmt"
"net/netip"
"slices"
"strings"
"sync"
"testing"
"time"
@@ -26,10 +23,6 @@ const (
testV6Prefix = "2001:db8:1::/48"
)
// secondUpdateAfter is how long after the first update the tests of first_seen
// and last_seen send the second, in the same batch.
const secondUpdateAfter = 5 * time.Second
// TestPrefixHandlerStoresPrefixesTheIPLookupFinds runs announcements and
// withdrawals through the prefix handler into a real database and looks the
// addresses up. The prefixes are written the way the feed sends them: IPv6
@@ -87,49 +80,6 @@ func TestPrefixHandlerStoresPrefixesTheIPLookupFinds(t *testing.T) {
}
}
// TestPrefixHandlerStoresANewPrefixFirstSeenAtItsEarliestUpdate announces a
// prefix the database does not hold yet twice in one batch, the second time 5
// seconds after the first, through the prefix handler into a real database. The
// prefix is stored first seen at the time of the first announcement, whether
// the two come from two peers or from one, and the IP lookup answers with that
// time.
func TestPrefixHandlerStoresANewPrefixFirstSeenAtItsEarliestUpdate(
t *testing.T,
) {
t.Parallel()
const otherPeerIP = "2001:db8:ffff::2"
tests := []struct {
name string
firstPeer, secondPeer string
}{
{name: "two peers", firstPeer: testPeerIP, secondPeer: otherPeerIP},
{name: "one peer", firstPeer: testPeerIP, secondPeer: testPeerIP},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
t.Parallel()
first := peerAnnouncement(tt.firstPeer, testV4Prefix)
second := peerAnnouncement(tt.secondPeer, testV4Prefix)
second.ParsedTimestamp = first.ParsedTimestamp.Add(secondUpdateAfter)
db := handleBatches(t, [][]*ristypes.RISMessage{{first, second}})
info, err := db.GetIPInfo("192.0.2.1")
if err != nil {
t.Fatalf("GetIPInfo(192.0.2.1): %v", err)
}
if !info.FirstSeen.Equal(first.ParsedTimestamp) {
t.Errorf("first_seen of %s = %s, want %s", testV4Prefix,
info.FirstSeen, first.ParsedTimestamp)
}
})
}
}
// TestPrefixHandlerAppliesEachPeersLastUpdateInABatch runs updates for one
// prefix through the prefix handler into a real database. All updates carry the
// same timestamp, as updates within one second of the feed do. Within a batch,
@@ -378,133 +328,6 @@ func TestPrefixHandlerDoesNotStoreAnAnnouncementWithAnEmptyPath(t *testing.T) {
assertPeers(t, db, map[string][]string{testV4Prefix: nil})
}
// TestPrefixHandlerRemovesTheRouteOnAnAnnouncementWithoutAnOrigin runs, through
// the prefix handler into a real database, a peer's announcement of a prefix
// with origin AS testASNB and then the peer's announcement of the prefix with
// an AS path that gives no origin: one whose last AS is 0, an empty one, or one
// that holds AS 0 before its end, which decodes as an empty path. As any later
// announcement from the peer does, the second replaces the peer's route, here
// with none, whether it comes in a later batch or follows a withdrawal in one
// batch. The in-memory route count and prefix distribution must match the
// route table.
func TestPrefixHandlerRemovesTheRouteOnAnAnnouncementWithoutAnOrigin(
t *testing.T,
) {
t.Parallel()
const prefixMaskLength = 24
noOrigin := func(path ristypes.ASPath) *ristypes.RISMessage {
msg := peerAnnouncement(testPeerIP, testV4Prefix)
msg.Path = path
return msg
}
endsInZero := ristypes.ASPath{Sequence: []int{testASNA, 0}}
withdrawal := &ristypes.RISMessage{
Type: ristypes.MessageTypeUpdate,
Peer: testPeerIP,
ParsedTimestamp: time.Date(2026, 1, 2, 3, 4, 5, 0, time.UTC),
Withdrawals: []string{testV4Prefix},
}
tests := []struct {
name string
// Handled and flushed as one batch after the announcement with origin
// AS testASNB.
batch []*ristypes.RISMessage
}{
{
name: "a path whose last AS is 0",
batch: []*ristypes.RISMessage{noOrigin(endsInZero)},
},
{
name: "an empty path",
batch: []*ristypes.RISMessage{noOrigin(ristypes.ASPath{})},
},
{
name: "a path with AS 0 before its end",
batch: []*ristypes.RISMessage{
noOrigin(decodePath(t, testASNA, 0, testASNB)),
},
},
{
name: "a withdrawal and then a path whose last AS is 0",
batch: []*ristypes.RISMessage{withdrawal, noOrigin(endsInZero)},
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
t.Parallel()
db := handleBatches(t, [][]*ristypes.RISMessage{
{peerAnnouncement(testPeerIP, testV4Prefix)},
tt.batch,
})
checkLivePeers(t, db, testV4Prefix, prefixMaskLength, nil)
})
}
}
// TestPrefixHandlerDoesNotStoreAPrefixThatDoesNotParse runs an announcement of
// "not-a-prefix" and a withdrawal of "also/bad" through the prefix handler into
// a real database. Neither is stored in the prefix tables: the prefix count in
// the stats stays 0, both as kept in memory and as counted from the tables when
// the database is opened again.
func TestPrefixHandlerDoesNotStoreAPrefixThatDoesNotParse(t *testing.T) {
t.Parallel()
cfg := &config.Config{StateDir: t.TempDir()}
db, err := database.New(cfg, logger.New())
if err != nil {
t.Fatalf("failed to create database: %v", err)
}
// Built without NewPrefixHandler's flush timer, so the test flushes itself.
h := &PrefixHandler{db: db, logger: logger.New()}
withdrawal := &ristypes.RISMessage{
Type: ristypes.MessageTypeUpdate,
Peer: testPeerIP,
ParsedTimestamp: time.Date(2026, 1, 2, 3, 4, 5, 0, time.UTC),
Withdrawals: []string{"also/bad"},
}
h.HandleMessage(peerAnnouncement(testPeerIP, "not-a-prefix"))
h.HandleMessage(withdrawal)
h.flushBatch()
checkNoPrefixes := func(when string) {
t.Helper()
stats, err := db.GetStatsContext(context.Background())
if err != nil {
t.Fatalf("GetStatsContext %s: %v", when, err)
}
if stats.Prefixes != 0 {
t.Errorf("prefix count %s = %d, want 0", when, stats.Prefixes)
}
}
checkNoPrefixes("after the flush")
err = db.Close()
if err != nil {
t.Fatalf("Close: %v", err)
}
db, err = database.New(cfg, logger.New())
if err != nil {
t.Fatalf("failed to open the database again: %v", err)
}
t.Cleanup(func() { _ = db.Close() })
checkNoPrefixes("after opening the database again")
}
// TestStateMessagesGoToThePrefixHandlerNotThePeerHandler checks that STATE
// messages go to the prefix handler, which acts on a down, and not to the peer
// handler, which would record a peer whose session never comes up as active.
@@ -737,258 +560,6 @@ func TestPrefixHandlerHandlesADownWhileAFlushWrites(t *testing.T) {
})
}
// errTestPeerDelete is the error a test store's failing delete of a peer's
// routes returns.
var errTestPeerDelete = errors.New("test: the delete of a peer's routes failed")
// firstPeerDeleteFailsStore is a real database whose first delete of a peer's
// live routes fails, deleting nothing. deletes counts the deletes asked for.
type firstPeerDeleteFailsStore struct {
*database.Database
// whileFailing, if not nil, runs during the delete that fails.
whileFailing func()
deletes int
}
// DeleteLiveRoutesOfPeer fails the first call and deletes the peer's routes on
// every later one.
func (s *firstPeerDeleteFailsStore) DeleteLiveRoutesOfPeer(
peerIP string,
) (int, error) {
s.deletes++
if s.deletes > 1 {
return s.Database.DeleteLiveRoutesOfPeer(peerIP)
}
if s.whileFailing != nil {
s.whileFailing()
}
return 0, errTestPeerDelete
}
// TestPrefixHandlerRetriesAFailedDeleteOfAPeersRoutes runs, through the prefix
// handler into a database whose first delete of a peer's routes fails, one flush
// after another: a peer's announcement, then its session going down and a new
// announcement from the peer, then nothing. The failed delete must be retried
// on the next flush, and the announcement the peer sent after the down written
// only after that retry, so it is kept. In the other cases the peer's session
// goes down again, and the peer announces another prefix, while the delete
// fails or after the flush whose delete failed: the announcement sent between
// the two downs must not be written, and the next flush must delete the peer's
// routes once, not once for the retry and once for the down.
func TestPrefixHandlerRetriesAFailedDeleteOfAPeersRoutes(t *testing.T) {
t.Parallel()
const (
beforeDown = "192.0.2.0/24"
afterDown = "198.51.100.0/24"
afterSecondDown = "203.0.113.0/24"
maskLength = 24
// The delete that fails and its retry, which is also the delete for
// a second down.
wantDeletes = 2
)
tests := []struct {
name string
// Handled while the delete fails, if not nil.
whileFailing []*ristypes.RISMessage
// Handled after the flush whose delete fails, if not nil.
afterFailing []*ristypes.RISMessage
// The peers left with a live route for each prefix.
wantPeers map[string][]string
}{
{
name: "the delete fails once",
wantPeers: map[string][]string{
beforeDown: nil, afterDown: {testPeerIP}, afterSecondDown: nil,
},
},
{
name: "the session goes down again while the delete fails",
whileFailing: []*ristypes.RISMessage{
sessionDown(), peerAnnouncement(testPeerIP, afterSecondDown),
},
wantPeers: map[string][]string{
beforeDown: nil, afterDown: nil, afterSecondDown: {testPeerIP},
},
},
{
name: "the session goes down again after the delete failed",
afterFailing: []*ristypes.RISMessage{
sessionDown(), peerAnnouncement(testPeerIP, afterSecondDown),
},
wantPeers: map[string][]string{
beforeDown: nil, afterDown: nil, afterSecondDown: {testPeerIP},
},
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
t.Parallel()
db := newTestDatabase(t)
store := &firstPeerDeleteFailsStore{Database: db}
// Built without NewPrefixHandler's flush timer, so the test
// flushes itself.
h := &PrefixHandler{db: store, logger: logger.New()}
store.whileFailing = func() {
for _, msg := range tt.whileFailing {
h.HandleMessage(msg)
}
}
h.HandleMessage(peerAnnouncement(testPeerIP, beforeDown))
h.flushBatch()
h.HandleMessage(sessionDown())
h.HandleMessage(peerAnnouncement(testPeerIP, afterDown))
h.flushBatch() // the delete fails
for _, msg := range tt.afterFailing {
h.HandleMessage(msg)
}
h.flushBatch()
if store.deletes != wantDeletes {
t.Errorf("the store was asked for %d deletes of the peer's "+
"routes, want %d", store.deletes, wantDeletes)
}
assertPeers(t, db, tt.wantPeers)
assertCounts(t, db, ipv4Version, 1, []database.PrefixDistribution{
{MaskLength: maskLength, Count: 1},
})
})
}
}
// peerDeleteAlwaysFailsStore is a real database whose every delete of a peer's
// live routes fails, deleting nothing.
type peerDeleteAlwaysFailsStore struct {
*database.Database
}
// DeleteLiveRoutesOfPeer fails.
func (peerDeleteAlwaysFailsStore) DeleteLiveRoutesOfPeer(string) (int, error) {
return 0, errTestPeerDelete
}
// TestPrefixHandlerKeepsAtMostABatchOfUpdatesWhileADeleteKeepsFailing runs,
// through the prefix handler into a database whose every delete of a peer's
// routes fails, a peer's session going down and then three flushes. Before each
// flush the peer announces prefixBatchSize new prefixes and another peer
// announces one. After each flush the handler must hold only the peer's last
// prefixBatchSize updates, not every update since the down, and the other
// peer's route must be written. After the last flush, one more update must not
// wake the flush goroutine: held updates do not count toward a full batch.
func TestPrefixHandlerKeepsAtMostABatchOfUpdatesWhileADeleteKeepsFailing(
t *testing.T,
) {
t.Parallel()
const otherPeerIP = "2001:db8:ffff::2"
// One prefix the other peer announces before each flush.
otherPeersPrefixes := []string{testV4Prefix, testV6Prefix, "2001:db8:2::/48"}
db := newTestDatabase(t)
// Built without NewPrefixHandler's flush timer, so the test flushes itself.
h := &PrefixHandler{
db: peerDeleteAlwaysFailsStore{Database: db}, logger: logger.New(),
}
h.HandleMessage(sessionDown())
// The peer announces a /32 for each address from 10.0.0.0 on, each
// address once.
addr := netip.MustParseAddr("10.0.0.0")
for flush, otherPeersPrefix := range otherPeersPrefixes {
prefixes := make([]string, prefixBatchSize)
for i := range prefixes {
prefixes[i] = netip.PrefixFrom(addr, addr.BitLen()).String()
addr = addr.Next()
}
h.HandleMessage(peerAnnouncement(testPeerIP, prefixes...))
h.HandleMessage(peerAnnouncement(otherPeerIP, otherPeersPrefix))
h.flushBatch()
kept := make([]string, 0, len(h.held))
for _, update := range h.held {
kept = append(kept, update.prefix)
}
if !slices.Equal(kept, prefixes) {
t.Fatalf("after flush %d the handler holds %d updates, not just "+
"the peer's last %d", flush+1, len(kept), len(prefixes))
}
assertPeers(t, db, map[string][]string{otherPeersPrefix: {otherPeerIP}})
}
// A new channel, so that a wake-up in it comes from the update below.
h.batchFullCh = make(chan struct{}, 1)
h.HandleMessage(peerAnnouncement(otherPeerIP, testV4Prefix))
if len(h.batchFullCh) > 0 {
t.Error("one update after a flush that held a batch of updates " +
"woke the flush goroutine")
}
}
// TestPrefixHandlerStopLogsAFailedDeleteWithoutTheDeadline stops a prefix
// handler, with no deadline, after a peer's session went down and the peer
// announced a prefix, into a database whose every delete of a peer's routes
// fails. Stop must log the peer down and the announcement as not written, and
// must not say that the stop deadline passed.
func TestPrefixHandlerStopLogsAFailedDeleteWithoutTheDeadline(t *testing.T) {
t.Parallel()
var logs bytes.Buffer
h := NewPrefixHandler(peerDeleteAlwaysFailsStore{Database: newTestDatabase(t)},
newErrorLogger(&logs))
h.HandleMessage(sessionDown())
h.HandleMessage(peerAnnouncement(testPeerIP, testV4Prefix))
h.Stop(t.Context())
logged := lastLogged(t, &logs)
if strings.Contains(logged.Msg, "deadline") {
t.Errorf("Stop logged %q although its deadline had not passed",
logged.Msg)
}
if logged.Updates != 1 || logged.PeersDown != 1 {
t.Errorf("Stop logged %d updates and %d peer downs not written, "+
"want 1 and 1", logged.Updates, logged.PeersDown)
}
}
// TestPrefixHandlerStopLogsNothingAfterADeleteThatSucceeds stops a prefix
// handler, with no deadline, after a peer's session went down, into a database
// whose delete of the peer's routes succeeds. Stop must log nothing: a peer
// whose routes are deleted no longer counts as a peer down left undone.
func TestPrefixHandlerStopLogsNothingAfterADeleteThatSucceeds(t *testing.T) {
t.Parallel()
var logs bytes.Buffer
h := NewPrefixHandler(newTestDatabase(t), newErrorLogger(&logs))
h.HandleMessage(sessionDown())
h.Stop(t.Context())
if logs.Len() != 0 {
t.Errorf("Stop logged %q after the routes of the peer whose session "+
"went down were deleted", logs.String())
}
}
// newTestDatabase opens a database in a new temporary directory. It is closed
// when the test ends.
func newTestDatabase(t *testing.T) *database.Database {
internal/server/handlers.go
+54 -182
View File
@@ -31,18 +31,6 @@ const (
healthCheckTimeout = 2 * time.Second
)
// Limits of one IP lookup.
const (
// maxLookupTargets is the most addresses and host names an IP lookup may
// list, and the most addresses it may have once the names are resolved:
// each address costs up to 113 database reads and a PTR lookup.
maxLookupTargets = 10
// ptrLookupTimeout is each PTR lookup's own time limit, so that ten
// addresses whose PTR lookups never answer fit in the 30-second limit.
ptrLookupTimeout = 2 * time.Second
)
// Keys of the JSON envelope the handlers and middleware write, and the status
// value of an error answer.
const (
@@ -53,15 +41,6 @@ const (
jsonStatusError = "error"
)
// Error messages an answer sends, in place of the error behind it where there
// is one.
const (
invalidPrefixMessage = "Invalid prefix"
hostNotResolvedMessage = "could not resolve host name"
hostNotPublicMessage = "host name has no public address"
tooManyTargetsMessage = "Too many addresses or host names"
)
// IP versions as database.LiveRoute.IPVersion records them.
const (
ipVersionV4 = 4
@@ -89,9 +68,7 @@ func (s *Server) handleHealthCheck() http.HandlerFunc {
// Check database connectivity with lightweight ping
err := s.db.Ping(ctx)
if err != nil {
s.logger.Error("Health check database ping failed", "error", err)
checks["database"] = jsonStatusError
checks["database"] = "error: " + err.Error()
healthy = false
} else {
checks["database"] = "ok"
@@ -131,12 +108,12 @@ func (s *Server) handleHealthCheck() http.HandlerFunc {
// handleIndex returns a handler that serves the home page.
func (s *Server) handleIndex() http.HandlerFunc {
return func(w http.ResponseWriter, r *http.Request) {
return func(w http.ResponseWriter, _ *http.Request) {
w.Header().Set("Content-Type", "text/html; charset=utf-8")
tmpl := templates.IndexTemplate()
err := tmpl.Execute(w, pageData{CSPNonce: cspNonce(r.Context())})
err := tmpl.Execute(w, nil)
if err != nil {
s.logger.Error("Failed to render index template", "error", err)
http.Error(w, "Internal Server Error", http.StatusInternalServerError)
@@ -144,12 +121,6 @@ func (s *Server) handleIndex() http.HandlerFunc {
}
}
// pageData is what the pages that show no data are rendered with: the nonce
// their script and style blocks carry.
type pageData struct {
CSPNonce string
}
// writeJSONError writes a standardized JSON error response with the given
// status code and error message.
func (s *Server) writeJSONError(w http.ResponseWriter, statusCode int, message string) {
@@ -280,7 +251,7 @@ func (s *Server) handleStatusJSON() http.HandlerFunc {
if err != nil {
s.logger.Error("Failed to get database stats", "error", err)
s.writeJSONError(w, http.StatusInternalServerError, internalErrorMessage)
s.writeJSONError(w, http.StatusInternalServerError, err.Error())
return
}
@@ -570,7 +541,7 @@ func (s *Server) handleStats() http.HandlerFunc {
if err != nil {
s.logger.Error("Failed to get database stats", "error", err)
s.writeJSONError(w, http.StatusInternalServerError, internalErrorMessage)
s.writeJSONError(w, http.StatusInternalServerError, err.Error())
return
}
@@ -651,12 +622,12 @@ func (s *Server) statsResponse(
// handleStatusHTML returns a handler that serves the HTML status page,
// which displays real-time statistics fetched via JavaScript.
func (s *Server) handleStatusHTML() http.HandlerFunc {
return func(w http.ResponseWriter, r *http.Request) {
return func(w http.ResponseWriter, _ *http.Request) {
w.Header().Set("Content-Type", "text/html; charset=utf-8")
tmpl := templates.StatusTemplate()
err := tmpl.Execute(w, pageData{CSPNonce: cspNonce(r.Context())})
err := tmpl.Execute(w, nil)
if err != nil {
s.logger.Error("Failed to render template", "error", err)
http.Error(w, "Internal Server Error", http.StatusInternalServerError)
@@ -690,18 +661,15 @@ func (s *Server) handleIPInfo() http.HandlerFunc {
}
ctx := r.Context()
response := IPLookupResponse{Query: target, Results: make([]*database.IPInfo, 0)}
// Collect all IPs to look up
ipsToLookup, resolveErrors, err := s.resolveLookupTargets(ctx, target)
if deadlinePassed(ctx) {
s.writeJSONError(w, http.StatusGatewayTimeout, requestTimeoutMessage)
return
response := IPLookupResponse{
Query: target,
Results: make([]*database.IPInfo, 0),
}
if err != nil {
s.writeJSONError(w, http.StatusBadRequest, tooManyTargetsMessage)
// Collect all IPs to look up
ipsToLookup, resolveErrors := resolveLookupTargets(ctx, target)
if deadlinePassed(ctx) {
s.writeJSONError(w, http.StatusGatewayTimeout, requestTimeoutMessage)
return
}
@@ -709,7 +677,7 @@ func (s *Server) handleIPInfo() http.HandlerFunc {
response.Errors = append(response.Errors, resolveErrors...)
if len(ipsToLookup) == 0 {
s.writeJSONError(w, http.StatusBadRequest, noTargetsMessage(resolveErrors))
s.writeJSONError(w, http.StatusBadRequest, "No valid IPs or hostnames provided")
return
}
@@ -726,20 +694,14 @@ func (s *Server) handleIPInfo() http.HandlerFunc {
return
}
switch {
case errors.Is(err, database.ErrNoRoute), errors.Is(err, database.ErrInvalidIP):
if err != nil {
response.Errors = append(response.Errors, ip+": "+err.Error())
continue
case err != nil:
s.logger.Error("Failed to look up IP", "ip", ip, "error", err)
s.writeJSONError(w, http.StatusInternalServerError, internalErrorMessage)
return
}
// Do PTR lookup for this IP
ptrs := s.lookupPTRNames(ctx, ip)
ptrs := lookupPTRNames(ctx, ip)
if deadlinePassed(ctx) {
s.writeJSONError(w, http.StatusGatewayTimeout, requestTimeoutMessage)
@@ -760,37 +722,17 @@ func (s *Server) handleIPInfo() http.HandlerFunc {
// Queue WHOIS refresh for stale ASNs (non-blocking)
s.queueWHOISRefresh(refreshASNs)
s.writeIPLookupAnswer(w, response)
}
}
// Return response (even if no results, include errors)
if len(response.Results) == 0 && len(response.Errors) > 0 {
s.writeJSONError(w, http.StatusNotFound, "No routes found: "+response.Errors[0])
// noTargetsMessage returns the message of the 400 an IP lookup answers when
// no address is left to look up, followed by the first of resolveErrors, which
// says why, when there is one.
func noTargetsMessage(resolveErrors []string) string {
const message = "No valid IPs or hostnames provided"
return
}
if len(resolveErrors) == 0 {
return message
}
return message + ": " + resolveErrors[0]
}
// writeIPLookupAnswer sends response, or 404 with its first error when no
// address had a route.
func (s *Server) writeIPLookupAnswer(
w http.ResponseWriter, response IPLookupResponse,
) {
if len(response.Results) == 0 && len(response.Errors) > 0 {
s.writeJSONError(w, http.StatusNotFound, "No routes found: "+response.Errors[0])
return
}
err := writeJSONSuccess(w, response)
if err != nil {
s.logger.Error("Failed to encode IP lookup response", "error", err)
err := writeJSONSuccess(w, response)
if err != nil {
s.logger.Error("Failed to encode IP lookup response", "error", err)
}
}
}
@@ -822,16 +764,12 @@ func (s *Server) queueWHOISRefresh(asns map[int]bool) {
}
}
// errTooManyTargets is what resolveLookupTargets returns for a list of more
// than maxLookupTargets addresses and host names, or of names that resolve to
// more than maxLookupTargets addresses.
var errTooManyTargets = errors.New("too many addresses or host names")
// splitLookupTargets returns the addresses and host names in a comma-separated
// list, leaving out blank entries, or errTooManyTargets when there are more
// than maxLookupTargets.
func splitLookupTargets(target string) ([]string, error) {
var targets []string
// resolveLookupTargets splits a comma-separated list of IP addresses and
// hostnames and returns the IP addresses to look up, resolving each hostname.
// A hostname that does not resolve is reported in the returned errors instead.
// It stops once ctx's deadline has passed; the caller then answers 504.
func resolveLookupTargets(ctx context.Context, target string) ([]string, []string) {
var ipsToLookup, resolveErrors []string
for t := range strings.SplitSeq(target, ",") {
t = strings.TrimSpace(t)
@@ -839,36 +777,6 @@ func splitLookupTargets(target string) ([]string, error) {
continue
}
if len(targets) == maxLookupTargets {
return nil, errTooManyTargets
}
targets = append(targets, t)
}
return targets, nil
}
// resolveLookupTargets splits a comma-separated list of IP addresses and
// hostnames and returns the IP addresses to look up, resolving each hostname.
// A hostname that does not resolve is reported in the returned errors instead,
// with the resolver's own error only in the log, since it names the DNS server.
// Only the public addresses a hostname resolves to are kept, so the answer
// does not show the deployment's private addresses; a hostname with none is
// reported in the errors. More than maxLookupTargets entries, or addresses once
// resolved, return errTooManyTargets. It stops once ctx's deadline has passed;
// the caller then answers 504.
func (s *Server) resolveLookupTargets(
ctx context.Context, target string,
) ([]string, []string, error) {
targets, err := splitLookupTargets(target)
if err != nil {
return nil, nil, err
}
var ipsToLookup, resolveErrors []string
for _, t := range targets {
// Check if this target is an IP address
if net.ParseIP(t) != nil {
ipsToLookup = append(ipsToLookup, t)
@@ -877,57 +785,27 @@ func (s *Server) resolveLookupTargets(
}
// It's a hostname - resolve it
resolved, err := s.resolver.LookupHost(ctx, t)
resolved, err := net.DefaultResolver.LookupHost(ctx, t)
if deadlinePassed(ctx) {
return nil, nil, nil
return nil, nil
}
if err != nil {
s.logger.Warn("Host name lookup failed", "host", t, "error", err)
resolveErrors = append(resolveErrors, t+": "+hostNotResolvedMessage)
resolveErrors = append(resolveErrors, t+": "+err.Error())
continue
}
public := publicAddresses(resolved)
if len(public) == 0 {
resolveErrors = append(resolveErrors, t+": "+hostNotPublicMessage)
continue
}
ipsToLookup = append(ipsToLookup, public...)
ipsToLookup = append(ipsToLookup, resolved...)
}
if len(ipsToLookup) > maxLookupTargets {
return nil, nil, errTooManyTargets
}
return ipsToLookup, resolveErrors, nil
}
// publicAddresses returns the addresses in addrs that are global unicast and
// not private; Go counts the private ranges as global unicast.
func publicAddresses(addrs []string) []string {
var public []string
for _, a := range addrs {
ip, err := netip.ParseAddr(a)
if err == nil && ip.IsGlobalUnicast() && !ip.IsPrivate() {
public = append(public, a)
}
}
return public
return ipsToLookup, resolveErrors
}
// lookupPTRNames returns the PTR records of ip without their trailing dots, or
// nil when the lookup fails or runs past ptrLookupTimeout.
func (s *Server) lookupPTRNames(ctx context.Context, ip string) []string {
ctx, cancel := context.WithTimeout(ctx, ptrLookupTimeout)
defer cancel()
ptrs, err := s.resolver.LookupAddr(ctx, ip)
// nil when the lookup fails.
func lookupPTRNames(ctx context.Context, ip string) []string {
ptrs, err := net.DefaultResolver.LookupAddr(ctx, ip)
if err != nil {
return nil
}
@@ -941,7 +819,7 @@ func (s *Server) lookupPTRNames(ctx context.Context, ip string) []string {
}
// extractClientIP extracts the client IP from the request.
// Works with RealIPMiddleware, which sets RemoteAddr.
// Works with chi's RealIP middleware which sets RemoteAddr.
func extractClientIP(r *http.Request) string {
// RemoteAddr is in the form "IP:port" or just "IP" for unix sockets
host, _, err := net.SplitHostPort(r.RemoteAddr)
@@ -976,9 +854,7 @@ func (s *Server) handleASDetailJSON() http.HandlerFunc {
if errors.Is(err, database.ErrNoRoute) {
s.writeJSONError(w, http.StatusNotFound, err.Error())
} else {
s.logger.Error("Failed to get AS details", "asn", asn, "error", err)
s.writeJSONError(w, http.StatusInternalServerError,
internalErrorMessage)
s.writeJSONError(w, http.StatusInternalServerError, err.Error())
}
return
@@ -1057,7 +933,7 @@ func (s *Server) handlePrefixDetailJSON() http.HandlerFunc {
// tables store, so that any valid way of writing it finds the route.
parsed, err := netip.ParsePrefix(prefixParam + "/" + lenParam)
if err != nil {
s.writeJSONError(w, http.StatusBadRequest, invalidPrefixMessage)
s.writeJSONError(w, http.StatusBadRequest, err.Error())
return
}
@@ -1075,10 +951,7 @@ func (s *Server) handlePrefixDetailJSON() http.HandlerFunc {
if errors.Is(err, database.ErrNoRoute) {
s.writeJSONError(w, http.StatusNotFound, err.Error())
} else {
s.logger.Error("Failed to get prefix details",
"prefix", prefix, "error", err)
s.writeJSONError(w, http.StatusInternalServerError,
internalErrorMessage)
s.writeJSONError(w, http.StatusInternalServerError, err.Error())
}
return
@@ -1129,7 +1002,7 @@ func (s *Server) handleASDetail() http.HandlerFunc {
http.Error(w, "AS not found", http.StatusNotFound)
} else {
s.logger.Error("Failed to get AS details", "error", err)
http.Error(w, internalErrorMessage, http.StatusInternalServerError)
http.Error(w, "Internal server error", http.StatusInternalServerError)
}
return
@@ -1166,7 +1039,6 @@ func (s *Server) handleASDetail() http.HandlerFunc {
IPv6Count int
Peers []database.ASPeer
PeerCount int
CSPNonce string
}{
ASN: asInfo,
IPv4Prefixes: ipv4Prefixes,
@@ -1176,7 +1048,6 @@ func (s *Server) handleASDetail() http.HandlerFunc {
IPv6Count: len(ipv6Prefixes),
Peers: peers,
PeerCount: len(peers),
CSPNonce: cspNonce(r.Context()),
}
// Check if context is still valid before writing response
@@ -1187,7 +1058,9 @@ func (s *Server) handleASDetail() http.HandlerFunc {
w.Header().Set("Content-Type", "text/html; charset=utf-8")
err = templates.ASDetailTemplate().Execute(w, data)
tmpl := templates.ASDetailTemplate()
err = tmpl.Execute(w, data)
if err != nil {
s.logger.Error("Failed to render AS detail template", "error", err)
http.Error(w, "Internal Server Error", http.StatusInternalServerError)
@@ -1212,7 +1085,7 @@ func (s *Server) handlePrefixDetail() http.HandlerFunc {
// tables store, so that any valid way of writing it finds the route.
parsed, err := netip.ParsePrefix(prefixParam + "/" + lenParam)
if err != nil {
http.Error(w, invalidPrefixMessage, http.StatusBadRequest)
http.Error(w, "Invalid prefix", http.StatusBadRequest)
return
}
@@ -1231,7 +1104,7 @@ func (s *Server) handlePrefixDetail() http.HandlerFunc {
http.Error(w, "Prefix not found", http.StatusNotFound)
} else {
s.logger.Error("Failed to get prefix details", "error", err)
http.Error(w, internalErrorMessage, http.StatusInternalServerError)
http.Error(w, "Internal server error", http.StatusInternalServerError)
}
return
@@ -1258,7 +1131,6 @@ func (s *Server) handlePrefixDetail() http.HandlerFunc {
Origins []*ASNInfo
PeerCount int
OriginCount int
CSPNonce string
}{
Prefix: prefix,
MaskLength: maskLength,
@@ -1267,7 +1139,6 @@ func (s *Server) handlePrefixDetail() http.HandlerFunc {
Origins: origins,
PeerCount: len(routes),
OriginCount: len(origins),
CSPNonce: cspNonce(r.Context()),
}
// Check if context is still valid before writing response
@@ -1280,7 +1151,9 @@ func (s *Server) handlePrefixDetail() http.HandlerFunc {
w.Header().Set("Content-Type", "text/html; charset=utf-8")
err = templates.PrefixDetailTemplate().Execute(w, data)
tmpl := templates.PrefixDetailTemplate()
err = tmpl.Execute(w, data)
if err != nil {
s.logger.Error("Failed to render prefix detail template", "error", err)
http.Error(w, "Internal Server Error", http.StatusInternalServerError)
@@ -1430,7 +1303,7 @@ func (s *Server) prefixLengthHandler(ipVersion, maxMaskLength int) http.HandlerF
if err != nil {
s.logger.Error("Failed to get prefixes by length", "error", err)
http.Error(w, internalErrorMessage, http.StatusInternalServerError)
http.Error(w, "Internal server error", http.StatusInternalServerError)
return
}
@@ -1451,7 +1324,6 @@ func (s *Server) prefixLengthHandler(ipVersion, maxMaskLength int) http.HandlerF
"IPVersion": ipVersion,
"Prefixes": prefixesWithASDescriptions(prefixes),
"Count": len(prefixes),
"CSPNonce": cspNonce(r.Context()),
}
// Check if context is still valid before writing response
internal/server/handlers_internal_test.go
+17 -493
View File
@@ -9,9 +9,7 @@ import (
"net/http/httptest"
"runtime"
"slices"
"strconv"
"strings"
"sync/atomic"
"testing"
"time"
@@ -30,14 +28,6 @@ const (
testPeerV6 = "2001:db8::1"
)
// Paths several tests request.
const (
asPagePath = "/as/64500"
prefixLengthPagePath = "/prefixlength/24"
statusJSONPath = "/status.json"
statsPath = "/api/v1/stats"
)
// The names the stats handler tests report the two stats handlers by.
const (
statusJSONHandlerName = "status.json"
@@ -193,10 +183,14 @@ func TestStatsHandlersAnswerFromMemory(t *testing.T) {
}
}
// newServerWithTestRoutes returns a server on a new database that holds a
// route for testPrefixV4 and one for 2001:db8::/32.
func newServerWithTestRoutes(t *testing.T) *Server {
t.Helper()
// TestPrefixDetailHandlersParsePrefix checks that both prefix pages find a
// stored route whatever valid text form the URL writes its prefix in, and
// answer 400 for a prefix that does not parse. The tables store the form
// net/netip prints. Before the fix the URL text was looked up as typed, so
// "2001:DB8:0:0::/32" was not found although "2001:db8::/32" was stored, and
// "foo/24" answered 500 (https://git.eeqj.de/sneak/routewatch/issues/71).
func TestPrefixDetailHandlersParsePrefix(t *testing.T) {
t.Parallel()
db, err := database.New(&config.Config{StateDir: t.TempDir()}, logger.New())
if err != nil {
@@ -223,22 +217,7 @@ func newServerWithTestRoutes(t *testing.T) *Server {
t.Fatalf("UpsertLiveRouteBatch: %v", err)
}
return New(db, streamer.New(logger.New(), metrics.New()), logger.New(),
&config.Config{})
}
// TestPrefixDetailHandlersParsePrefix checks that both prefix pages find a
// stored route whatever valid text form the URL writes its prefix in, and
// answer 400 for a prefix that does not parse. The tables store the form
// net/netip prints. Before the fix the URL text was looked up as typed, so
// "2001:DB8:0:0::/32" was not found although "2001:db8::/32" was stored, and
// "foo/24" answered 500 (https://git.eeqj.de/sneak/routewatch/issues/71). A
// prefix with its colons percent-encoded, as JavaScript's encodeURIComponent
// writes it, answered 400 (https://git.eeqj.de/sneak/routewatch/issues/103).
func TestPrefixDetailHandlersParsePrefix(t *testing.T) {
t.Parallel()
s := newServerWithTestRoutes(t)
s := New(db, streamer.New(logger.New(), metrics.New()), logger.New(), &config.Config{})
tests := []struct {
path string
@@ -250,8 +229,6 @@ func TestPrefixDetailHandlersParsePrefix(t *testing.T) {
{"/prefix/2001:DB8:0:0::/32", http.StatusOK},
{"/api/v1/prefix/192.0.2.1/24", http.StatusOK},
{"/prefix/192.0.2.1/24", http.StatusOK},
{"/api/v1/prefix/2001%3Adb8%3A%3A/32", http.StatusOK},
{"/prefix/2001%3Adb8%3A%3A/32", http.StatusOK},
}
for _, tt := range tests {
@@ -266,61 +243,6 @@ func TestPrefixDetailHandlersParsePrefix(t *testing.T) {
}
}
// TestIPLookupsReadPercentEncodedAddresses checks that the IP lookups find the
// addresses a URL asks for when it percent-encodes them, as the home page's
// lookup form does with encodeURIComponent. Before the fix the handler read
// "2001%3Adb8%3A%3A1" undecoded as a host name, which did not resolve, and
// answered 400, so the home page could look up no IPv6 address
// (https://git.eeqj.de/sneak/routewatch/issues/103).
func TestIPLookupsReadPercentEncodedAddresses(t *testing.T) {
t.Parallel()
s := newServerWithTestRoutes(t)
tests := []struct {
path string
wantIPs string // the addresses of the results, comma-separated
}{
{"/ip/2001%3Adb8%3A%3A1", "2001:db8::1"},
{"/api/v1/ip/192.0.2.1%2C192.0.2.2", "192.0.2.1,192.0.2.2"},
}
for _, tt := range tests {
rec := httptest.NewRecorder()
s.router.ServeHTTP(rec, httptest.NewRequestWithContext(
t.Context(), http.MethodGet, tt.path, nil))
if rec.Code != http.StatusOK {
t.Errorf("%s: status %d, want %d; body %s",
tt.path, rec.Code, http.StatusOK, rec.Body)
continue
}
var body struct {
Data IPLookupResponse `json:"data"`
}
err := json.Unmarshal(rec.Body.Bytes(), &body)
if err != nil {
t.Errorf("%s: decoding the answer %q: %v", tt.path, rec.Body, err)
continue
}
gotIPs := make([]string, 0, len(body.Data.Results))
for _, result := range body.Data.Results {
gotIPs = append(gotIPs, result.IP)
}
got := strings.Join(gotIPs, ",")
if got != tt.wantIPs {
t.Errorf("%s: results for %q, want %q", tt.path, got, tt.wantIPs)
}
}
}
// errASDetailsLookupFailed is what failingASDetailsDB answers every AS details
// lookup with.
var errASDetailsLookupFailed = errors.New("AS details lookup failed")
@@ -439,7 +361,7 @@ func TestPagesSayOriginZeroIsNoKnownOrigin(t *testing.T) {
s := New(db, streamer.New(logger.New(), metrics.New()), logger.New(), &config.Config{})
for _, path := range []string{"/prefix/" + testPrefixV4, prefixLengthPagePath} {
for _, path := range []string{"/prefix/" + testPrefixV4, "/prefixlength/24"} {
rec := httptest.NewRecorder()
s.router.ServeHTTP(rec, httptest.NewRequestWithContext(
t.Context(), http.MethodGet, path, nil))
@@ -620,13 +542,13 @@ func TestRoutesAnswer504WhenDatabaseReadIsCutOff(t *testing.T) {
path string
wantJSONBody bool
}{
{path: asPagePath},
{path: "/as/64500"},
{path: "/prefix/192.0.2.0/24"},
{path: prefixLengthPagePath},
{path: "/prefixlength/24"},
{path: "/prefixlength6/48"},
{path: statusJSONPath, wantJSONBody: true},
{path: "/status.json", wantJSONBody: true},
{path: "/ip/192.0.2.1", wantJSONBody: true},
{path: statsPath, wantJSONBody: true},
{path: "/api/v1/stats", wantJSONBody: true},
{path: "/api/v1/ip/192.0.2.1", wantJSONBody: true},
{path: "/api/v1/as/64500", wantJSONBody: true},
{path: "/api/v1/prefix/192.0.2.0/24", wantJSONBody: true},
@@ -699,9 +621,9 @@ func TestRoutesAnswer504WhenSecondDatabaseReadIsCutOff(t *testing.T) {
path string
wantJSONBody bool
}{
{path: asPagePath},
{path: statusJSONPath, wantJSONBody: true},
{path: statsPath, wantJSONBody: true},
{path: "/as/64500"},
{path: "/status.json", wantJSONBody: true},
{path: "/api/v1/stats", wantJSONBody: true},
}
for _, tt := range tests {
@@ -768,404 +690,6 @@ func checkJSONErrorCode(
}
}
// errDiskIO is the database driver's error failingDB answers with.
var errDiskIO = errors.New("disk I/O error")
// failingDB embeds database.Store (left nil) and fails the reads of the JSON
// routes and the health check's ping with errDiskIO. Every other Store method
// is unused and would panic if called.
type failingDB struct {
database.Store
}
func (failingDB) Ping(_ context.Context) error {
return errDiskIO
}
func (failingDB) GetStatsContext(_ context.Context) (database.Stats, error) {
return database.Stats{}, errDiskIO
}
func (failingDB) GetIPInfoContext(
_ context.Context, _ string,
) (*database.IPInfo, error) {
return nil, errDiskIO
}
func (failingDB) GetASDetailsContext(
_ context.Context, _ int,
) (*database.ASN, []database.LiveRoute, error) {
return nil, nil, errDiskIO
}
func (failingDB) GetPrefixDetailsContext(
_ context.Context, _ string,
) ([]database.LiveRoute, map[int]database.ASN, error) {
return nil, nil, errDiskIO
}
// TestJSONRoutesDoNotSendErrorDetails sends a request to each JSON route and
// the health check with a store whose reads fail with the driver's error text,
// and a prefix that does not parse, and checks that no answer carries the
// driver's or the prefix parser's text. Before the fix the answers sent it to
// the client (https://git.eeqj.de/sneak/routewatch/issues/108).
func TestJSONRoutesDoNotSendErrorDetails(t *testing.T) {
t.Parallel()
s := New(failingDB{}, streamer.New(logger.New(), metrics.New()), logger.New(),
&config.Config{})
tests := []struct {
path string
wantCode int
}{
{"/.well-known/healthcheck.json", http.StatusServiceUnavailable},
{statusJSONPath, http.StatusInternalServerError},
{statsPath, http.StatusInternalServerError},
{"/api/v1/as/64500", http.StatusInternalServerError},
{"/api/v1/prefix/192.0.2.0/24", http.StatusInternalServerError},
{"/api/v1/ip/192.0.2.1", http.StatusInternalServerError},
{"/api/v1/prefix/foo/24", http.StatusBadRequest},
}
for _, tt := range tests {
rec := httptest.NewRecorder()
s.router.ServeHTTP(rec, httptest.NewRequestWithContext(
t.Context(), http.MethodGet, tt.path, nil))
if rec.Code != tt.wantCode {
t.Errorf("%s: status %d, want %d; body %s",
tt.path, rec.Code, tt.wantCode, rec.Body)
}
for _, detail := range []string{errDiskIO.Error(), "netip"} {
if strings.Contains(rec.Body.String(), detail) {
t.Errorf("%s: body %s carries the error text %q",
tt.path, rec.Body, detail)
}
}
}
}
// The addresses failingIPLookupDB fails the read of and finds a route for.
const (
failingLookupIP = "192.0.2.1"
routedLookupIP = "192.0.2.2"
)
// failingIPLookupDB embeds database.Store (left nil) and answers the IP lookup
// with errDiskIO for failingLookupIP, a route for routedLookupIP, and no route
// for any other address. Every other Store method is unused and would panic if
// called.
type failingIPLookupDB struct {
database.Store
}
func (failingIPLookupDB) GetIPInfoContext(
_ context.Context, ip string,
) (*database.IPInfo, error) {
switch ip {
case failingLookupIP:
return nil, errDiskIO
case routedLookupIP:
return &database.IPInfo{IP: ip}, nil
default:
return nil, database.ErrNoRoute
}
}
// TestIPLookupAnswers500WhenDatabaseReadFails sends IP lookups through the
// server's router with a store whose read of one address fails with the
// driver's error. A lookup that asks for that address answers 500 with the
// JSON error body, as the AS and prefix JSON routes do; an address with no
// route still answers 404. Before the fix the failed read was reported as if
// the address had no route: 404 for that address alone, and 200 beside a
// routed address (https://git.eeqj.de/sneak/routewatch/issues/109).
func TestIPLookupAnswers500WhenDatabaseReadFails(t *testing.T) {
t.Parallel()
s := New(failingIPLookupDB{}, streamer.New(logger.New(), metrics.New()),
logger.New(), &config.Config{})
tests := []struct {
path string
wantCode int
}{
{
path: "/api/v1/ip/" + failingLookupIP,
wantCode: http.StatusInternalServerError,
},
{
path: "/api/v1/ip/" + failingLookupIP + "," + routedLookupIP,
wantCode: http.StatusInternalServerError,
},
{
path: "/api/v1/ip/203.0.113.1",
wantCode: http.StatusNotFound,
},
}
for _, tt := range tests {
rec := httptest.NewRecorder()
s.router.ServeHTTP(rec, httptest.NewRequestWithContext(
t.Context(), http.MethodGet, tt.path, nil))
if rec.Code != tt.wantCode {
t.Errorf("%s: status %d, want %d; body %s",
tt.path, rec.Code, tt.wantCode, rec.Body)
}
checkJSONErrorCode(t, tt.path, rec, tt.wantCode)
}
}
// errNoSuchHost is what fakeResolver answers a lookup it has no answer for
// with.
var errNoSuchHost = errors.New("no such host")
// fakeResolver answers a host name lookup with the addresses hosts lists for
// the name, and a PTR lookup with no names or, when ptrNeverAnswers is set,
// not at all: it waits until the lookup's context ends. hostLookups counts the
// host name lookups.
type fakeResolver struct {
hosts map[string][]string
ptrNeverAnswers bool
hostLookups atomic.Int64
}
func (r *fakeResolver) LookupHost(_ context.Context, host string) ([]string, error) {
r.hostLookups.Add(1)
addrs, ok := r.hosts[host]
if !ok {
return nil, errNoSuchHost
}
return addrs, nil
}
func (r *fakeResolver) LookupAddr(ctx context.Context, _ string) ([]string, error) {
if r.ptrNeverAnswers {
<-ctx.Done()
return nil, ctx.Err()
}
return nil, errNoSuchHost
}
// countingIPLookupDB embeds database.Store (left nil) and counts its IP
// lookups, answering each that the address has no route. Every other Store
// method is unused and would panic if called.
type countingIPLookupDB struct {
database.Store
lookups atomic.Int64
}
func (d *countingIPLookupDB) GetIPInfoContext(
_ context.Context, _ string,
) (*database.IPInfo, error) {
d.lookups.Add(1)
return nil, database.ErrNoRoute
}
// TestIPLookupRefusesTooManyAddresses sends IP lookups of maxLookupTargets
// addresses and of one more, listed or as a host name that resolves to them,
// and a lookup listing maxLookupTargets+1 host names. The longer lookups answer
// 400 without reading the database, and the list of host names without
// resolving any of them; the others read the database once per address. Before
// the fix every address was read and every name resolved, however many there
// were (https://git.eeqj.de/sneak/routewatch/issues/110).
func TestIPLookupRefusesTooManyAddresses(t *testing.T) {
t.Parallel()
addrs := make([]string, maxLookupTargets+1)
names := make([]string, maxLookupTargets+1)
for i := range addrs {
addrs[i] = "192.0.2." + strconv.Itoa(i+1)
names[i] = "host" + strconv.Itoa(i+1) + ".test"
}
hosts := map[string][]string{
"most.test": addrs[:maxLookupTargets],
"too-many.test": addrs,
}
tests := []struct {
target string
wantCode int
wantLookups int64
wantHostLookups int64
}{
{
strings.Join(addrs[:maxLookupTargets], ","),
http.StatusNotFound, maxLookupTargets, 0,
},
{strings.Join(addrs, ","), http.StatusBadRequest, 0, 0},
{"most.test", http.StatusNotFound, maxLookupTargets, 1},
{"too-many.test", http.StatusBadRequest, 0, 1},
{strings.Join(names, ","), http.StatusBadRequest, 0, 0},
}
for _, tt := range tests {
db := &countingIPLookupDB{}
s := New(db, streamer.New(logger.New(), metrics.New()), logger.New(),
&config.Config{})
resolver := &fakeResolver{hosts: hosts}
s.resolver = resolver
rec := httptest.NewRecorder()
s.router.ServeHTTP(rec, httptest.NewRequestWithContext(
t.Context(), http.MethodGet, "/api/v1/ip/"+tt.target, nil))
if rec.Code != tt.wantCode {
t.Errorf("%s: status %d, want %d; body %s",
tt.target, rec.Code, tt.wantCode, rec.Body)
}
checkJSONErrorCode(t, tt.target, rec, tt.wantCode)
got := db.lookups.Load()
if got != tt.wantLookups {
t.Errorf("%s: %d database reads, want %d",
tt.target, got, tt.wantLookups)
}
gotHostLookups := resolver.hostLookups.Load()
if gotHostLookups != tt.wantHostLookups {
t.Errorf("%s: %d host name lookups, want %d",
tt.target, gotHostLookups, tt.wantHostLookups)
}
}
}
// addrInTestPrefixV4 is an address in testPrefixV4, which
// newServerWithTestRoutes stores a route for.
const addrInTestPrefixV4 = "192.0.2.1"
// TestIPLookupKeepsRouteWhenPTRLookupDoesNotAnswer looks up a routed address
// with a resolver that never answers its PTR lookup. Once the PTR lookup's own
// limit passes, the lookup answers 200 with the route. Before the fix the PTR
// lookup ran until the request's time limit, and the lookup answered 504
// without the route (https://git.eeqj.de/sneak/routewatch/issues/110).
func TestIPLookupKeepsRouteWhenPTRLookupDoesNotAnswer(t *testing.T) {
t.Parallel()
s := newServerWithTestRoutes(t)
s.resolver = &fakeResolver{ptrNeverAnswers: true}
// The request's time limit passes after the PTR lookup's.
ctx, cancel := context.WithTimeout(t.Context(), 2*ptrLookupTimeout)
defer cancel()
rec := httptest.NewRecorder()
s.router.ServeHTTP(rec, httptest.NewRequestWithContext(
ctx, http.MethodGet, "/api/v1/ip/"+addrInTestPrefixV4, nil))
if rec.Code != http.StatusOK {
t.Fatalf("status %d, want %d; body %s", rec.Code, http.StatusOK, rec.Body)
}
var body struct {
Data IPLookupResponse `json:"data"`
}
err := json.Unmarshal(rec.Body.Bytes(), &body)
if err != nil {
t.Fatalf("decoding the answer %q: %v", rec.Body, err)
}
if len(body.Data.Results) != 1 || body.Data.Results[0].Netblock != testPrefixV4 {
t.Errorf("answer %s, want one result in %s", rec.Body, testPrefixV4)
}
}
// The host name TestIPLookupHidesPrivateAddresses resolves to a private
// address, and that address.
const (
internalHostName = "internal.test"
internalHostAddr = "10.0.0.5"
)
// TestIPLookupHidesPrivateAddresses looks up host names that resolve to a
// private address, alone or beside a routed address. No answer carries the
// private address, and a name with no other address is reported as having no
// public address, in the 400 when nothing else is left to look up. Before the
// fix the private address was looked up and named in the answer, "No routes
// found: 10.0.0.5: no route found for IP 10.0.0.5", so the API showed what the
// deployment's own DNS knows (https://git.eeqj.de/sneak/routewatch/issues/110).
func TestIPLookupHidesPrivateAddresses(t *testing.T) {
t.Parallel()
s := newServerWithTestRoutes(t)
s.resolver = &fakeResolver{hosts: map[string][]string{
internalHostName: {internalHostAddr},
"mixed.test": {internalHostAddr, addrInTestPrefixV4},
}}
notPublic := internalHostName + ": " + hostNotPublicMessage
tests := []struct {
target string
wantCode int
wantMessage string
wantErrors []string
}{
{
internalHostName, http.StatusBadRequest,
"No valid IPs or hostnames provided: " + notPublic, nil,
},
{
internalHostName + "," + addrInTestPrefixV4, http.StatusOK, "",
[]string{notPublic},
},
{"mixed.test", http.StatusOK, "", nil},
}
for _, tt := range tests {
rec := httptest.NewRecorder()
s.router.ServeHTTP(rec, httptest.NewRequestWithContext(
t.Context(), http.MethodGet, "/api/v1/ip/"+tt.target, nil))
if rec.Code != tt.wantCode {
t.Errorf("%s: status %d, want %d; body %s",
tt.target, rec.Code, tt.wantCode, rec.Body)
}
if strings.Contains(rec.Body.String(), internalHostAddr) {
t.Errorf("%s: body %s carries %s", tt.target, rec.Body, internalHostAddr)
}
var body struct {
Data IPLookupResponse `json:"data"`
Error struct {
Msg string `json:"msg"`
} `json:"error"`
}
err := json.Unmarshal(rec.Body.Bytes(), &body)
if err != nil {
t.Errorf("%s: decoding the answer %q: %v", tt.target, rec.Body, err)
continue
}
if body.Error.Msg != tt.wantMessage {
t.Errorf("%s: error message %q, want %q",
tt.target, body.Error.Msg, tt.wantMessage)
}
if !slices.Equal(body.Data.Errors, tt.wantErrors) {
t.Errorf("%s: errors %q, want %q",
tt.target, body.Data.Errors, tt.wantErrors)
}
}
}
// settledGoroutineCount lets transient goroutines finish, then reports the
// current count.
func settledGoroutineCount() int {
internal/server/middleware.go
+4 -135
View File
@@ -3,19 +3,13 @@ package server
import (
"bytes"
"context"
"crypto/rand"
"encoding/json"
"errors"
"fmt"
"log/slog"
"net/http"
"net/netip"
"slices"
"strings"
"sync"
"time"
"github.com/go-chi/chi/v5"
)
// responseWriter wraps http.ResponseWriter to capture the response
@@ -122,10 +116,6 @@ func JSONResponseMiddleware(next http.Handler) http.Handler {
// requestTimeoutMessage is the error message sent with 504 Gateway Timeout.
const requestTimeoutMessage = "Request timeout"
// internalErrorMessage is the error message sent with 500 Internal Server
// Error, whatever went wrong; the handler logs the detail.
const internalErrorMessage = "Internal server error"
// deadlinePassed reports whether ctx ended because its deadline passed.
//
// A handler calls it right after each database read or host name lookup,
@@ -142,9 +132,8 @@ func deadlinePassed(ctx context.Context) bool {
//
// If a response is not valid JSON or is empty when JSON is expected, the middleware
// returns a properly formatted JSON error response. If the request's time limit
// has passed, it answers 504 with the error message "Request timeout". A 4xx
// keeps its status, with the status's standard text as the message. For other
// errors, it returns "Internal server error", with status 500 if the original
// has passed, it answers 504 with the error message "Request timeout". For other
// errors, it returns "Internal server error" with status 500 if the original
// status was 200.
//
// Non-JSON responses are passed through unchanged.
@@ -203,14 +192,9 @@ func JSONValidationMiddleware(next http.Handler) http.Handler {
w.WriteHeader(statusCode)
errorMsg := internalErrorMessage
switch {
case statusCode == http.StatusGatewayTimeout:
errorMsg := "Internal server error"
if statusCode == http.StatusGatewayTimeout {
errorMsg = requestTimeoutMessage
case statusCode >= http.StatusBadRequest &&
statusCode < http.StatusInternalServerError:
errorMsg = http.StatusText(statusCode)
}
response := map[string]any{
@@ -297,118 +281,3 @@ func RequestLoggerMiddleware(logger *slog.Logger) func(http.Handler) http.Handle
})
}
}
// DecodedPathMiddleware makes the router match routes against the decoded
// path, so that the handlers read their URL parameters decoded. Without it,
// chi matches against the escaped path whenever the request has one and hands
// the handlers their parameters still escaped: /ip/2001%3Adb8%3A%3A1 would
// look up "2001%3Adb8%3A%3A1".
func DecodedPathMiddleware(next http.Handler) http.Handler {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
chi.RouteContext(r.Context()).RoutePath = r.URL.Path
next.ServeHTTP(w, r)
})
}
// The values of the headers SecurityHeadersMiddleware sets.
const (
// strictTransportSecurity tells browsers to reach this host and its
// subdomains only over HTTPS for a year. The server listens on plain
// HTTP behind a TLS-terminating proxy, which passes the header on.
strictTransportSecurity = "max-age=31536000; includeSubDomains"
// contentSecurityPolicy lets a page load content only from this server,
// and run or apply an inline script or style block only when the block
// carries the nonce made for that answer, which replaces %[1]s. No other
// site may frame the pages.
contentSecurityPolicy = "default-src 'self'; " +
"script-src 'nonce-%[1]s'; style-src 'nonce-%[1]s'; " +
"frame-ancestors 'none'; base-uri 'none'; form-action 'self'"
frameOptions = "DENY"
contentTypeOptions = "nosniff"
referrerPolicy = "strict-origin-when-cross-origin"
// permissionsPolicy turns off browser features the pages do not use.
permissionsPolicy = "accelerometer=(), camera=(), geolocation=(), " +
"gyroscope=(), magnetometer=(), microphone=(), payment=(), usb=()"
)
// cspNonceKey is the request context key of the nonce SecurityHeadersMiddleware
// made for the answer.
type cspNonceKey struct{}
// SecurityHeadersMiddleware sets the security headers REPO_POLICIES.md requires
// on every answer. Each answer's Content-Security-Policy names a nonce of its
// own; a page reads it with cspNonce and writes it on its script and style
// blocks.
func SecurityHeadersMiddleware(next http.Handler) http.Handler {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
nonce := rand.Text()
header := w.Header()
header.Set("Strict-Transport-Security", strictTransportSecurity)
header.Set("Content-Security-Policy",
fmt.Sprintf(contentSecurityPolicy, nonce))
header.Set("X-Frame-Options", frameOptions)
header.Set("X-Content-Type-Options", contentTypeOptions)
header.Set("Referrer-Policy", referrerPolicy)
header.Set("Permissions-Policy", permissionsPolicy)
ctx := context.WithValue(r.Context(), cspNonceKey{}, nonce)
next.ServeHTTP(w, r.WithContext(ctx))
})
}
// cspNonce returns the nonce SecurityHeadersMiddleware made for the answer to
// the request with ctx.
func cspNonce(ctx context.Context) string {
nonce, _ := ctx.Value(cspNonceKey{}).(string)
return nonce
}
// RealIPMiddleware gives a request from one of trustedProxies the address of
// the client the proxies forwarded it for as its RemoteAddr: the last
// X-Forwarded-For entry that is not a trusted proxy. Each proxy appends the
// address it took the request from, so the entries left of that one are
// whatever the client sent. A request from any other peer keeps the peer's
// address, whatever headers it carries.
func RealIPMiddleware(trustedProxies []netip.Prefix) func(http.Handler) http.Handler {
isTrusted := func(addr netip.Addr) bool {
return slices.ContainsFunc(trustedProxies, func(proxy netip.Prefix) bool {
return proxy.Contains(addr.Unmap())
})
}
return func(next http.Handler) http.Handler {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
peer, err := netip.ParseAddrPort(r.RemoteAddr)
if err != nil || !isTrusted(peer.Addr()) {
next.ServeHTTP(w, r)
return
}
// A header sent on several lines is one list.
entries := strings.Split(
strings.Join(r.Header.Values("X-Forwarded-For"), ","), ",")
for _, entry := range slices.Backward(entries) {
addr, err := netip.ParseAddr(strings.TrimSpace(entry))
if err != nil {
break
}
r.RemoteAddr = addr.String()
if !isTrusted(addr) {
break
}
}
next.ServeHTTP(w, r)
})
}
}
internal/server/middleware_internal_test.go
-233
View File
@@ -1,233 +0,0 @@
package server
import (
"encoding/json"
"net/http"
"net/http/httptest"
"net/netip"
"strings"
"testing"
"time"
)
// The addresses the forwarding tests use: the client a proxy names in
// X-Forwarded-For, a peer inside the trusted range of
// TestRealIPMiddlewareTrustsOnlyConfiguredProxies, and a peer outside it,
// which lies in testPrefixV4 so that an IP lookup of it finds a route.
const (
forwardedClient = "203.0.113.9"
trustedPeer = "10.0.0.2:1234"
untrustedPeerAddr = "192.0.2.1"
untrustedPeer = untrustedPeerAddr + ":1234"
)
// TestResponsesCarrySecurityHeaders checks that a page, a JSON route and a
// path no route matches each answer with the security headers REPO_POLICIES.md
// requires, and that the Content-Security-Policy allows no inline content
// without a nonce (https://git.eeqj.de/sneak/routewatch/issues/108).
func TestResponsesCarrySecurityHeaders(t *testing.T) {
t.Parallel()
s := newServerWithTestRoutes(t)
headers := []string{
"Strict-Transport-Security",
"Content-Security-Policy",
"X-Frame-Options",
"X-Content-Type-Options",
"Referrer-Policy",
"Permissions-Policy",
}
for _, path := range []string{"/", statsPath, "/no-such-page"} {
rec := httptest.NewRecorder()
s.router.ServeHTTP(rec, httptest.NewRequestWithContext(
t.Context(), http.MethodGet, path, nil))
for _, name := range headers {
if rec.Header().Get(name) == "" {
t.Errorf("%s: no %s header", path, name)
}
}
policy := rec.Header().Get("Content-Security-Policy")
if strings.Contains(policy, "'unsafe-") {
t.Errorf("%s: Content-Security-Policy %q allows unsafe content",
path, policy)
}
}
}
// TestJSONValidationKeeps4xxWithoutBody wraps a handler that answers 400
// without a body in JSONValidationMiddleware and checks that the JSON error
// body keeps the 400 and gives the status's own text. Before the fix it said
// "Internal server error"
// (https://git.eeqj.de/sneak/routewatch/issues/111).
func TestJSONValidationKeeps4xxWithoutBody(t *testing.T) {
t.Parallel()
handler := JSONValidationMiddleware(http.HandlerFunc(
func(w http.ResponseWriter, _ *http.Request) {
w.WriteHeader(http.StatusBadRequest)
}))
rec := httptest.NewRecorder()
handler.ServeHTTP(rec, httptest.NewRequestWithContext(
t.Context(), http.MethodGet, "/", nil))
const name = "400 without a body"
if rec.Code != http.StatusBadRequest {
t.Errorf("%s: status %d, want %d; body %s",
name, rec.Code, http.StatusBadRequest, rec.Body)
}
checkJSONErrorCode(t, name, rec, http.StatusBadRequest)
checkJSONErrorMessage(t, name, rec, http.StatusText(http.StatusBadRequest))
}
// TestPagesCarryTheCSPNonce checks that every script and style block of each
// page carries the nonce its Content-Security-Policy header names, so that the
// browser runs and applies them, that each answer has a nonce of its own, and
// that no page uses a style attribute, which the policy does not allow.
func TestPagesCarryTheCSPNonce(t *testing.T) {
t.Parallel()
s := newServerWithTestRoutes(t)
_, err := s.db.GetOrCreateASN(64500, time.Date(2026, 1, 2, 3, 4, 5, 0, time.UTC))
if err != nil {
t.Fatalf("GetOrCreateASN: %v", err)
}
paths := []string{
"/", "/status", asPagePath, "/prefix/" + testPrefixV4,
prefixLengthPagePath, "/prefixlength6/32",
}
seen := make(map[string]bool)
for _, path := range paths {
rec := httptest.NewRecorder()
s.router.ServeHTTP(rec, httptest.NewRequestWithContext(
t.Context(), http.MethodGet, path, nil))
if rec.Code != http.StatusOK {
t.Errorf("%s: status %d, want %d; body %s",
path, rec.Code, http.StatusOK, rec.Body)
continue
}
policy := rec.Header().Get("Content-Security-Policy")
_, nonce, found := strings.Cut(policy, "'nonce-")
if !found {
t.Errorf("%s: Content-Security-Policy %q names no nonce", path, policy)
continue
}
nonce, _, _ = strings.Cut(nonce, "'")
if seen[nonce] {
t.Errorf("%s: nonce %q was already used for another answer", path, nonce)
}
seen[nonce] = true
page := rec.Body.String()
for _, tag := range []string{"<script", "<style"} {
withNonce := tag + ` nonce="` + nonce + `"`
if strings.Count(page, tag) != strings.Count(page, withNonce) {
t.Errorf("%s: a %s tag does not carry the nonce %q", path, tag, nonce)
}
}
if strings.Contains(page, "style=") {
t.Errorf("%s: the page uses a style attribute", path)
}
}
}
// TestForwardedForFromUntrustedPeerIsIgnored asks /ip/ for the caller's own
// address from a peer that is not a trusted proxy, with forwarding headers
// naming another address, and checks that the lookup is for the TCP peer.
// Before the fix any client could choose the address it was looked up and
// logged as (https://git.eeqj.de/sneak/routewatch/issues/108).
func TestForwardedForFromUntrustedPeerIsIgnored(t *testing.T) {
t.Parallel()
s := newServerWithTestRoutes(t)
req := httptest.NewRequestWithContext(t.Context(), http.MethodGet, "/ip/", nil)
req.RemoteAddr = untrustedPeer
req.Header.Set("X-Forwarded-For", forwardedClient)
req.Header.Set("X-Real-IP", forwardedClient)
req.Header.Set("True-Client-IP", forwardedClient)
rec := httptest.NewRecorder()
s.router.ServeHTTP(rec, req)
var body struct {
Data IPLookupResponse `json:"data"`
}
err := json.Unmarshal(rec.Body.Bytes(), &body)
if err != nil {
t.Fatalf("decoding the answer %q: %v", rec.Body, err)
}
if body.Data.Query != untrustedPeerAddr {
t.Errorf("looked up %q, want the TCP peer %s; answer %s",
body.Data.Query, untrustedPeerAddr, rec.Body)
}
}
// TestRealIPMiddlewareTrustsOnlyConfiguredProxies checks which address
// RealIPMiddleware gives a request: the TCP peer, unless the peer is a trusted
// proxy; then the last X-Forwarded-For entry that is not a trusted proxy,
// since the entries left of it are whatever the client sent.
func TestRealIPMiddlewareTrustsOnlyConfiguredProxies(t *testing.T) {
t.Parallel()
trusted := []netip.Prefix{netip.MustParsePrefix("10.0.0.0/8")}
tests := []struct {
name string
peer string
forwardedFor string
want string
}{
{"untrusted peer", untrustedPeer, forwardedClient, untrustedPeer},
{"trusted peer", trustedPeer, forwardedClient, forwardedClient},
{
"entries the client wrote", trustedPeer,
"198.51.100.7, " + forwardedClient, forwardedClient,
},
{
"two trusted proxies", trustedPeer,
forwardedClient + ", 10.0.0.3", forwardedClient,
},
{"no header", trustedPeer, "", trustedPeer},
{"entry not an address", trustedPeer, "client", trustedPeer},
}
for _, tt := range tests {
var got string
handler := RealIPMiddleware(trusted)(http.HandlerFunc(
func(_ http.ResponseWriter, r *http.Request) { got = r.RemoteAddr }))
req := httptest.NewRequestWithContext(t.Context(), http.MethodGet, "/", nil)
req.RemoteAddr = tt.peer
if tt.forwardedFor != "" {
req.Header.Set("X-Forwarded-For", tt.forwardedFor)
}
handler.ServeHTTP(httptest.NewRecorder(), req)
if got != tt.want {
t.Errorf("%s: RemoteAddr %q, want %q", tt.name, got, tt.want)
}
}
}
internal/server/routes.go
+2 -37
View File
@@ -1,36 +1,19 @@
package server
import (
"net/http"
"time"
"github.com/go-chi/chi/v5"
"github.com/go-chi/chi/v5/middleware"
)
// The error messages the JSON routes under /api/v1 and /ip answer with when no
// route matches the path, and when none takes the method.
const (
notFoundMessage = "Not found"
methodNotAllowedMessage = "Method not allowed"
// jsonRoutesAllow is the Allow header of their 405 answer: every one of
// them takes GET and HEAD, and no other method.
jsonRoutesAllow = "GET, HEAD"
)
// setupRoutes configures the HTTP routes
func (s *Server) setupRoutes() {
r := chi.NewRouter()
// Middleware
r.Use(DecodedPathMiddleware)
// GetHead serves HEAD with the GET handler. It looks the route up by the
// path DecodedPathMiddleware sets, so it comes after it.
r.Use(middleware.GetHead)
r.Use(SecurityHeadersMiddleware)
r.Use(middleware.RequestID)
r.Use(RealIPMiddleware(s.trustedProxies))
r.Use(middleware.RealIP)
r.Use(RequestLoggerMiddleware(s.logger.Logger)) // Structured request logging
r.Use(middleware.Recoverer)
r.Use(JSONResponseMiddleware)
@@ -55,34 +38,16 @@ func (s *Server) setupRoutes() {
r.Get("/prefixlength/{length}", s.handlePrefixLength())
r.Get("/prefixlength6/{length}", s.handlePrefixLength6())
// The JSON routes answer a path or method they do not take with the JSON
// error body, as they answer everything else.
jsonNotFound := func(w http.ResponseWriter, _ *http.Request) {
s.writeJSONError(w, http.StatusNotFound, notFoundMessage)
}
jsonMethodNotAllowed := func(w http.ResponseWriter, _ *http.Request) {
// chi's own 405 answer sets Allow; one set in its place must too.
w.Header().Set("Allow", jsonRoutesAllow)
s.writeJSONError(w, http.StatusMethodNotAllowed, methodNotAllowedMessage)
}
// IP info JSON endpoints (replaces old /ip redirect)
r.Route("/ip", func(r chi.Router) {
r.Use(JSONValidationMiddleware)
r.NotFound(jsonNotFound)
r.MethodNotAllowed(jsonMethodNotAllowed)
r.Get("/", s.handleIPInfo()) // Client IP
// GetHead does not reach this route: mounting the group registers
// /ip and /ip/ for every method, so HEAD arrives here as HEAD.
r.Head("/", s.handleIPInfo())
r.Get("/", s.handleIPInfo()) // Client IP
r.Get("/{addr}", s.handleIPInfo()) // Specified IP
})
// API routes
r.Route("/api/v1", func(r chi.Router) {
r.Use(JSONValidationMiddleware)
r.NotFound(jsonNotFound)
r.MethodNotAllowed(jsonMethodNotAllowed)
r.Get("/stats", s.handleStats())
r.Get("/ip/{ip}", s.handleIPLookup())
r.Get("/as/{asn}", s.handleASDetailJSON())
internal/server/routes_internal_test.go
-100
View File
@@ -1,100 +0,0 @@
package server
import (
"encoding/json"
"net/http"
"net/http/httptest"
"testing"
)
// TestRoutesAnswerHEAD sends HEAD for a page and JSON routes through the
// server's router and checks that each answers as GET does. /ip and /ip/, the
// lookup of the caller's own address, are in the list because chi's GetHead
// does not reach them. Before the fix every route answered HEAD with 405
// (https://git.eeqj.de/sneak/routewatch/issues/111).
func TestRoutesAnswerHEAD(t *testing.T) {
t.Parallel()
s := newServerWithTestRoutes(t)
for _, path := range []string{"/", statsPath, "/ip", "/ip/"} {
rec := httptest.NewRecorder()
s.router.ServeHTTP(rec, httptest.NewRequestWithContext(
t.Context(), http.MethodHead, path, nil))
if rec.Code != http.StatusOK {
t.Errorf("HEAD %s: status %d, want %d; body %s",
path, rec.Code, http.StatusOK, rec.Body)
}
}
}
// TestJSONRoutesAnswerUnmatchedRequestsWithJSONErrors sends a method the JSON
// routes do not take, and a path none of them matches, to each group of JSON
// routes through the server's router. Each must answer with the JSON error
// body, and a 405 must name the methods the routes take in its Allow header.
// Before the fix a 405 said "Internal server error" and a 404 was chi's plain
// text answer (https://git.eeqj.de/sneak/routewatch/issues/111).
func TestJSONRoutesAnswerUnmatchedRequestsWithJSONErrors(t *testing.T) {
t.Parallel()
s := newServerWithTestRoutes(t)
tests := []struct {
method string
path string
wantCode int
wantMsg string
}{
{http.MethodPost, statsPath, http.StatusMethodNotAllowed, methodNotAllowedMessage},
{http.MethodPost, "/ip/", http.StatusMethodNotAllowed, methodNotAllowedMessage},
{http.MethodGet, "/api/v1/nothing", http.StatusNotFound, notFoundMessage},
{http.MethodGet, "/ip/192.0.2.1/nothing", http.StatusNotFound, notFoundMessage},
}
for _, tt := range tests {
name := tt.method + " " + tt.path
rec := httptest.NewRecorder()
s.router.ServeHTTP(rec, httptest.NewRequestWithContext(
t.Context(), tt.method, tt.path, nil))
if rec.Code != tt.wantCode {
t.Errorf("%s: status %d, want %d; body %s",
name, rec.Code, tt.wantCode, rec.Body)
}
checkJSONErrorCode(t, name, rec, tt.wantCode)
checkJSONErrorMessage(t, name, rec, tt.wantMsg)
allow := rec.Header().Get("Allow")
if tt.wantCode == http.StatusMethodNotAllowed && allow != jsonRoutesAllow {
t.Errorf("%s: Allow %q, want %q", name, allow, jsonRoutesAllow)
}
}
}
// checkJSONErrorMessage checks that rec holds the JSON error body with the
// given message.
func checkJSONErrorMessage(
t *testing.T, name string, rec *httptest.ResponseRecorder, want string,
) {
t.Helper()
var body struct {
Error struct {
Msg string `json:"msg"`
} `json:"error"`
}
err := json.Unmarshal(rec.Body.Bytes(), &body)
if err != nil {
t.Errorf("%s: decoding the answer %q: %v", name, rec.Body, err)
return
}
if body.Error.Msg != want {
t.Errorf("%s: error message %q, want %q", name, body.Error.Msg, want)
}
}
internal/server/server.go
+7 -33
View File
@@ -7,7 +7,6 @@ import (
"fmt"
"net"
"net/http"
"net/netip"
"strconv"
"time"
@@ -32,13 +31,6 @@ type ASNFetcher interface {
GetStats() ASNFetcherStats
}
// dnsResolver looks up the addresses of host names and the names of addresses
// in DNS, as *net.Resolver does.
type dnsResolver interface {
LookupHost(ctx context.Context, host string) ([]string, error)
LookupAddr(ctx context.Context, addr string) ([]string, error)
}
// Server provides HTTP endpoints for status monitoring
type Server struct {
router *chi.Mux
@@ -48,14 +40,6 @@ type Server struct {
port int
srv *http.Server
asnFetcher ASNFetcher
// trustedProxies are the reverse proxies whose X-Forwarded-For header
// RealIPMiddleware believes
trustedProxies []netip.Prefix
// resolver makes the IP lookup's DNS lookups: net.DefaultResolver, or a
// fake in tests
resolver dnsResolver
}
// New creates a new HTTP server
@@ -66,12 +50,10 @@ func New(
cfg *config.Config,
) *Server {
s := &Server{
db: db,
streamer: streamer,
logger: logger,
port: cfg.Port,
trustedProxies: cfg.TrustedProxies,
resolver: net.DefaultResolver,
db: db,
streamer: streamer,
logger: logger,
port: cfg.Port,
}
s.setupRoutes()
@@ -121,9 +103,7 @@ func (s *Server) Start(ctx context.Context) error {
return nil
}
// Stop stops the HTTP server. It waits, until ctx ends, for the requests in
// progress to finish, and then closes the connections of those still running
// and returns ctx's error.
// Stop gracefully stops the HTTP server
func (s *Server) Stop(ctx context.Context) error {
if s.srv == nil {
return nil
@@ -131,16 +111,10 @@ func (s *Server) Stop(ctx context.Context) error {
s.logger.Info("Stopping HTTP server")
err := s.srv.Shutdown(ctx)
if err != nil {
return errors.Join(err, s.srv.Close())
}
return nil
return s.srv.Shutdown(ctx)
}
// SetASNFetcher sets the ASN WHOIS fetcher for on-demand lookups. Call it
// before Start: the handlers read the fetcher without a lock.
// SetASNFetcher sets the ASN WHOIS fetcher for on-demand lookups.
func (s *Server) SetASNFetcher(fetcher ASNFetcher) {
s.asnFetcher = fetcher
}
internal/server/server_internal_test.go
-92
View File
@@ -1,92 +0,0 @@
package server
import (
"context"
"net"
"net/http"
"testing"
"time"
"git.eeqj.de/sneak/routewatch/internal/logger"
)
// TestStopClosesTheConnectionOfARequestStillRunning serves a request whose
// handler does not return until the test ends, and stops the server with a
// deadline that passes while the request runs. The client must then see its
// connection closed. Before the fix Stop returned at the deadline and left the
// connection open until the handler returned.
func TestStopClosesTheConnectionOfARequestStillRunning(t *testing.T) {
t.Parallel()
const (
stopDeadline = 100 * time.Millisecond
// waitTimeout is how long the test waits for the request to reach
// the handler, and then for the client to see its connection closed.
waitTimeout = 2 * time.Second
// readHeaderTimeout bounds the test server's read of the request
// header, as every http.Server must.
readHeaderTimeout = time.Second
)
release := make(chan struct{})
defer close(release)
started := make(chan struct{})
var listenConfig net.ListenConfig
listener, err := listenConfig.Listen(t.Context(), "tcp", "127.0.0.1:0")
if err != nil {
t.Fatalf("failed to open the test listener: %v", err)
}
s := &Server{logger: logger.New(), srv: &http.Server{
Handler: http.HandlerFunc(func(http.ResponseWriter, *http.Request) {
close(started)
<-release
}),
ReadHeaderTimeout: readHeaderTimeout,
}}
go func() { _ = s.srv.Serve(listener) }()
answered := make(chan error, 1)
go func() {
req, err := http.NewRequestWithContext(t.Context(), http.MethodGet,
"http://"+listener.Addr().String()+"/", nil)
if err != nil {
answered <- err
return
}
resp, err := http.DefaultClient.Do(req)
if err == nil {
_ = resp.Body.Close()
}
answered <- err
}()
select {
case <-started:
case <-time.After(waitTimeout):
t.Fatal("the request did not reach the handler")
}
ctx, cancel := context.WithTimeout(t.Context(), stopDeadline)
defer cancel()
_ = s.Stop(ctx)
select {
case err := <-answered:
if err == nil {
t.Error("the request still running at the deadline was answered")
}
case <-time.After(waitTimeout):
t.Fatal("the connection of the request still running at the " +
"deadline was open after Stop returned")
}
}
internal/streamer/streamer.go
+13 -42
View File
@@ -106,12 +106,12 @@ type RawMessageHandler func(line string)
// handlerMetrics tracks performance metrics for a handler
type handlerMetrics struct {
processedCount uint64 // Total messages processed
droppedCount atomic.Uint64 // Total messages dropped; not under mu
droppedCount uint64 // Total messages dropped
totalTime time.Duration // Total processing time (for average calculation)
minTime time.Duration // Minimum processing time
maxTime time.Duration // Maximum processing time
queueHighWaterMark int // Maximum queue length seen
mu sync.Mutex // Protects the other metrics
mu sync.Mutex // Protects the metrics
}
// handlerInfo wraps a handler with its queue and metrics
@@ -253,10 +253,9 @@ func (s *Streamer) Start(ctx context.Context) error {
// Stop ends the stream and stops the handler workers. It returns once each
// worker has handed its handler every message left on its queue and then the
// STATE messages held for it (see queueMessage), or once ctx ends, and then
// logs how many messages were not handed over. This method is safe to call
// STATE messages held for it (see queueMessage). This method is safe to call
// multiple times.
func (s *Streamer) Stop(ctx context.Context) {
func (s *Streamer) Stop() {
s.mu.Lock()
if s.cancel == nil {
// Not started, or already stopped: closing the queues again would panic.
@@ -265,56 +264,28 @@ func (s *Streamer) Stop(ctx context.Context) {
return
}
// Once the stream's context is cancelled the read loop queues and holds
// nothing more (see dispatchMessage).
// Once ctx is cancelled the read loop queues and holds nothing more (see
// dispatchMessage).
s.cancel()
s.cancel = nil
notHandedOver := 0
for _, info := range s.handlers {
// The held messages came after everything on the queue. Each waits
// for the worker, which does not take s.mu, to make room, until ctx
// ends.
for len(info.heldState) > 0 && ctx.Err() == nil {
select {
case info.queue <- info.heldState[0]:
info.heldState = info.heldState[1:]
case <-ctx.Done():
}
// for the worker, which does not take s.mu, to make room.
for _, msg := range info.heldState {
info.queue <- msg
}
notHandedOver += len(info.heldState)
info.heldState = nil
// The worker returns once it has emptied the closed queue.
close(info.queue)
}
handlers := s.handlers
s.running = false
s.mu.Unlock()
s.metrics.SetConnected(false)
workersDone := make(chan struct{})
go func() {
s.workers.Wait()
close(workersDone)
}()
select {
case <-workersDone:
case <-ctx.Done():
for _, info := range handlers {
notHandedOver += len(info.queue)
}
}
if notHandedOver > 0 {
s.logger.Error("Stop deadline passed before the handlers took every "+
"message", "messages", notHandedOver)
}
s.workers.Wait()
}
// IsRunning reports whether the streamer is currently connected and processing
@@ -371,7 +342,7 @@ func (s *Streamer) GetHandlerStats() []HandlerStats {
QueueCapacity: cap(info.queue),
QueueHighWaterMark: info.metrics.queueHighWaterMark,
ProcessedCount: info.metrics.processedCount,
DroppedCount: info.metrics.droppedCount.Load(),
DroppedCount: info.metrics.droppedCount,
MinProcessTime: info.metrics.minTime,
MaxProcessTime: info.metrics.maxTime,
}
@@ -488,7 +459,7 @@ func (s *Streamer) logMetrics() {
"queue_len", len(info.queue),
"queue_cap", cap(info.queue),
"processed", info.metrics.processedCount,
"dropped", info.metrics.droppedCount.Load(),
"dropped", info.metrics.droppedCount,
"avg_time", avgTime,
"min_time", info.metrics.minTime,
"max_time", info.metrics.maxTime,
@@ -858,7 +829,7 @@ func (s *Streamer) dispatchMessage(
}
if !s.queueMessage(info, msg) {
info.metrics.droppedCount.Add(1)
atomic.AddUint64(&info.metrics.droppedCount, 1)
s.totalDropped.Add(1)
continue
internal/streamer/streamer_internal_test.go
+4 -131
View File
@@ -1,12 +1,9 @@
package streamer
import (
"bytes"
"context"
"encoding/json"
"errors"
"io"
"log/slog"
"net/http"
"net/http/httptest"
"runtime"
@@ -380,7 +377,7 @@ func TestStopQueuesTheHeldStateMessages(t *testing.T) {
stopped := make(chan struct{})
go func() {
s.Stop(t.Context())
s.Stop()
close(stopped)
}()
@@ -415,68 +412,6 @@ func TestStopQueuesTheHeldStateMessages(t *testing.T) {
}
}
// TestStopReturnsByItsDeadline stops a streamer whose handler has a full queue
// and a STATE message held for it, and a worker that stays busy until the test
// ends, as one does whose handler waits for a database write. Stop must return
// within its deadline plus a small margin instead of waiting for room on the
// queue and for the worker, and log as not handed over the held message and
// every message on the queue.
func TestStopReturnsByItsDeadline(t *testing.T) {
t.Parallel()
const (
deadline = 100 * time.Millisecond
margin = 2 * time.Second
)
var logs bytes.Buffer
s := newStreamerWithFullStateQueue("")
s.logger = &logger.Logger{Logger: slog.New(slog.NewJSONHandler(&logs,
&slog.HandlerOptions{Level: slog.LevelError}))}
s.handlers[0].heldState = []*ristypes.RISMessage{
{Type: ristypes.MessageTypeState},
}
wantNotHandedOver := 1 + len(s.handlers[0].queue)
// Stop cancels the stream Start would have started; there is none here.
s.cancel = func() {}
release := make(chan struct{})
defer close(release)
s.workers.Go(func() { <-release })
ctx, cancel := context.WithTimeout(t.Context(), deadline)
defer cancel()
stopped := make(chan struct{})
go func() {
s.Stop(ctx)
close(stopped)
}()
select {
case <-stopped:
case <-time.After(deadline + margin):
t.Fatalf("Stop did not return within %s of its deadline", margin)
}
var logged struct {
Messages int `json:"messages"`
}
err := json.Unmarshal(logs.Bytes(), &logged)
if err != nil {
t.Fatalf("want the error Stop logs, got %q: %v", logs.String(), err)
}
if logged.Messages != wantNotHandedOver {
t.Errorf("Stop logged %d messages not handed over, want %d",
logged.Messages, wantNotHandedOver)
}
}
// TestStopBeforeMessageReachesQueues stops the streamer after the read loop
// has checked for cancellation but before it hands the message to the handler
// queues. That is the gap a stop from another goroutine can land in, and it
@@ -494,7 +429,7 @@ func TestStopBeforeMessageReachesQueues(t *testing.T) {
s := New(logger.New(), metrics.New())
s.url = srv.URL
s.RegisterHandler(updateHandler{})
s.RegisterRawHandler(func(string) { s.Stop(t.Context()) })
s.RegisterRawHandler(func(string) { s.Stop() })
// Start would run the stream in the background, where the test cannot
// wait for it. Setting cancel as Start does lets Stop cancel the stream
@@ -508,7 +443,7 @@ func TestStopBeforeMessageReachesQueues(t *testing.T) {
}
// A second Stop must not close the queues again.
s.Stop(t.Context())
s.Stop()
}
// TestReconnectWhenStreamEnds has the server end the first response
@@ -621,68 +556,6 @@ func TestStreamReadsLongLine(t *testing.T) {
}
}
// blockedHandler wants UPDATE messages, has room on its queue for one, and
// does not return from HandleMessage until release is closed.
type blockedHandler struct {
release chan struct{}
}
func (blockedHandler) WantsMessage(messageType string) bool {
return messageType == ristypes.MessageTypeUpdate
}
func (blockedHandler) QueueCapacity() int { return 1 }
func (h blockedHandler) HandleMessage(*ristypes.RISMessage) { <-h.release }
// TestHandlerStatsReadWhileMessagesAreDropped has the server send a few UPDATE
// lines to a blockedHandler, so the read loop drops all but the one the
// handler holds and the one on its queue, while the test reads the handler's
// statistics. The drop count must reach the number dropped, and the race
// detector must find the count read and written safely.
func TestHandlerStatsReadWhileMessagesAreDropped(t *testing.T) {
t.Parallel()
const (
lines = 5
wantDropped = lines - 2
maxWait = 5 * time.Second
)
srv := httptest.NewServer(http.HandlerFunc(
func(w http.ResponseWriter, r *http.Request) {
_, _ = io.WriteString(w, strings.Repeat(updateLine, lines))
flush(w)
<-r.Context().Done()
}))
defer srv.Close()
s := New(logger.New(), metrics.New())
s.url = srv.URL
handler := blockedHandler{release: make(chan struct{})}
s.RegisterHandler(handler)
err := s.Start(t.Context())
if err != nil {
t.Fatalf("Start returned error: %v", err)
}
defer s.Stop(t.Context())
defer close(handler.release)
// The test polls instead of waiting for a signal from the read loop: a
// signal would order the drops before the reads here, and the race
// detector would have nothing to report.
deadline := time.Now().Add(maxWait)
for s.GetHandlerStats()[0].DroppedCount < wantDropped {
if time.Now().After(deadline) {
t.Fatalf("fewer than %d messages counted as dropped within %s",
wantDropped, maxWait)
}
}
}
// TestPeerCountCountsEveryPeerHeardFrom has the server send an UPDATE from
// stateDownPeer and a KEEPALIVE from another peer, with no OPEN from either,
// and then a message that ends stateDownPeer's session: a NOTIFICATION, or a
@@ -803,7 +676,7 @@ func checkReconnect(
if err != nil {
t.Fatalf("Start returned error: %v", err)
}
defer s.Stop(t.Context())
defer s.Stop()
select {
case <-handler.handled:
internal/templates/as_detail.html
+1 -1
View File
@@ -4,7 +4,7 @@
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>AS{{.ASN.ASN}} - {{.ASN.Handle}} - RouteWatch</title>
<style nonce="{{.CSPNonce}}">
<style>
* {
box-sizing: border-box;
}
internal/templates/index.html
+4 -11
View File
@@ -4,7 +4,7 @@
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>RouteWatch - BGP Route Monitor</title>
<style nonce="{{.CSPNonce}}">
<style>
* {
box-sizing: border-box;
}
@@ -167,13 +167,6 @@
color: #95a5a6;
margin-top: 8px;
}
#asname-results p {
color: #7f8c8d;
margin-top: 12px;
}
#asname-results .search-message {
font-size: 13px;
}
/* IP Lookup result */
.ip-result {
@@ -336,7 +329,7 @@
<span><a href="{{appGitCommitURL}}">{{appGitRevision}}</a></span>
</footer>
<script nonce="{{.CSPNonce}}">
<script>
function formatNumber(num) {
if (num >= 1000000) {
return (num / 1000000).toFixed(1) + 'M';
@@ -394,7 +387,7 @@
if (!query) return;
const resultsDiv = document.getElementById('asname-results');
resultsDiv.innerHTML = '<p class="loading">Searching...</p>';
resultsDiv.innerHTML = '<p class="loading" style="color: #7f8c8d; margin-top: 12px;">Searching...</p>';
// Use a simple client-side search against the asinfo data
// For now, redirect to AS page if it looks like an ASN
@@ -404,7 +397,7 @@
}
// Show a message that server-side search is coming
resultsDiv.innerHTML = '<p class="search-message">AS name search coming soon. For now, try an AS number.</p>';
resultsDiv.innerHTML = '<p style="color: #7f8c8d; margin-top: 12px; font-size: 13px;">AS name search coming soon. For now, try an AS number.</p>';
});
// IP lookup
internal/templates/prefix_detail.html
+2 -6
View File
@@ -4,7 +4,7 @@
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>{{.Prefix}} - RouteWatch</title>
<style nonce="{{.CSPNonce}}">
<style>
* {
box-sizing: border-box;
}
@@ -221,10 +221,6 @@
border-radius: 4px;
border: 1px solid #e0e0e0;
}
.origin-item .peer-count {
color: #7f8c8d;
margin-left: 10px;
}
.empty-state {
text-align: center;
padding: 40px;
@@ -292,7 +288,7 @@
{{else}}
No known origin
{{end}}
<span class="peer-count">{{.PeerCount}} peer{{if ne .PeerCount 1}}s{{end}}</span>
<span style="color: #7f8c8d; margin-left: 10px;">{{.PeerCount}} peer{{if ne .PeerCount 1}}s{{end}}</span>
</div>
{{end}}
</div>
internal/templates/prefix_length.html
+1 -1
View File
@@ -4,7 +4,7 @@
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Prefixes with /{{ .MaskLength }} - RouteWatch</title>
<style nonce="{{.CSPNonce}}">
<style>
* {
box-sizing: border-box;
}
internal/templates/status.html
+4 -5
View File
@@ -4,7 +4,7 @@
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>RouteWatch Status</title>
<style nonce="{{.CSPNonce}}">
<style>
* {
box-sizing: border-box;
}
@@ -130,7 +130,6 @@
color: #ef4444;
}
.error {
display: none;
background: #fee;
color: #c00;
padding: 10px;
@@ -174,7 +173,7 @@
</nav>
<main class="main-content">
<div id="error" class="error"></div>
<div id="error" class="error" style="display: none;"></div>
<div class="status-grid">
<div class="status-card">
<h2>RouteWatch</h2>
@@ -384,8 +383,8 @@
<div id="handler-stats-container" class="status-grid">
<!-- Handler stats will be dynamically added here -->
</div>
<script nonce="{{.CSPNonce}}">
<script>
function formatBytes(bytes) {
if (bytes === 0) return '0 B';
const k = 1024;
internal/templates/templates_test.go
-54
View File
@@ -1,54 +0,0 @@
package templates_test
import (
"bytes"
"html/template"
"strings"
"testing"
"git.eeqj.de/sneak/routewatch/internal/templates"
)
// TestFooterLicense renders the home and status pages and expects each footer
// to name the licence in LICENSE, MIT, and not WTFPL.
func TestFooterLicense(t *testing.T) {
t.Parallel()
const (
license = "MIT"
wrongLicense = "WTFPL"
)
pages := map[string]*template.Template{
"index": templates.IndexTemplate(),
"status": templates.StatusTemplate(),
}
for name, tmpl := range pages {
t.Run(name, func(t *testing.T) {
t.Parallel()
var page bytes.Buffer
err := tmpl.Execute(&page, struct{ CSPNonce string }{})
if err != nil {
t.Fatalf("rendering the page: %v", err)
}
_, footer, found := strings.Cut(page.String(), "<footer")
if !found {
t.Fatal("the page has no footer")
}
footer, _, _ = strings.Cut(footer, "</footer>")
if !strings.Contains(footer, license) {
t.Errorf("footer does not name %s: %q", license, footer)
}
if strings.Contains(footer, wrongLicense) {
t.Errorf("footer names %s: %q", wrongLicense, footer)
}
})
}
}
internal/version/version.go
+2 -2
View File
@@ -19,8 +19,8 @@ const (
Author = "@sneak"
// AuthorURL is the author's website
AuthorURL = "https://sneak.berlin"
// License is the program license, as LICENSE grants it
License = "MIT"
// License is the program license
License = "WTFPL"
// RepoURL is the git repository URL
RepoURL = "https://git.eeqj.de/sneak/routewatch"
)
internal/whois/whois.go
+66 -86
View File
@@ -4,9 +4,7 @@ package whois
import (
"bufio"
"context"
"errors"
"fmt"
"io"
"net"
"regexp"
"strings"
@@ -20,10 +18,6 @@ const (
writeTimeout = 5 * time.Second
)
// maxAnswerSize caps the bytes read of one WHOIS answer: a real one is a
// few KiB, so one this large is from a server that does not stop sending.
const maxAnswerSize = 1 << 20
// Parsing constants.
const (
keyValueParts = 2 // Expected parts when splitting "key: value"
@@ -49,19 +43,6 @@ const (
RIRAFRNIC = "AFRINIC"
)
// Errors LookupASN returns when a WHOIS answer has nothing to store.
var (
// ErrNoReferral is returned when IANA refers the AS to no RIR, as it
// does for a reserved, private-use or unallocated AS number.
ErrNoReferral = errors.New("IANA refers the AS to no RIR")
errNoASBlock = errors.New("IANA answered with neither a referral " +
"nor an AS block")
errNoFields = errors.New("RIR answered with no WHOIS fields")
errUnknownServer = errors.New("IANA refers the AS to a server no RIR runs")
errAnswerTooLarge = errors.New("WHOIS answer reached the size cap")
)
// ASNInfo contains parsed WHOIS information for an ASN.
type ASNInfo struct {
ASN int
@@ -82,25 +63,20 @@ type ASNInfo struct {
// Client performs WHOIS lookups for ASNs.
type Client struct {
// DialContext opens the connection to a WHOIS server, given as
// host:port. NewClient sets a net.Dialer's; tests replace it to reach
// local servers instead.
DialContext func(ctx context.Context, network, address string) (net.Conn, error)
// Dialer for creating connections (can be overridden for testing)
dialer *net.Dialer
}
// NewClient creates a new WHOIS client.
func NewClient() *Client {
dialer := &net.Dialer{Timeout: dialTimeout}
return &Client{DialContext: dialer.DialContext}
return &Client{
dialer: &net.Dialer{
Timeout: dialTimeout,
},
}
}
// LookupASN asks IANA which RIR holds the given ASN, asks that RIR, and
// returns what the RIR's answer says. It returns ErrNoReferral when IANA
// refers the AS to no RIR, and another error when a query fails, when IANA's
// answer has neither a referral nor the AS's block, when IANA refers the AS
// to a server none of the five RIRs run, or when the RIR's answer carries
// none of ASNInfo's fields, as a refusal of service does.
// LookupASN queries WHOIS for the given ASN and returns parsed information.
func (c *Client) LookupASN(ctx context.Context, asn int) (*ASNInfo, error) {
// Query IANA first to find the authoritative RIR
query := fmt.Sprintf("AS%d", asn)
@@ -111,53 +87,34 @@ func (c *Client) LookupASN(ctx context.Context, asn int) (*ASNInfo, error) {
}
// Determine RIR from IANA response
rir, whoisServer, err := c.parseIANAReferral(ianaResp)
if err != nil {
return nil, fmt.Errorf("%s: %w", query, err)
}
rir, whoisServer := c.parseIANAReferral(ianaResp)
if whoisServer == "" {
// IANA names the block of an AS it refers to no RIR. An answer
// without that, such as an empty one, says nothing about the AS.
if !strings.Contains(ianaResp, "as-block:") {
return nil, fmt.Errorf("%s: %w", query, errNoASBlock)
}
return nil, fmt.Errorf("%s: %w", query, ErrNoReferral)
// No referral, try to parse what we have
return c.parseResponse(asn, rir, ianaResp), nil
}
// Query the authoritative RIR
rirResp, err := c.query(ctx, whoisServer, query)
if err != nil {
return nil, fmt.Errorf("RIR query failed: %w", err)
// Return partial data from IANA if RIR query fails
info := c.parseResponse(asn, rir, ianaResp)
info.RawResponse = fmt.Sprintf("%s\n--- RIR query failed: %v ---\n",
ianaResp, err)
return info, nil
}
// The fields come from the RIR's answer alone: IANA's describes the
// block the AS is in.
info := c.parseResponse(asn, rir, rirResp)
if info.isEmpty() {
return nil, fmt.Errorf("%s from %s: %w", query, whoisServer, errNoFields)
}
info.RawResponse = ianaResp + "\n" + rirResp
// Combine responses and parse
fullResponse := ianaResp + "\n" + rirResp
info := c.parseResponse(asn, rir, fullResponse)
info.RawResponse = fullResponse
return info, nil
}
// isEmpty reports whether info holds none of the fields a WHOIS answer
// fills, as when the answer is only comment and error lines.
func (info *ASNInfo) isEmpty() bool {
return info.ASName == "" && info.OrgName == "" && info.OrgID == "" &&
info.Address == "" && info.CountryCode == "" &&
info.AbuseEmail == "" && info.AbusePhone == "" &&
info.TechEmail == "" && info.TechPhone == "" &&
info.RegDate == nil && info.LastMod == nil
}
// query performs a raw WHOIS query to the specified server. An answer that
// reaches maxAnswerSize is an error.
// query performs a raw WHOIS query to the specified server.
func (c *Client) query(ctx context.Context, server, query string) (string, error) {
conn, err := c.DialContext(ctx, "tcp", server)
conn, err := c.dialer.DialContext(ctx, "tcp", server)
if err != nil {
return "", fmt.Errorf("dial %s: %w", server, err)
}
@@ -183,11 +140,7 @@ func (c *Client) query(ctx context.Context, server, query string) (string, error
var sb strings.Builder
// The reader ends at the cap; N, the bytes it has left, is 0 once the
// answer has reached it.
answer := &io.LimitedReader{R: conn, N: maxAnswerSize}
scanner := bufio.NewScanner(answer)
scanner := bufio.NewScanner(conn)
for scanner.Scan() {
sb.WriteString(scanner.Text())
sb.WriteString("\n")
@@ -198,18 +151,13 @@ func (c *Client) query(ctx context.Context, server, query string) (string, error
return sb.String(), fmt.Errorf("read response: %w", err)
}
if answer.N == 0 {
return "", fmt.Errorf("read response: %w", errAnswerTooLarge)
}
return sb.String(), nil
}
// parseIANAReferral extracts the RIR and WHOIS server from an IANA response.
// The server is "" when the response has no referral. A referral to a server
// that is none of the five RIRs' is an error, so a forged or broken answer
// cannot send the client to any other host or port.
func (c *Client) parseIANAReferral(response string) (string, string, error) {
func (c *Client) parseIANAReferral(response string) (string, string) {
var rir string
for line := range strings.SplitSeq(response, "\n") {
line = strings.TrimSpace(line)
@@ -220,22 +168,54 @@ func (c *Client) parseIANAReferral(response string) (string, string, error) {
switch {
case strings.Contains(server, "arin"):
return RIRARIN, whoisServerARIN, nil
return RIRARIN, whoisServerARIN
case strings.Contains(server, "ripe"):
return RIRRIPE, whoisServerRIPE, nil
return RIRRIPE, whoisServerRIPE
case strings.Contains(server, "apnic"):
return RIRAPNIC, whoisServerAPNIC, nil
return RIRAPNIC, whoisServerAPNIC
case strings.Contains(server, "lacnic"):
return RIRLACNIC, whoisServerLACNIC, nil
return RIRLACNIC, whoisServerLACNIC
case strings.Contains(server, "afrinic"):
return RIRAFRNIC, whoisServerAFRINIC, nil
return RIRAFRNIC, whoisServerAFRINIC
default:
return "", "", fmt.Errorf("%w: %q", errUnknownServer, server)
// Unknown server, add port if missing
if !strings.Contains(server, ":") {
server += ":43"
}
return "", server
}
}
// Also check organisation line for RIR hints
if strings.HasPrefix(strings.ToLower(line), "organisation:") {
orgRIR := rirInOrganisationLine(strings.ToLower(line))
if orgRIR != "" {
rir = orgRIR
}
}
}
return "", "", nil
return rir, ""
}
// rirInOrganisationLine returns the RIR a lowercased IANA "organisation:"
// line names, or "" when it names none.
func rirInOrganisationLine(org string) string {
switch {
case strings.Contains(org, "arin"):
return RIRARIN
case strings.Contains(org, "ripe"):
return RIRRIPE
case strings.Contains(org, "apnic"):
return RIRAPNIC
case strings.Contains(org, "lacnic"):
return RIRLACNIC
case strings.Contains(org, "afrinic"):
return RIRAFRNIC
default:
return ""
}
}
// parseResponse extracts ASN information from a WHOIS response.
internal/whois/whois_test.go
-246
View File
@@ -1,246 +0,0 @@
package whois_test
import (
"bufio"
"context"
"errors"
"io"
"net"
"strings"
"sync"
"syscall"
"testing"
"time"
"git.eeqj.de/sneak/routewatch/internal/whois"
)
// The servers LookupASN asks: IANA first, then the RIR IANA refers it to.
const (
ianaServer = "whois.iana.org:43"
ripeServer = "whois.ripe.net:43"
)
// testASN is the AS the tests look up.
const testASN = 3333
// ianaReferral is IANA's answer for an AS in a block it gave to RIPE.
const ianaReferral = `% IANA WHOIS server
refer: whois.ripe.net
as-block: 3154-3353
organisation: Assigned by RIPE NCC
whois: whois.ripe.net
changed: 2002-09-01
source: IANA
`
// ripeRecord is RIPE's answer for the AS.
const ripeRecord = `% This is the RIPE Database query service.
aut-num: AS3333
as-name: RIPE-NCC-AS
org: ORG-RIEN1-RIPE
created: 2002-09-17T13:08:32Z
last-modified: 2024-01-09T08:55:41Z
source: RIPE
`
// ripeRefusal is RIPE's answer to a host it no longer serves.
const ripeRefusal = `% This is the RIPE Database query service.
% The objects are in RPSL format.
%
%ERROR:201: access denied for 192.0.2.1
%
% Sorry, access from your host has been permanently
% denied because of a repeated excessive querying.
`
// ianaPrivateUse is IANA's answer for a private-use AS: its block and no
// referral.
const ianaPrivateUse = `% IANA WHOIS server
as-block: 64512-65534
organisation: Reserved for Private Use
descr: Reserved for Private Use
source: IANA
`
// ianaNoObject is IANA's answer when it has nothing for the query.
const ianaNoObject = `% IANA WHOIS server
% This query returned 0 objects.
`
// fillerLine is one of the lines a server that does not stop keeps sending.
const fillerLine = "remarks: and more\n"
// endlessAnswerSize is how many bytes of filler lines that server sends:
// four times the client's cap on an answer.
const endlessAnswerSize = 4 << 20
// fakeServers returns a DialContext for a whois.Client that connects to a
// fake server for each address in answers, which reads the query and sends
// answers[address]. Connecting to any other address is refused.
func fakeServers(
t *testing.T, answers map[string]string,
) func(context.Context, string, string) (net.Conn, error) {
t.Helper()
var servers sync.WaitGroup
t.Cleanup(servers.Wait)
return func(_ context.Context, _, address string) (net.Conn, error) {
answer, ok := answers[address]
if !ok {
return nil, syscall.ECONNREFUSED
}
client, server := net.Pipe()
servers.Go(func() {
defer func() { _ = server.Close() }()
_, _ = bufio.NewReader(server).ReadString('\n')
_, _ = io.WriteString(server, answer)
})
return client, nil
}
}
// lookUp looks testASN up with fake servers sending answers.
func lookUp(
t *testing.T, answers map[string]string,
) (*whois.ASNInfo, error) {
t.Helper()
client := whois.NewClient()
client.DialContext = fakeServers(t, answers)
return client.LookupASN(t.Context(), testASN)
}
// TestLookupASNReadsTheRIRAnswer looks up an AS IANA refers to RIPE. The
// fields come from RIPE's answer, the last-modified date included: IANA's
// "changed:" line dates its block, not the AS.
func TestLookupASNReadsTheRIRAnswer(t *testing.T) {
t.Parallel()
info, err := lookUp(t, map[string]string{
ianaServer: ianaReferral,
ripeServer: ripeRecord,
})
if err != nil {
t.Fatalf("LookupASN: %v", err)
}
if info.ASName != "RIPE-NCC-AS" || info.OrgID != "ORG-RIEN1-RIPE" ||
info.RIR != whois.RIRRIPE {
t.Errorf("LookupASN = AS name %q, org %q, RIR %q; "+
"want RIPE-NCC-AS, ORG-RIEN1-RIPE, RIPE",
info.ASName, info.OrgID, info.RIR)
}
wantLastMod := time.Date(2024, 1, 9, 8, 55, 41, 0, time.UTC)
if info.LastMod == nil || !info.LastMod.Equal(wantLastMod) {
t.Errorf("LookupASN last modified = %v, want %v",
info.LastMod, wantLastMod)
}
}
// TestLookupASNFailsWhenTheRIRGivesNoAnswer looks up an AS IANA refers to
// RIPE when RIPE's server refuses the connection, and when it answers only
// with an error. Neither is a lookup that found the AS.
func TestLookupASNFailsWhenTheRIRGivesNoAnswer(t *testing.T) {
t.Parallel()
for name, answers := range map[string]map[string]string{
"connection refused": {ianaServer: ianaReferral},
"only an error": {ianaServer: ianaReferral, ripeServer: ripeRefusal},
} {
t.Run(name, func(t *testing.T) {
t.Parallel()
_, err := lookUp(t, answers)
if err == nil || errors.Is(err, whois.ErrNoReferral) {
t.Errorf("LookupASN error = %v, want one that is not "+
"ErrNoReferral", err)
}
})
}
}
// TestLookupASNReportsNoReferral looks up a private-use AS, which IANA
// refers to no RIR.
func TestLookupASNReportsNoReferral(t *testing.T) {
t.Parallel()
_, err := lookUp(t, map[string]string{ianaServer: ianaPrivateUse})
if !errors.Is(err, whois.ErrNoReferral) {
t.Errorf("LookupASN error = %v, want ErrNoReferral", err)
}
}
// TestLookupASNFailsWhenIANAAnswersNoBlock looks up an AS when IANA answers
// with neither a referral nor the AS's block. That is not IANA saying no
// RIR holds the AS.
func TestLookupASNFailsWhenIANAAnswersNoBlock(t *testing.T) {
t.Parallel()
_, err := lookUp(t, map[string]string{ianaServer: ianaNoObject})
if err == nil || errors.Is(err, whois.ErrNoReferral) {
t.Errorf("LookupASN error = %v, want one that is not ErrNoReferral",
err)
}
}
// TestLookupASNFailsWhenTheRIRAnswerDoesNotEnd looks up an AS whose RIR
// server follows the record with filler lines far past any real answer.
// The client stops reading at its cap and fails instead of keeping them.
func TestLookupASNFailsWhenTheRIRAnswerDoesNotEnd(t *testing.T) {
t.Parallel()
endless := ripeRecord +
strings.Repeat(fillerLine, endlessAnswerSize/len(fillerLine))
_, err := lookUp(t, map[string]string{
ianaServer: ianaReferral,
ripeServer: endless,
})
if err == nil || errors.Is(err, whois.ErrNoReferral) {
t.Errorf("LookupASN error = %v, want one that is not ErrNoReferral",
err)
}
}
// TestLookupASNFailsOnAReferralToAnotherServer looks up an AS when IANA's
// answer refers it to a server none of the five RIRs run, with a port and
// without. The client follows a referral only to an RIR's server, so it
// fails instead of asking that server, which would answer with a record.
func TestLookupASNFailsOnAReferralToAnotherServer(t *testing.T) {
t.Parallel()
// Each case maps the server on the "refer:" line to the address the
// client would dial to follow it. IANA's answer also names the AS's
// block, so a client that dropped the referral would report no RIR.
for refer, address := range map[string]string{
"whois.example.net:8080": "whois.example.net:8080",
"whois.example.net": "whois.example.net:43",
} {
t.Run(refer, func(t *testing.T) {
t.Parallel()
_, err := lookUp(t, map[string]string{
ianaServer: "refer: " + refer + "\n\n" +
"as-block: 3154-3353\n",
address: ripeRecord,
})
if err == nil || errors.Is(err, whois.ErrNoReferral) {
t.Errorf("LookupASN error = %v, want one that is not "+
"ErrNoReferral", err)
}
})
}
}