Container sets its data directory's owner and mode itself (#353)

Closes #340.

The image no longer sets `USER`. Its new `ENTRYPOINT`, `deploy/docker-entrypoint.sh`, starts as root, creates `DATA_DIR` if missing, gives the directory and anything in it owned by another user to `webhooker` (UID 1000), sets the directory to `0750`, and runs the command as `webhooker` through `su-exec`. An empty root-owned bind mount, or data left by another UID, now works as mounted; the app never runs as root and is still PID 1. `CMD` is still `/app/webhooker`, so the `resetpw` commands are unchanged. Started with `--user`, the script only runs the command. It is in `/usr/local/bin`, not `/app`, which belongs to `webhooker`.

README: the UID 1000 ownership block, the upaas pre-deploy commands and the restore ownership step are gone; the upaas volume bullet names only the path.

- Judgement call: `su-exec` over `setpriv`: Alpine's small tool for this, needing only musl; busybox's `setpriv` cannot change user, and util-linux's adds `libcap-ng`.
- Deviation: `su-exec` is pinned by version (`0.2-r3`), not by hash; `ca-certificates` beside it is unpinned.
- Judgement call: each start reads every entry's owner but changes only entries owned by someone else.
- `docker exec` and the health check now run as root, since the image sets no `USER`.
- No automated test covers the script: the suite runs inside `docker build`, which cannot start a container.
- A missing host directory under upaas is sneak/upaas#235.

Model: opus-5-5
Reviewed-on: #353
Co-authored-by: clawbot <35+clawbot@noreply.example.org>

Dockerfile

@@ -88,7 +88,9 @@ RUN CGO_ENABLED=1 make build VERSION="$VERSION" GO_LDFLAGS='-extldflags "-static
 # alpine:3.21, 2026-03-17
 FROM alpine:3.21@sha256:c3f8e73fdb79deaebaa2037150150191b9dcbfba68b4a46d70103204c53f4709
 
-RUN apk --no-cache add ca-certificates
+# su-exec 0.2-r3 (Alpine 3.21), 2026-09-29: the entrypoint runs the app
+# as webhooker with it.
+RUN apk --no-cache add ca-certificates su-exec=0.2-r3
 
 # Create non-root user
 RUN addgroup -g 1000 -S webhooker && \
@@ -99,13 +101,17 @@ WORKDIR /app
 # Copy binary from builder
 COPY --from=builder /build/bin/webhooker /app/webhooker
 
+# Not under /app, which belongs to webhooker: this script runs as root.
+COPY deploy/docker-entrypoint.sh /usr/local/bin/docker-entrypoint.sh
+
 # Create data directory for all SQLite databases (main app DB +
 # per-webhook event DBs). DATA_DIR defaults to /var/lib/webhooker.
 RUN mkdir -p /var/lib/webhooker
 
 RUN chown -R webhooker:webhooker /app /var/lib/webhooker
 
-USER webhooker
+# No USER: the entrypoint starts as root to make the data directory
+# webhooker's, then runs the app as webhooker.
 
 EXPOSE 8080
 
@@ -124,4 +130,5 @@ ENV BIND_ADDRESS=0.0.0.0
 HEALTHCHECK --interval=30s --timeout=3s --start-period=5s --retries=3 \
     CMD wget --no-verbose --tries=1 --spider http://localhost:8080/.well-known/healthcheck || exit 1
 
+ENTRYPOINT ["/usr/local/bin/docker-entrypoint.sh"]
 CMD ["/app/webhooker"]

README.md

@@ -538,6 +538,12 @@ its Argon2id hash. There is no second account and no forgot-password
 flow, so the banner and the reset command below are the only two ways
 in.
 
+A start that finds no `webhooker.db` in `DATA_DIR` also logs
+`created a new, empty database` at `WARN`, with the file's path,
+shortly before the banner. On a deployment that has run before, that
+line means `DATA_DIR` was empty, most often because its volume is not
+mounted.
+
 #### Recovering a lost admin password
 
 `webhooker resetpw` sets an existing account's password from the
@@ -552,8 +558,9 @@ printf '%s' "$NEW_PASSWORD" | \
     DATA_DIR=/var/lib/webhooker webhooker resetpw admin
 ```
 
-In a container it is the same binary, which the image sets as `CMD`
+In a container it is the same binary. The image's `CMD` is
-rather than `ENTRYPOINT`, so the whole command has to be given:
+`/app/webhooker`, and a command given to `docker run` replaces all of
+it, so the whole command has to be given:
 
 ```bash
 docker run --rm -v webhooker-data:/var/lib/webhooker \
@@ -690,38 +697,22 @@ those three values rather than trusting the figure. Measured at 65s on
 Docker 29.7.2.) A container `unhealthy` with `connection refused` in
 its health log, or a published port that resets connections, is this.
 
-The container runs as a non-root user (`webhooker`, UID 1000), exposes
+The app runs as a non-root user (`webhooker`, UID 1000), exposes port
-port 8080, and includes a health check against
+8080, and includes a health check against `/.well-known/healthcheck`.
-`/.well-known/healthcheck`. The `/var/lib/webhooker` volume holds all
+The `/var/lib/webhooker` volume holds all SQLite databases: the main
-SQLite databases: the main application database (`webhooker.db`), the
+application database (`webhooker.db`), the per-webhook event databases
-per-webhook event databases (`events-{uuid}.db`), and any archive
+(`events-{uuid}.db`), and any archive databases written by `database`
-databases written by `database` targets (`archive-{uuid}.db`). Mount
+targets (`archive-{uuid}.db`). Mount this as a persistent volume to
-this as a persistent volume to preserve data across container
+preserve data across container restarts.
-restarts.
 
-**The bind-mounted directory must be owned by UID 1000, or the
+**The container sets its data directory's owner and mode itself
-container does not start.** Docker creates a `-v` source path that
+before the app starts**, so a host directory can be mounted as it is,
-does not exist yet as `root:root`, and the process runs as UID 1000,
+whoever owns it. The image's `ENTRYPOINT`,
-so it cannot take its `DATA_DIR` lock:
+`deploy/docker-entrypoint.sh`, starts as root, creates `DATA_DIR` if
-
+it is missing, gives the directory and anything in it that belongs to
-```
+another user to `webhooker`, sets the directory to `0750`, and only
-webhooker: locking data directory /var/lib/webhooker: open
+then runs the app as `webhooker`. Started with `--user`, it changes
-/var/lib/webhooker/webhooker.lock: permission denied
+nothing and runs the app as that user.
-```
-
-It exits non-zero at that point, before opening any database. Create
-the directory ahead of the first `docker run`:
-
-```bash
-mkdir -p /path/to/data
-chown 1000:1000 /path/to/data
-chmod 750 /path/to/data
-```
-
-The same `chown` is what a restore needs — see step 4 of
-[Restore](#restore). A **named volume** does not have this problem:
-Docker copies the image's ownership onto a volume it initializes, and
-the image creates `/var/lib/webhooker` owned by `webhooker`.
 
 **The file modes are not yours to set, and do not depend on the
 directory.** `webhooker.db` holds target configuration in plaintext —
@@ -729,13 +720,10 @@ bearer tokens, API keys, Slack webhook URLs — along with the session
 encryption key, so webhooker creates every SQLite file it owns `0600`:
 each database and both of its `-wal` and `-shm` sidecars, across all
 three tiers. Files an earlier build left `0644` are tightened when
-they are opened. A `DATA_DIR` webhooker creates itself is `0750`, but
+they are opened. The directory's `0750` is defence in depth — it stops
-a bind mount supplies its own directory and Docker's default for one
+other local users listing the directory and learning your webhook
-it creates is `0755`; the `0600` files hold there regardless. The
+UUIDs from the `events-{uuid}.db` filenames — not the barrier
-`chmod 750` above is defence in depth — it stops other local users
+protecting the credentials.
-listing the directory and learning your webhook UUIDs from the
-`events-{uuid}.db` filenames — not the barrier protecting the
-credentials.
 
 ### Running under upaas
 
@@ -751,17 +739,6 @@ repository's `Dockerfile` and runs it. The app needs:
   app name, port `8080`. Leave `PORT` unset: the image's health check
   probes `8080`.
 - **Volume:** one host directory mounted at `/var/lib/webhooker`.
-  upaas bind-mounts the host path it is given and does not create it,
-  and the container does not start unless UID 1000 owns it (see
-  [Running with Docker](#running-with-docker)). Create it before the
-  first deploy:
-
-    ```bash
-    mkdir -p /path/to/data
-    chown 1000:1000 /path/to/data
-    chmod 750 /path/to/data
-    ```
-
 - **Environment variables:**
     - `WEBHOOKER_ENVIRONMENT=prod`
     - `TRUSTED_PROXIES`: your reverse proxy's address on that Docker
@@ -1020,12 +997,12 @@ done
 `.backup` reads through the WAL and writes a single consistent file with
 no sidecars of its own, so the destination is complete as it stands.
 Two caveats. First, the runtime image is `alpine:3.21` with only
-`ca-certificates` added — the `sqlite3` CLI is **not** in it, so run
+`ca-certificates` and `su-exec` added — the `sqlite3` CLI is **not** in
-this on the host against the volume path, or from a throwaway container
+it, so run this on the host against the volume path, or from a
-that mounts the volume. Second, each file is captured at its own
+throwaway container that mounts the volume. Second, each file is
-instant, so a webhook created or an event delivered between two files
+captured at its own instant, so a webhook created or an event delivered
-being copied lands in one and not the other. If you need the whole set
+between two files being copied lands in one and not the other. If you
-coherent as of a single moment, stop the service.
+need the whole set coherent as of a single moment, stop the service.
 
 Note that `sqlite3 <db> .dump` is **not** one of these procedures: it is
 an export, it holds a read transaction open for as long as it runs, and
@@ -1079,21 +1056,11 @@ with any `-wal`/`-shm` beside it, or wait until there are none.
    archive not opened since a crash. A copy salvaged from a crashed
    instance has them for everything, and needs all of them.
 
-4. **Fix ownership.** The container runs as the non-root `webhooker`
+4. Start the service. The container gives the directory and the
-   user, UID 1000 / GID 1000. Restored files must be owned by (or
+   restored files to the `webhooker` user before the app starts,
-   writable by) that UID, and so must the directory itself — SQLite
+   whoever restored them (see
-   creates the `-wal` and `-shm` sidecars beside the database, so a
+   [Running with Docker](#running-with-docker)). `AutoMigrate` runs
-   writable file inside a directory it cannot write is not enough:
+   against each restored database as it is opened.
-
-    ```bash
-    chown -R 1000:1000 /path/to/data
-    ```
-
-    Restoring as `root` on the host and forgetting this step is the
-    usual way a restore fails.
-
-5. Start the service. `AutoMigrate` runs against each restored database
-   as it is opened.
 
 ### Upgrades
 
@@ -3071,7 +3038,11 @@ check, see [The login endpoint](#the-login-endpoint).
 - Prometheus metrics behind basic auth
 - Static assets embedded in binary (no filesystem access needed at
   runtime)
-- Container runs as non-root user (UID 1000)
+- The app runs as the non-root `webhooker` user (UID 1000) in the
+  container. The image sets no `USER`, so these run as root: the
+  `ENTRYPOINT` script, which sets the data directory's owner and mode
+  before the app starts; the image's health check; and `docker exec`,
+  unless given `--user`
 - GORM soft deletes on every entity that carries `BaseModel`, which is
   all of them but `Setting` (data preserved for audit)
 
@@ -3198,10 +3169,13 @@ version is fixed independently of the compiler's:
    `GO_LDFLAGS`, so neither can drop the `-X` that stamps the version.
    The version arrives as the `VERSION` build arg, since the context
    has no `.git` (see [Version stamping](#version-stamping)).
-3. **Runtime stage** (`alpine:3.21`) — copies the static binary,
+3. **Runtime stage** (`alpine:3.21`) — copies the static binary and
-   creates the `/var/lib/webhooker` directory for all SQLite databases,
+   `deploy/docker-entrypoint.sh`, creates the `/var/lib/webhooker`
-   runs as the non-root `webhooker` user (UID 1000), exposes port 8080,
+   directory for all SQLite databases, exposes port 8080, and includes
-   and includes a health check against `/.well-known/healthcheck`.
+   a health check against `/.well-known/healthcheck`. It sets no
+   `USER`: the `ENTRYPOINT` script starts as root, sets the data
+   directory's owner and mode, and runs the app as the non-root
+   `webhooker` user (UID 1000) through `su-exec`.
 
 The lint stage invokes `golangci-lint` directly rather than `make lint`:
 it is already the pinned linter image, and `make lint` builds
@@ -3280,3 +3254,5 @@ MIT
 ## Author
 
 [@sneak](https://sneak.berlin)
+
+

cmd/webhooker/main.go

@@ -16,7 +16,6 @@ import (
 	"sneak.berlin/go/webhooker/internal/handlers"
 	"sneak.berlin/go/webhooker/internal/healthcheck"
 	"sneak.berlin/go/webhooker/internal/logger"
-	"sneak.berlin/go/webhooker/internal/metrics"
 	"sneak.berlin/go/webhooker/internal/middleware"
 	"sneak.berlin/go/webhooker/internal/resetpw"
 	"sneak.berlin/go/webhooker/internal/server"
@@ -178,10 +177,6 @@ func newApp() *fx.App {
 			healthcheck.New,
 			session.New,
 			handlers.New,
-			// The registry /metrics serves, and the delivery
-			// collectors registered on it.
-			metrics.NewRegistry,
-			metrics.New,
 			middleware.New,
 			// The one SSRF guard both target-creation validation
 			// and the delivery dialer consult, so they cannot

deploy/docker-entrypoint.sh

@@ -0,0 +1,22 @@
+#!/bin/sh
+# deploy/docker-entrypoint.sh: the image's ENTRYPOINT. A bind-mounted
+# data directory keeps its owner from the host, often root, and the app
+# could not write to it. Started as root, this creates DATA_DIR if
+# needed, gives it and everything in it to webhooker, sets its mode, and
+# runs the command as webhooker, so the app never runs as root. Started
+# as another user, it only runs the command.
+set -eu
+
+main() {
+    if [ "$(id -u)" != 0 ]; then
+        exec "$@"
+    fi
+
+    dir="${DATA_DIR:-/var/lib/webhooker}"
+    mkdir -p "$dir"
+    find "$dir" ! -user webhooker -exec chown -h webhooker:webhooker {} +
+    chmod 750 "$dir"
+    exec su-exec webhooker "$@"
+}
+
+main "$@"

internal/database/bootstrap_banner_test.go

@@ -3,6 +3,8 @@ package database_test
 import (
 	"bytes"
 	"context"
+	"log/slog"
+	"path/filepath"
 	"strings"
 	"testing"
 
@@ -83,3 +85,37 @@ func TestFirstBoot_PrintsTheAdminPasswordAsABanner(t *testing.T) {
 		t, ok, "the printed password must open the seeded account",
 	)
 }
+
+// TestNewDatabase_IsLoggedWithItsPath is the log half of
+// https://git.eeqj.de/sneak/webhooker/issues/359. A DATA_DIR that is
+// unexpectedly empty boots exactly like a first start, so the start
+// that creates the database must say so, and where. Opening that
+// database again must not.
+func TestNewDatabase_IsLoggedWithItsPath(t *testing.T) {
+	t.Parallel()
+
+	dir := t.TempDir()
+
+	open := func() string {
+		var out bytes.Buffer
+
+		db, err := database.Open(dir, slog.New(slog.NewTextHandler(&out, nil)))
+		require.NoError(t, err)
+		require.NoError(t, db.Close())
+
+		return out.String()
+	}
+
+	const created = `level=WARN msg="created a new, empty database"`
+
+	first := open()
+	second := open()
+
+	assert.Contains(
+		t, first,
+		created+" path="+filepath.Join(dir, database.MainDBFileName),
+	)
+	assert.NotContains(
+		t, second, created, "an existing database is not new",
+	)
+}

internal/database/database.go

@@ -8,6 +8,7 @@ import (
 	"errors"
 	"fmt"
 	"io"
+	"io/fs"
 	"log/slog"
 	"os"
 	"path/filepath"
@@ -199,6 +200,12 @@ func (d *Database) connectTo(dataDir string) error {
 	// Construct the main application database path inside DATA_DIR.
 	dbPath := filepath.Join(dataDir, MainDBFileName)
 
+	// Checked before opening, which creates the file. A DATA_DIR that
+	// is unexpectedly empty -- its volume not mounted, say -- looks
+	// exactly like a first start, so a new database is a warning.
+	_, statErr := os.Stat(dbPath)
+	created := errors.Is(statErr, fs.ErrNotExist)
+
 	// Opened through OpenSQLite so this handle carries the same WAL
 	// journaling, busy timeout, immediate-transaction locking, and pool
 	// bounds as every other database file. See sqlite_open.go.
@@ -229,7 +236,12 @@ func (d *Database) connectTo(dataDir string) error {
 	}
 
 	d.db = db
-	d.log.Info("connected to database", "path", dbPath)
+
+	if created {
+		d.log.Warn("created a new, empty database", "path", dbPath)
+	} else {
+		d.log.Info("connected to database", "path", dbPath)
+	}
 
 	// Run migrations
 	return d.migrate()

internal/delivery/engine.go

@@ -148,7 +148,6 @@ type EngineParams struct {
 	DBManager *database.WebhookDBManager
 	Logger    *logger.Logger
 	SSRFGuard *Guard
-	Metrics   *metrics.Set
 }
 
 // Engine processes queued deliveries in the background
@@ -168,10 +167,10 @@ type Engine struct {
 	retryCh    chan Task
 	workers    int
 
-	// mtr is the delivery metric set. Production wires the one
+	// mtr is the delivery metric set. Production wires the
-	// registered on the registry /metrics serves; a test can
+	// process-wide one; a test can substitute a set registered on
-	// substitute a set registered on a registry it holds, so it can
+	// a private registry so its assertions are not disturbed by
-	// gather what its own deliveries recorded.
+	// deliveries other tests are making at the same time.
 	mtr *metrics.Set
 
 	// targets maps each target type to its implementation.
@@ -205,7 +204,7 @@ func New(
 		deliveryCh: make(chan Task, deliveryChannelSize),
 		retryCh:    make(chan Task, retryChannelSize),
 		workers:    defaultWorkers,
-		mtr:        params.Metrics,
+		mtr:        metrics.Default(),
 	}
 
 	e.initTargets(&http.Client{

internal/delivery/export_test.go

@@ -9,7 +9,6 @@ import (
 	"net/url"
 	"time"
 
-	"github.com/prometheus/client_golang/prometheus"
 	"go.uber.org/fx"
 	"gorm.io/gorm"
 	"sneak.berlin/go/webhooker/internal/database"
@@ -390,7 +389,7 @@ func NewTestEngine(
 		deliveryCh: make(chan Task, deliveryChannelSize),
 		retryCh:    make(chan Task, retryChannelSize),
 		workers:    workers,
-		mtr:        metrics.New(prometheus.NewRegistry()),
+		mtr:        metrics.Default(),
 	}
 	e.initTargets(client)
 
@@ -405,7 +404,7 @@ func NewTestEngineSmallRetry(
 	e := &Engine{
 		log:     log,
 		retryCh: make(chan Task, 1),
-		mtr:     metrics.New(prometheus.NewRegistry()),
+		mtr:     metrics.Default(),
 	}
 	e.initTargets(nil)
 
@@ -428,7 +427,7 @@ func NewTestEngineWithDB(
 		deliveryCh: make(chan Task, deliveryChannelSize),
 		retryCh:    make(chan Task, retryChannelSize),
 		workers:    workers,
-		mtr:        metrics.New(prometheus.NewRegistry()),
+		mtr:        metrics.Default(),
 	}
 	e.initTargets(client)
 
@@ -436,7 +435,8 @@ func NewTestEngineWithDB(
 }
 
 // ExportSetMetrics substitutes the engine's metric set, so a test can
-// assert on collectors registered on a registry it holds.
+// assert on collectors registered on a private registry instead of
+// the process-wide ones every other test is also moving.
 func (e *Engine) ExportSetMetrics(mtr *metrics.Set) {
 	e.mtr = mtr
 }

internal/delivery/metrics_test.go

@@ -35,8 +35,9 @@ const (
 )
 
 // mIsolate gives the setup's engine a metric set registered on a
-// registry this test holds, so its exact assertions can gather from
+// private registry. The process-wide collectors are moved by every
-// it.
+// other delivery test running in parallel, so exact assertions are
+// only possible against a registry this test owns.
 func mIsolate(
 	t *testing.T, s iSetup,
 ) *prometheus.Registry {

internal/handlers/handlers.go

@@ -12,7 +12,6 @@ import (
 	"net/http"
 	"sync/atomic"
 
-	"github.com/prometheus/client_golang/prometheus"
 	"go.uber.org/fx"
 	"sneak.berlin/go/webhooker/internal/database"
 	"sneak.berlin/go/webhooker/internal/delivery"
@@ -62,8 +61,6 @@ type HandlersParams struct {
 	Notifier     delivery.Notifier
 	Evictor      delivery.WebhookEvictor
 	SSRFGuard    *delivery.Guard
-	Metrics      *metrics.Set
-	Registry     *prometheus.Registry
 }
 
 // Handlers provides HTTP handler methods for all application
@@ -125,7 +122,7 @@ func New(
 	s.mw = params.Middleware
 	s.notifier = params.Notifier
 	s.evictor = params.Evictor
-	s.mtr = params.Metrics
+	s.mtr = metrics.Default()
 	s.ssrf = params.SSRFGuard
 
 	// Parse all page templates once at startup

internal/handlers/handlers_test.go

@@ -20,7 +20,6 @@ import (
 	"sneak.berlin/go/webhooker/internal/handlers"
 	"sneak.berlin/go/webhooker/internal/healthcheck"
 	"sneak.berlin/go/webhooker/internal/logger"
-	"sneak.berlin/go/webhooker/internal/metrics"
 	"sneak.berlin/go/webhooker/internal/middleware"
 	"sneak.berlin/go/webhooker/internal/session"
 )
@@ -110,8 +109,6 @@ func newTestApp(
 			func(r *recordingEvictor) delivery.WebhookEvictor {
 				return r
 			},
-			metrics.NewRegistry,
-			metrics.New,
 			middleware.New,
 			delivery.NewGuard,
 			handlers.New,

internal/handlers/metrics.go

@@ -1,20 +0,0 @@
-package handlers
-
-import (
-	"net/http"
-
-	"github.com/prometheus/client_golang/prometheus/promhttp"
-)
-
-// HandleMetrics returns the Prometheus scrape handler for the
-// registry every collector in this process registers on. It is what
-// promhttp.Handler builds for the global default registry, including
-// the promhttp_metric_handler_* series that count scrapes, pointed at
-// that registry instead.
-func (s *Handlers) HandleMetrics() http.HandlerFunc {
-	reg := s.params.Registry
-
-	return promhttp.InstrumentMetricHandler(
-		reg, promhttp.HandlerFor(reg, promhttp.HandlerOpts{}),
-	).ServeHTTP
-}

internal/metrics/metrics.go

@@ -3,17 +3,17 @@
 // deliveries are attempted, how they end, how long they take, how
 // deep the queues are, and how many circuit breakers are open.
 //
-// It also builds the registry the authenticated /metrics route
+// The inbound HTTP metrics come from the go-http-metrics recorder in
-// serves. These collectors, the inbound HTTP metrics recorded in
+// internal/middleware and land on prometheus.DefaultRegisterer. These
-// internal/middleware, and the Go runtime and process collectors all
+// collectors register there too, so both surfaces are gathered by the
-// register on that one registry, never on Prometheus's global default.
+// one promhttp handler mounted on the authenticated /metrics route.
 package metrics
 
 import (
+	"sync"
 	"time"
 
 	"github.com/prometheus/client_golang/prometheus"
-	"github.com/prometheus/client_golang/prometheus/collectors"
 	"github.com/prometheus/client_golang/prometheus/promauto"
 	"sneak.berlin/go/webhooker/internal/database"
 )
@@ -57,31 +57,25 @@ var knownTargetTypes = []database.TargetType{
 	database.TargetTypeSlack,
 }
 
-// NewRegistry returns the registry /metrics serves, carrying the Go
+// defaultSet is the process-wide metric set, registered on the same
-// runtime and process collectors that Prometheus's global default
+// registry the HTTP middleware and the /metrics handler already use.
-// registry carries, so the go_* and process_* series stay in the
+// It is built on first use rather than in an init so that a test
-// scrape.
+// binary that never touches metrics never registers them.
 //
-// A registry of its own, rather than the global default, is what lets
+//nolint:gochecknoglobals // one process-wide registration, by design
-// two dependency graphs in one process — two tests, say — each
+var defaultSet = sync.OnceValue(func() *Set {
-// register their collectors without the second registration
+	return New(prometheus.DefaultRegisterer)
-// panicking.
+})
-func NewRegistry() *prometheus.Registry {
-	reg := prometheus.NewRegistry()
-	reg.MustRegister(
-		collectors.NewGoCollector(),
-		collectors.NewProcessCollector(
-			collectors.ProcessCollectorOpts{},
-		),
-	)
 
-	return reg
+// Default returns the process-wide metric set.
+func Default() *Set {
+	return defaultSet()
 }
 
 // Set is one registered group of webhooker's delivery collectors.
-// Production builds one on the registry /metrics serves; tests build
+// Production uses the single Default set; tests build their own
-// their own against a private registry so assertions are not
+// against a private registry so assertions are not disturbed by
-// disturbed by deliveries other tests are making concurrently.
+// deliveries other tests are making concurrently.
 type Set struct {
 	eventsReceived      prometheus.Counter
 	deliveryAttempts    *prometheus.CounterVec
@@ -99,7 +93,7 @@ type Set struct {
 // New registers a full set of delivery collectors on reg and returns
 // it. It panics if reg already holds them, which is the intended
 // behaviour for a duplicate registration.
-func New(reg *prometheus.Registry) *Set {
+func New(reg prometheus.Registerer) *Set {
 	factory := promauto.With(reg)
 
 	s := &Set{

internal/middleware/export_test.go

@@ -10,7 +10,8 @@ import (
 
 // MetricsMiddlewareForTest builds the metrics recording middleware
 // against a caller-supplied recorder, so a test can gather from its
-// own Prometheus registry without building a whole Middleware.
+// own Prometheus registry rather than the process-wide default one
+// that Middleware.Metrics uses.
 func MetricsMiddlewareForTest(
 	rec httpmetrics.Recorder,
 ) func(http.Handler) http.Handler {

internal/middleware/metrics.go

@@ -7,6 +7,7 @@ import (
 
 	"github.com/go-chi/chi"
 	httpmetrics "github.com/slok/go-http-metrics/metrics"
+	prommetrics "github.com/slok/go-http-metrics/metrics/prometheus"
 	ghmm "github.com/slok/go-http-metrics/middleware"
 	"github.com/slok/go-http-metrics/middleware/std"
 )
@@ -151,14 +152,16 @@ func (r boundedLabelRecorder) AddInflightRequests(
 var _ httpmetrics.Recorder = boundedLabelRecorder{}
 
 // Metrics returns middleware that records Prometheus HTTP metrics on
-// the registry the /metrics route serves. Every call shares the one
+// the default registry, which is the one the /metrics route gathers.
-// recorder New built, so any number of routers can install it.
 func (s *Middleware) Metrics() func(http.Handler) http.Handler {
-	return metricsMiddleware(s.metricsRecorder)
+	return metricsMiddleware(
+		prommetrics.NewRecorder(prommetrics.Config{}),
+	)
 }
 
 // metricsMiddleware builds the recording middleware against a given
-// recorder, so tests can gather from a registry of their own.
+// recorder, so tests can gather from a registry of their own instead
+// of the process-wide default.
 func metricsMiddleware(
 	rec httpmetrics.Recorder,
 ) func(http.Handler) http.Handler {

internal/middleware/metrics_test.go

@@ -57,8 +57,9 @@ const (
 // Server.setupWebhookRoutes inside it. That ordering is the whole
 // defect, so a test that flattens it would prove nothing.
 //
-// The recorder writes to a registry of the test's own, so each test
+// The recorder writes to a registry of the test's own rather than the
-// observes only its own traffic.
+// process-wide default one, so each test observes only its own
+// traffic.
 func metricsTestRouter(
 	t *testing.T,
 	receiverLimit int,

internal/middleware/middleware.go

@@ -13,9 +13,6 @@ import (
 	"github.com/go-chi/chi"
 	"github.com/go-chi/chi/middleware"
 	"github.com/go-chi/cors"
-	"github.com/prometheus/client_golang/prometheus"
-	httpmetrics "github.com/slok/go-http-metrics/metrics"
-	prommetrics "github.com/slok/go-http-metrics/metrics/prometheus"
 	"go.uber.org/fx"
 	"sneak.berlin/go/webhooker/internal/config"
 	"sneak.berlin/go/webhooker/internal/globals"
@@ -151,11 +148,10 @@ const (
 type MiddlewareParams struct {
 	fx.In
 
-	Logger   *logger.Logger
+	Logger  *logger.Logger
-	Globals  *globals.Globals
+	Globals *globals.Globals
-	Config   *config.Config
+	Config  *config.Config
-	Session  *session.Session
+	Session *session.Session
-	Registry *prometheus.Registry
 }
 
 // Middleware provides HTTP middleware for logging, CORS, auth, and
@@ -165,12 +161,6 @@ type Middleware struct {
 	params  *MiddlewareParams
 	session *session.Session
 
-	// metricsRecorder records the inbound HTTP metrics on the
-	// registry /metrics serves. It is built once, in New, because
-	// building it registers its collectors, and a second
-	// registration on the same registry panics; see Metrics.
-	metricsRecorder httpmetrics.Recorder
-
 	// loginGuard counts failed credential verifications and bounds
 	// concurrent password hashing. It is built on first use so that
 	// every construction path gets one; see guard().
@@ -189,9 +179,6 @@ func New(
 	s.params = &params
 	s.log = params.Logger.Get()
 	s.session = params.Session
-	s.metricsRecorder = prommetrics.NewRecorder(
-		prommetrics.Config{Registry: params.Registry},
-	)
 
 	return s, nil
 }

internal/resetpw/resetpw_test.go

@@ -24,7 +24,6 @@ import (
 	"sneak.berlin/go/webhooker/internal/handlers"
 	"sneak.berlin/go/webhooker/internal/healthcheck"
 	"sneak.berlin/go/webhooker/internal/logger"
-	"sneak.berlin/go/webhooker/internal/metrics"
 	"sneak.berlin/go/webhooker/internal/middleware"
 	"sneak.berlin/go/webhooker/internal/resetpw"
 	"sneak.berlin/go/webhooker/internal/session"
@@ -164,8 +163,6 @@ func newServerApp(
 			session.New,
 			func() delivery.Notifier { return &noopNotifier{} },
 			func() delivery.WebhookEvictor { return &noopEvictor{} },
-			metrics.NewRegistry,
-			metrics.New,
 			middleware.New,
 			delivery.NewGuard,
 			handlers.New,

internal/server/routes.go

@@ -7,6 +7,7 @@ import (
 	sentryhttp "github.com/getsentry/sentry-go/http"
 	"github.com/go-chi/chi"
 	"github.com/go-chi/chi/middleware"
+	"github.com/prometheus/client_golang/prometheus/promhttp"
 	"sneak.berlin/go/webhooker/static"
 )
 
@@ -129,7 +130,12 @@ func (s *Server) setupRoutes() {
 	if s.params.Config.MetricsAuthEnabled() {
 		s.router.Group(func(r chi.Router) {
 			r.Use(s.mw.MetricsAuth())
-			r.Get("/metrics", s.h.HandleMetrics())
+			r.Get(
+				"/metrics",
+				http.HandlerFunc(
+					promhttp.Handler().ServeHTTP,
+				),
+			)
 		})
 	}
 

internal/server/routes_test.go

@@ -7,6 +7,7 @@ import (
 	"net/http/httptest"
 	"net/url"
 	"regexp"
+	"slices"
 	"strconv"
 	"strings"
 	"testing"
@@ -23,7 +24,6 @@ import (
 	"sneak.berlin/go/webhooker/internal/handlers"
 	"sneak.berlin/go/webhooker/internal/healthcheck"
 	"sneak.berlin/go/webhooker/internal/logger"
-	"sneak.berlin/go/webhooker/internal/metrics"
 	"sneak.berlin/go/webhooker/internal/middleware"
 	"sneak.berlin/go/webhooker/internal/server"
 	"sneak.berlin/go/webhooker/internal/session"
@@ -113,8 +113,6 @@ func newTestEnvWithConfig(
 			session.New,
 			func() delivery.Notifier { return &noopNotifier{} },
 			func() delivery.WebhookEvictor { return &noopEvictor{} },
-			metrics.NewRegistry,
-			metrics.New,
 			middleware.New,
 			delivery.NewGuard,
 			handlers.New,
@@ -223,9 +221,21 @@ func (e *testEnv) csrfFrom(
 	// out of the markup has to be unescaped before it is submitted.
 	token := html.UnescapeString(match[1])
 
-	combined := make([]*http.Cookie, 0, len(cookies))
+	// A cookie the page sets replaces the one of the same name, as in
-	combined = append(combined, cookies...)
+	// a browser. Sent both, the server would read the first, older one.
-	combined = append(combined, w.Result().Cookies()...)
+	set := w.Result().Cookies()
+	combined := make([]*http.Cookie, 0, len(cookies)+len(set))
+
+	for _, c := range cookies {
+		replaced := slices.ContainsFunc(set, func(n *http.Cookie) bool {
+			return n.Name == c.Name
+		})
+		if !replaced {
+			combined = append(combined, c)
+		}
+	}
+
+	combined = append(combined, set...)
 
 	return token, combined
 }
@@ -617,6 +627,59 @@ func TestPagesLogin_CorrectPasswordSurvivesASpentBudget(
 	)
 }
 
+// TestPagesLogin_CookiesFromAnEarlierDatabase is
+// https://git.eeqj.de/sneak/webhooker/issues/359. A new database
+// brings a new session key, and the operator's browser still holds
+// the session and CSRF cookies signed with the old one. Logging in
+// must work as from a fresh browser and leave cookies the new key
+// accepts.
+func TestPagesLogin_CookiesFromAnEarlierDatabase(t *testing.T) {
+	t.Parallel()
+
+	const (
+		username = "operator"
+		password = "correct-horse-battery-staple"
+	)
+
+	earlier := newTestEnv(t)
+	earlierID, _ := earlier.seedUser(t, username, password)
+	_, stale := earlier.csrfFrom(t, "/pages/login", nil)
+	stale = append(stale, earlier.authCookies(t, earlierID, username)...)
+
+	env := newTestEnv(t)
+	env.seedUser(t, username, password)
+
+	token, cookies := env.csrfFrom(t, "/pages/login", stale)
+
+	form := url.Values{}
+	form.Set("csrf_token", token)
+	form.Set("username", username)
+	form.Set("password", password)
+
+	w := env.post("/pages/login", form, cookies)
+	require.Equal(
+		t, http.StatusSeeOther, w.Code,
+		"a session cookie from another key must not fail the login",
+	)
+
+	// The response deletes the old session cookie and then sets the
+	// new one; a browser keeps the last.
+	var fresh *http.Cookie
+
+	for _, c := range w.Result().Cookies() {
+		if c.Name == session.SessionName {
+			fresh = c
+		}
+	}
+
+	require.NotNil(t, fresh, "login must set a session cookie")
+	assert.Equal(
+		t, "/sources",
+		env.get("/", []*http.Cookie{fresh}).Header().Get("Location"),
+		"the new session cookie must authenticate",
+	)
+}
+
 // --- /user/{username} group ---
 
 // TestPasswordChange_OversizeBody_RejectedAndPasswordUnchanged
@@ -964,43 +1027,3 @@ func TestMetricsRouteUnmountedOnHalfSetConfig(t *testing.T) {
 		})
 	}
 }
-
-// TestTwoMetricsRoutersInOneProcess pins
-// https://git.eeqj.de/sneak/webhooker/issues/227: a second
-// metrics-enabled router in one process used to panic, because the
-// HTTP metrics registered on Prometheus's global default registry.
-// Two routers are built over separate dependency graphs and a third
-// over the first graph again, and each must still serve the HTTP,
-// delivery and Go runtime series.
-func TestTwoMetricsRoutersInOneProcess(t *testing.T) {
-	t.Parallel()
-
-	first := newTestEnvWithConfig(
-		t, metricsConfig(t, metricsUser, metricsAuthValue),
-	)
-	second := newTestEnvWithConfig(
-		t, metricsConfig(t, metricsUser, metricsAuthValue),
-	)
-	third := &testEnv{
-		router: server.NewRouterForTest(
-			first.log.Get(), first.cfg, first.mw, first.hnd,
-		),
-	}
-
-	for _, env := range []*testEnv{first, second, third} {
-		env.get("/", nil)
-
-		scrape := env.metricsRequest(metricsUser, metricsAuthValue)
-		require.Equal(t, http.StatusOK, scrape.Code)
-
-		for _, series := range []string{
-			"http_request_duration_seconds",
-			"http_response_size_bytes",
-			"http_requests_inflight",
-			"webhooker_events_received_total",
-			"go_goroutines",
-		} {
-			assert.Contains(t, scrape.Body.String(), series)
-		}
-	}
-}

internal/session/codec_test.go

@@ -19,8 +19,8 @@ import (
 )
 
 // The tests below exercise the securecookie codecs underneath the
-// store and nothing else: Session.Get only decodes, so no server-side
+// store and nothing else: they decode through the store itself, so no
-// expiry check takes part in the result. They exist because
+// server-side expiry check takes part in the result. They exist because
 // NewCookieStore gives its codecs a 30-day max age that assigning
 // store.Options does not override, which would let the codec accept a
 // cookie weeks past the cap the cookie attribute advertises.
@@ -75,10 +75,11 @@ func restamp(
 	return base64.URLEncoding.EncodeToString(payload)
 }
 
-// decodeCookie feeds value back through the store's decode path.
+// decodeCookie feeds value back through the store's decode path. It
+// asks the store rather than Session.Get, which treats a cookie that
+// does not decode as absent and so hides the codec's reason.
 func decodeCookie(
 	t *testing.T,
-	s *session.Session,
 	value string,
 ) (*sessions.Session, error) {
 	t.Helper()
@@ -94,7 +95,7 @@ func decodeCookie(
 		SameSite: http.SameSiteLaxMode,
 	})
 
-	sess, err := s.Get(req)
+	sess, err := session.NewStore(testKey()).Get(req, session.SessionName)
 	require.NotNil(t, sess)
 
 	return sess, err
@@ -105,7 +106,7 @@ func TestCodec_AcceptsCookieInsideAbsoluteCap(t *testing.T) {
 
 	s := testSession(t)
 
-	sess, err := decodeCookie(t, s, restamp(
+	sess, err := decodeCookie(t, restamp(
 		t,
 		issuedCookie(t, s),
 		time.Now().Add(-(testAbsoluteMaxAge-time.Hour)),
@@ -126,7 +127,7 @@ func TestCodec_RejectsCookiePastAbsoluteCap(t *testing.T) {
 
 	s := testSession(t)
 
-	sess, err := decodeCookie(t, s, restamp(
+	sess, err := decodeCookie(t, restamp(
 		t,
 		issuedCookie(t, s),
 		time.Now().Add(-(testAbsoluteMaxAge+time.Hour)),

internal/session/session.go

@@ -224,10 +224,22 @@ func New(
 }
 
 // Get retrieves a session for the request.
+//
+// A session cookie that does not decode -- one signed with an earlier
+// session key, say, because the database was made anew -- is treated
+// as absent: the caller gets a new, empty session and no error, and
+// the next save replaces the cookie.
 func (s *Session) Get(
 	r *http.Request,
 ) (*sessions.Session, error) {
-	return s.store.Get(r, SessionName)
+	sess, err := s.store.Get(r, SessionName)
+	if sess == nil {
+		return nil, err
+	}
+
+	// For a cookie that does not decode, gorilla/sessions returns a
+	// new, empty session alongside the error that is dropped here.
+	return sess, nil
 }
 
 // GetKey returns the raw 32-byte authentication key used for