Deploy main eacf1e7 to prod

Replaces #559, which conflicted.

Model: opus-5-5

internal/database/bootstrap_banner_test.go

@@ -4,6 +4,7 @@ import (
 	"bytes"
 	"context"
 	"log/slog"
+	"os"
 	"path/filepath"
 	"strings"
 	"testing"
@@ -119,3 +120,26 @@ func TestNewDatabase_IsLoggedWithItsPath(t *testing.T) {
 		t, second, created, "an existing database is not new",
 	)
 }
+
+// TestZeroLengthDatabase_IsLoggedAsNew covers what
+// https://git.eeqj.de/sneak/webhooker/issues/290 found: SQLite opens a
+// zero-length file as an empty database, so a start on one is a first
+// start, and it must say so exactly as a start with no file does.
+func TestZeroLengthDatabase_IsLoggedAsNew(t *testing.T) {
+	t.Parallel()
+
+	dir := t.TempDir()
+	path := filepath.Join(dir, database.MainDBFileName)
+	require.NoError(t, os.WriteFile(path, nil, database.SQLiteFilePerm))
+
+	var out bytes.Buffer
+
+	db, err := database.Open(dir, slog.New(slog.NewTextHandler(&out, nil)))
+	require.NoError(t, err)
+	require.NoError(t, db.Close())
+
+	assert.Contains(
+		t, out.String(),
+		`level=WARN msg="created a new, empty database" path=`+path,
+	)
+}

internal/database/database.go

@@ -8,7 +8,6 @@ import (
 	"errors"
 	"fmt"
 	"io"
-	"io/fs"
 	"log/slog"
 	"os"
 	"path/filepath"
@@ -19,6 +18,7 @@ import (
 	"sneak.berlin/go/webhooker/internal/banner"
 	"sneak.berlin/go/webhooker/internal/config"
 	"sneak.berlin/go/webhooker/internal/datadir"
+	"sneak.berlin/go/webhooker/internal/db/migrations"
 	"sneak.berlin/go/webhooker/internal/gormlog"
 	"sneak.berlin/go/webhooker/internal/logger"
 )
@@ -203,8 +203,7 @@ func (d *Database) connectTo(dataDir string) error {
 	// 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)
+	created := missingOrEmpty(dbPath)
 
 	// Opened through OpenSQLite so this handle carries the same WAL
 	// journaling, busy timeout, immediate-transaction locking, and pool
@@ -213,13 +212,15 @@ func (d *Database) connectTo(dataDir string) error {
 	if err != nil {
 		d.log.Error(
 			"failed to open database",
+			"path", dbPath,
 			"error", err,
 		)
 
 		return err
 	}
 
-	// Then use it with GORM
+	// Then use it with GORM. Its errors are SQLite's alone and name no
+	// file, so the path is added to them here.
 	db, err := gorm.Open(sqlite.Dialector{
 		Conn: sqlDB,
 	}, &gorm.Config{
@@ -229,10 +230,11 @@ func (d *Database) connectTo(dataDir string) error {
 	if err != nil {
 		d.log.Error(
 			"failed to connect to database",
+			"path", dbPath,
 			"error", err,
 		)
 
-		return err
+		return fmt.Errorf("connecting to %s: %w", dbPath, err)
 	}
 
 	d.db = db
@@ -243,13 +245,26 @@ func (d *Database) connectTo(dataDir string) error {
 		d.log.Info("connected to database", "path", dbPath)
 	}
 
-	// Run migrations
-	return d.migrate()
+	err = d.migrate()
+	if err != nil {
+		return fmt.Errorf("migrating %s: %w", dbPath, err)
+	}
+
+	err = parseModels(
+		db, &Setting{}, &User{}, &APIKey{},
+		&Webhook{}, &Entrypoint{}, &Target{},
+	)
+	if err != nil {
+		return fmt.Errorf("connecting to %s: %w", dbPath, err)
+	}
+
+	return nil
 }
 
+// migrate applies the main database's SQL files. See
+// internal/db/migrations.
 func (d *Database) migrate() error {
-	// Run GORM auto-migrations
-	err := d.Migrate()
+	err := migrations.Apply(d.db, migrations.Main)
 	if err != nil {
 		d.log.Error(
 			"failed to run database migrations",
@@ -264,6 +279,21 @@ func (d *Database) migrate() error {
 	return nil
 }
 
+// parseModels has GORM parse each model for db before the handle is
+// shared. GORM parses a model the first time a handle uses it, and two
+// goroutines doing that at once for models that refer to each other,
+// as Webhook and Target do, race inside GORM.
+func parseModels(db *gorm.DB, models ...any) error {
+	for _, model := range models {
+		err := (&gorm.Statement{DB: db}).Parse(model)
+		if err != nil {
+			return fmt.Errorf("parsing model %T: %w", model, err)
+		}
+	}
+
+	return nil
+}
+
 // ensureAdminUser creates the bootstrap admin account when the
 // deployment has no users at all.
 func (d *Database) ensureAdminUser() error {

internal/database/database_test.go

@@ -1,9 +1,15 @@
 package database_test
 
 import (
+	"bytes"
 	"context"
+	"log/slog"
+	"os"
+	"path/filepath"
 	"testing"
 
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
 	"go.uber.org/fx/fxtest"
 	"sneak.berlin/go/webhooker/internal/config"
 	"sneak.berlin/go/webhooker/internal/database"
@@ -100,3 +106,22 @@ func TestDatabaseConnection(t *testing.T) {
 		)
 	}
 }
+
+// TestOpen_UnreadableDatabaseIsNamed pins
+// https://git.eeqj.de/sneak/webhooker/issues/459: when SQLite cannot
+// read webhooker.db, the error that stops the server and `webhooker
+// resetpw` names the file, not only SQLite's own message.
+func TestOpen_UnreadableDatabaseIsNamed(t *testing.T) {
+	t.Parallel()
+
+	dir := t.TempDir()
+	path := filepath.Join(dir, database.MainDBFileName)
+	require.NoError(t, os.WriteFile(
+		path, bytes.Repeat([]byte("junk"), 1024), database.SQLiteFilePerm,
+	))
+
+	_, err := database.Open(dir, slog.New(slog.DiscardHandler))
+	require.Error(t, err)
+	assert.Contains(t, err.Error(), path)
+	assert.Contains(t, err.Error(), "file is not a database")
+}

internal/database/databasetest/databasetest.go

@@ -0,0 +1,55 @@
+// Package databasetest builds a WebhookDBManager for tests in other
+// packages.
+package databasetest
+
+import (
+	"log/slog"
+	"os"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+	"go.uber.org/fx/fxtest"
+	"sneak.berlin/go/webhooker/internal/config"
+	"sneak.berlin/go/webhooker/internal/database"
+)
+
+// NewWebhookDBManager creates a WebhookDBManager backed by the given
+// data directory, logging at DEBUG to standard error.
+func NewWebhookDBManager(
+	t *testing.T, dataDir string,
+) *database.WebhookDBManager {
+	t.Helper()
+
+	return NewWebhookDBManagerWithLogger(
+		t,
+		dataDir,
+		slog.New(slog.NewTextHandler(
+			os.Stderr,
+			&slog.HandlerOptions{Level: slog.LevelDebug},
+		)),
+	)
+}
+
+// NewWebhookDBManagerWithLogger is NewWebhookDBManager with the
+// logger supplied by the caller. The per-webhook databases this manager
+// opens hand that logger to gormlog, so a test that needs to see the SQL
+// the service emits can capture it.
+//
+// It is built through database.NewWebhookDBManager on a lifecycle that
+// is never started, so nothing closes its databases but the caller.
+func NewWebhookDBManagerWithLogger(
+	t *testing.T, dataDir string, log *slog.Logger,
+) *database.WebhookDBManager {
+	t.Helper()
+
+	mgr, err := database.NewWebhookDBManager(
+		fxtest.NewLifecycle(t),
+		database.WebhookDBManagerParams{
+			Config: &config.Config{DataDir: dataDir},
+			Logger: log,
+		},
+	)
+	require.NoError(t, err)
+
+	return mgr
+}

internal/database/event_db_isolation.go

@@ -2,7 +2,6 @@ package database
 
 import (
 	"fmt"
-	"log/slog"
 
 	"gorm.io/gorm"
 	"gorm.io/gorm/clause"
@@ -18,14 +17,14 @@ const omitAssociationsCallback = "webhooker:omit_associations"
 //
 // A per-webhook database holds the event tier only, but Delivery
 // declares belongs-to Event and Target and the delivery engine fills
-// both in memory before writing. Without this guard GORM upserts
-// those parent rows here on the delivery and retry write paths,
-// copying targets.config, which holds destination URLs and bearer
-// credentials, into the file most likely to be backed up or handed
-// to someone else. Registering the guard on the connection covers
-// every write path, including writes inside a transaction and write
-// paths added later. Every event-tier row this file holds is written
-// explicitly, so nothing depends on the automatic save.
+// both in memory before writing. Without this guard GORM upserts those
+// parent rows here on the delivery and retry write paths, and the
+// target, whose config holds destination URLs and bearer credentials,
+// goes to a targets table this file does not have, which fails the
+// write. Registering the guard on the connection
+// covers every write path, including writes inside a transaction and
+// write paths added later. Every event-tier row this file holds is
+// written explicitly, so nothing depends on the automatic save.
 func omitAssociations(db *gorm.DB) error {
 	omit := func(tx *gorm.DB) {
 		tx.Statement.Omits = append(
@@ -53,107 +52,3 @@ func omitAssociations(db *gorm.DB) error {
 
 	return nil
 }
-
-// eventDBSweptVersion is the PRAGMA user_version purgeTargetRows
-// stamps into a per-webhook database once it has removed any leaked
-// target rows *and* the VACUUM that removes their bytes has returned.
-// Nothing else in the tree uses user_version, so 0 means "not swept
-// by this build".
-//
-// The stamp, not the DELETE, is what records that a file is done. A
-// DELETE commits on its own, so a sweep that is interrupted or whose
-// VACUUM fails leaves a file whose rows are gone but whose credential
-// bytes are still in the free pages -- indistinguishable, by row
-// count, from a file that never leaked. Both leave the stamp unset,
-// so the next open sweeps again.
-const eventDBSweptVersion = 1
-
-// purgeTargetRows deletes target rows that an earlier build's
-// association upsert wrote into a per-webhook database, and rewrites
-// the file so their bytes are gone with them. AutoMigrate creates a
-// targets table in every one of these files because Delivery declares
-// a belongs-to Target, but nothing in the event tier may put rows in
-// it. The rows it did put there are junk, not history: they carry an
-// empty webhook_id, and delivery rows resolve their target against
-// the main database, so nothing here refers to them.
-//
-// The DELETE only unlinks the rows: modernc.org/sqlite leaves
-// secure_delete at SQLite's default of off, so the credential bytes
-// stay readable in the file's free pages and a backup of a swept file
-// would still hand them over. VACUUM rewrites the file without them.
-//
-// This runs before every migration and is gated on
-// eventDBSweptVersion, so a file pays for the rewrite once, on the
-// first open that finds it unstamped, and every open after that is a
-// PRAGMA read. A file this build created is stamped before its
-// targets table exists, so it never vacuums at all. A failure here
-// fails the open with the stamp left unset, so the sweep is retried
-// rather than skipped -- a webhook whose file cannot be swept stays
-// unusable instead of quietly serving from a file that still holds
-// recoverable credentials.
-func purgeTargetRows(
-	db *gorm.DB, log *slog.Logger, webhookID string,
-) error {
-	var version int
-
-	// Row().Scan, not (*gorm.DB).Scan: see internal/gormlog.
-	err := db.Raw("PRAGMA user_version").Row().Scan(&version)
-	if err != nil {
-		return fmt.Errorf(
-			"reading sweep marker of webhook database %s: %w",
-			webhookID, err,
-		)
-	}
-
-	if version >= eventDBSweptVersion {
-		return nil
-	}
-
-	var purged int64
-
-	if db.Migrator().HasTable("targets") {
-		res := db.Exec("DELETE FROM targets")
-		if res.Error != nil {
-			return fmt.Errorf(
-				"purging target rows from webhook database %s: %w",
-				webhookID, res.Error,
-			)
-		}
-
-		purged = res.RowsAffected
-
-		// Unconditional: a zero row count here does not mean there is
-		// nothing to remove, only that no *live* row is left. See
-		// eventDBSweptVersion.
-		err = db.Exec("VACUUM").Error
-		if err != nil {
-			return fmt.Errorf(
-				"purged %d leaked target rows from webhook database "+
-					"%s but vacuuming it failed, so the deleted "+
-					"target credentials are still recoverable from "+
-					"the file; it stays marked unswept and the next "+
-					"open retries: %w",
-				purged, webhookID, err,
-			)
-		}
-	}
-
-	err = db.Exec(fmt.Sprintf(
-		"PRAGMA user_version = %d", eventDBSweptVersion,
-	)).Error
-	if err != nil {
-		return fmt.Errorf(
-			"marking webhook database %s swept: %w", webhookID, err,
-		)
-	}
-
-	if purged > 0 {
-		log.Warn(
-			"purged leaked target rows from per-webhook database",
-			"webhook_id", webhookID,
-			"rows", purged,
-		)
-	}
-
-	return nil
-}

internal/database/event_db_isolation_test.go

@@ -1,391 +1,15 @@
 package database_test
 
 import (
-	"bytes"
-	"database/sql"
-	"fmt"
-	"os"
-	"path/filepath"
 	"testing"
 
 	"github.com/google/uuid"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
-	_ "modernc.org/sqlite"
 	"sneak.berlin/go/webhooker/internal/database"
+	"sneak.berlin/go/webhooker/internal/database/databasetest"
 )
 
-// testDataDirPerm is the mode the test data directory is created
-// with.
-const testDataDirPerm = 0o750
-
-// eventDBDataDir returns a data directory that a WebhookDBManager
-// can be pointed at.
-func eventDBDataDir(t *testing.T) string {
-	t.Helper()
-
-	dir := filepath.Join(t.TempDir(), "events")
-	require.NoError(t, os.MkdirAll(dir, testDataDirPerm))
-
-	return dir
-}
-
-// openRawEventDB opens the per-webhook database file directly,
-// without the manager, so a test can put a file on disk in a state
-// the manager has to cope with, or inspect one afterwards.
-func openRawEventDB(
-	t *testing.T, dataDir, webhookID string,
-) *sql.DB {
-	t.Helper()
-
-	path := filepath.Join(
-		dataDir, fmt.Sprintf("events-%s.db", webhookID),
-	)
-
-	sqlDB, err := sql.Open(
-		"sqlite",
-		fmt.Sprintf("file:%s?mode=rwc", path),
-	)
-	require.NoError(t, err)
-
-	t.Cleanup(func() { _ = sqlDB.Close() })
-
-	return sqlDB
-}
-
-// eventDBFileBytes reads a per-webhook database file off disk, so a
-// test can assert on what the file itself still holds rather than on
-// what a query returns.
-func eventDBFileBytes(t *testing.T, dataDir, webhookID string) []byte {
-	t.Helper()
-
-	//nolint:gosec // reads a file the test just created under t.TempDir()
-	raw, err := os.ReadFile(filepath.Join(
-		dataDir, fmt.Sprintf("events-%s.db", webhookID),
-	))
-	require.NoError(t, err)
-
-	return raw
-}
-
-// eventDBUserVersion returns the PRAGMA user_version of a per-webhook
-// database file, which is the marker purgeTargetRows stamps once it
-// has swept and vacuumed.
-func eventDBUserVersion(t *testing.T, sqlDB *sql.DB) int {
-	t.Helper()
-
-	var version int
-
-	require.NoError(t, sqlDB.QueryRowContext(
-		t.Context(), "PRAGMA user_version",
-	).Scan(&version))
-
-	return version
-}
-
-// clearEventDBSweptMarker resets the sweep marker to 0, which is what
-// a file written by a build without the sweep looks like. Tests that
-// seed a leaked row have to create the file through the manager to
-// get the real targets table shape, and that stamps it.
-func clearEventDBSweptMarker(t *testing.T, sqlDB *sql.DB) {
-	t.Helper()
-
-	_, err := sqlDB.ExecContext(t.Context(), "PRAGMA user_version = 0")
-	require.NoError(t, err)
-}
-
-// countTargetRows returns the number of rows in the targets table of
-// a per-webhook database file, or -1 if the table does not exist.
-func countTargetRows(t *testing.T, sqlDB *sql.DB) int {
-	t.Helper()
-
-	var tables int
-
-	require.NoError(t, sqlDB.QueryRowContext(
-		t.Context(),
-		"SELECT count(*) FROM sqlite_master "+
-			"WHERE type = 'table' AND name = 'targets'",
-	).Scan(&tables))
-
-	if tables == 0 {
-		return -1
-	}
-
-	var rows int
-
-	require.NoError(t, sqlDB.QueryRowContext(
-		t.Context(), "SELECT count(*) FROM targets",
-	).Scan(&rows))
-
-	return rows
-}
-
-// TestOpenPurgesLeakedTargetRows covers the sweep for event
-// databases written by a build that let GORM upsert target rows
-// into them: opening the database clears them, and opening it again
-// is a no-op.
-func TestOpenPurgesLeakedTargetRows(t *testing.T) {
-	t.Parallel()
-
-	dataDir := eventDBDataDir(t)
-	webhookID := uuid.New().String()
-
-	// Create the file the way the application does, so the targets
-	// table has exactly the shape AutoMigrate gives it, then write
-	// a leaked row into it the way the association upsert did.
-	initial := database.NewTestWebhookDBManager(dataDir)
-
-	_, err := initial.GetDB(webhookID)
-	require.NoError(t, err)
-	require.NoError(t, initial.CloseAll())
-
-	seed := openRawEventDB(t, dataDir, webhookID)
-
-	_, err = seed.ExecContext(
-		t.Context(),
-		"INSERT INTO targets "+
-			"(id, webhook_id, name, type, config) "+
-			"VALUES (?, '', ?, ?, ?)",
-		uuid.New().String(),
-		"leaked-target",
-		"slack",
-		`{"webhookUrl":"https://hooks.example/T000/B000/secret"}`,
-	)
-	require.NoError(t, err)
-	require.Equal(t, 1, countTargetRows(t, seed))
-	clearEventDBSweptMarker(t, seed)
-	require.NoError(t, seed.Close())
-
-	mgr := database.NewTestWebhookDBManager(dataDir)
-
-	_, err = mgr.GetDB(webhookID)
-	require.NoError(t, err)
-	require.NoError(t, mgr.CloseAll())
-
-	check := openRawEventDB(t, dataDir, webhookID)
-	assert.Zero(t, countTargetRows(t, check))
-	assert.Equal(
-		t, 1, eventDBUserVersion(t, check),
-		"a completed sweep must mark the file so later opens skip it",
-	)
-	require.NoError(t, check.Close())
-
-	// Idempotent: a second open leaves it at zero and does not
-	// error.
-	again := database.NewTestWebhookDBManager(dataDir)
-
-	_, err = again.GetDB(webhookID)
-	require.NoError(t, err)
-	require.NoError(t, again.CloseAll())
-
-	recheck := openRawEventDB(t, dataDir, webhookID)
-	assert.Zero(t, countTargetRows(t, recheck))
-}
-
-// TestOpenPurgeRemovesCredentialBytes covers the sweep at the level
-// that matters for a backup handed to someone else: the leaked
-// credential must be gone from the raw bytes of the file, not merely
-// unreachable by query. A bare DELETE unlinks the row and leaves the
-// bytes readable in the free pages, so this fails without the VACUUM
-// in purgeTargetRows.
-func TestOpenPurgeRemovesCredentialBytes(t *testing.T) {
-	t.Parallel()
-
-	dataDir := eventDBDataDir(t)
-	webhookID := uuid.New().String()
-	credential := "T00000000/B00000000/" + uuid.New().String()
-
-	initial := database.NewTestWebhookDBManager(dataDir)
-
-	_, err := initial.GetDB(webhookID)
-	require.NoError(t, err)
-	require.NoError(t, initial.CloseAll())
-
-	seed := openRawEventDB(t, dataDir, webhookID)
-
-	_, err = seed.ExecContext(
-		t.Context(),
-		"INSERT INTO targets "+
-			"(id, webhook_id, name, type, config) "+
-			"VALUES (?, '', ?, ?, ?)",
-		uuid.New().String(),
-		"leaked-target",
-		"slack",
-		fmt.Sprintf(
-			`{"webhookUrl":"https://hooks.example/%s"}`, credential,
-		),
-	)
-	require.NoError(t, err)
-	clearEventDBSweptMarker(t, seed)
-	require.NoError(t, seed.Close())
-
-	// The seed has to be in the file for its absence later to mean
-	// anything.
-	require.True(
-		t,
-		bytes.Contains(
-			eventDBFileBytes(t, dataDir, webhookID),
-			[]byte(credential),
-		),
-		"seeded credential is not in the file, so this test proves nothing",
-	)
-
-	mgr := database.NewTestWebhookDBManager(dataDir)
-
-	_, err = mgr.GetDB(webhookID)
-	require.NoError(t, err)
-	require.NoError(t, mgr.CloseAll())
-
-	assert.NotContains(
-		t,
-		string(eventDBFileBytes(t, dataDir, webhookID)),
-		credential,
-		"leaked credential is still recoverable from the raw file",
-	)
-}
-
-// TestOpenRevacuumsAfterIncompleteSweep covers the case a row count
-// cannot see: the rows are already deleted but the file was never
-// vacuumed, because an earlier sweep died between the two or its
-// VACUUM failed. The credential bytes are still recoverable, and the
-// unset marker is the only thing that says so, so the next open must
-// vacuum rather than conclude from the empty table that there is
-// nothing to do.
-func TestOpenRevacuumsAfterIncompleteSweep(t *testing.T) {
-	t.Parallel()
-
-	dataDir := eventDBDataDir(t)
-	webhookID := uuid.New().String()
-	credential := "T00000000/B00000000/" + uuid.New().String()
-
-	initial := database.NewTestWebhookDBManager(dataDir)
-
-	_, err := initial.GetDB(webhookID)
-	require.NoError(t, err)
-	require.NoError(t, initial.CloseAll())
-
-	seed := openRawEventDB(t, dataDir, webhookID)
-
-	_, err = seed.ExecContext(
-		t.Context(),
-		"INSERT INTO targets "+
-			"(id, webhook_id, name, type, config) "+
-			"VALUES (?, '', ?, ?, ?)",
-		uuid.New().String(),
-		"leaked-target",
-		"slack",
-		fmt.Sprintf(
-			`{"webhookUrl":"https://hooks.example/%s"}`, credential,
-		),
-	)
-	require.NoError(t, err)
-
-	// Exactly the state an interrupted sweep leaves: rows gone,
-	// marker unset, bytes still in the free pages.
-	_, err = seed.ExecContext(t.Context(), "DELETE FROM targets")
-	require.NoError(t, err)
-	require.Zero(t, countTargetRows(t, seed))
-	clearEventDBSweptMarker(t, seed)
-	require.NoError(t, seed.Close())
-
-	require.True(
-		t,
-		bytes.Contains(
-			eventDBFileBytes(t, dataDir, webhookID),
-			[]byte(credential),
-		),
-		"the deleted row's bytes must still be in the file, or this "+
-			"test proves nothing",
-	)
-
-	mgr := database.NewTestWebhookDBManager(dataDir)
-
-	_, err = mgr.GetDB(webhookID)
-	require.NoError(t, err)
-	require.NoError(t, mgr.CloseAll())
-
-	assert.NotContains(
-		t,
-		string(eventDBFileBytes(t, dataDir, webhookID)),
-		credential,
-		"an interrupted sweep was not retried, so the credential is "+
-			"still recoverable from the raw file",
-	)
-
-	check := openRawEventDB(t, dataDir, webhookID)
-	assert.Equal(t, 1, eventDBUserVersion(t, check))
-}
-
-// TestOpenSkipsSweptDatabase covers the other half of the marker: a
-// file this build created is marked without ever being vacuumed, and
-// a marked file is not swept again.
-func TestOpenSkipsSweptDatabase(t *testing.T) {
-	t.Parallel()
-
-	dataDir := eventDBDataDir(t)
-	webhookID := uuid.New().String()
-
-	mgr := database.NewTestWebhookDBManager(dataDir)
-
-	_, err := mgr.GetDB(webhookID)
-	require.NoError(t, err)
-	require.NoError(t, mgr.CloseAll())
-
-	marked := openRawEventDB(t, dataDir, webhookID)
-	assert.Equal(t, 1, eventDBUserVersion(t, marked))
-
-	// A marked file is left alone, so a row written into it survives
-	// a reopen. Nothing writes target rows any more; this stands in
-	// for the sweep having run.
-	_, err = marked.ExecContext(
-		t.Context(),
-		"INSERT INTO targets "+
-			"(id, webhook_id, name, type, config) "+
-			"VALUES (?, '', ?, ?, ?)",
-		uuid.New().String(), "sentinel", "slack", `{}`,
-	)
-	require.NoError(t, err)
-	require.NoError(t, marked.Close())
-
-	again := database.NewTestWebhookDBManager(dataDir)
-
-	_, err = again.GetDB(webhookID)
-	require.NoError(t, err)
-	require.NoError(t, again.CloseAll())
-
-	check := openRawEventDB(t, dataDir, webhookID)
-	assert.Equal(
-		t, 1, countTargetRows(t, check),
-		"a marked file must not be swept again",
-	)
-}
-
-// TestOpenSucceedsWithoutTargetsTable covers an existing event
-// database that never grew a targets table. The sweep must not fail
-// startup on it.
-func TestOpenSucceedsWithoutTargetsTable(t *testing.T) {
-	t.Parallel()
-
-	dataDir := eventDBDataDir(t)
-	webhookID := uuid.New().String()
-
-	seed := openRawEventDB(t, dataDir, webhookID)
-
-	_, err := seed.ExecContext(
-		t.Context(),
-		"CREATE TABLE events (id text PRIMARY KEY)",
-	)
-	require.NoError(t, err)
-	require.NoError(t, seed.Close())
-
-	mgr := database.NewTestWebhookDBManager(dataDir)
-
-	db, err := mgr.GetDB(webhookID)
-	require.NoError(t, err)
-	assert.NotNil(t, db)
-	require.NoError(t, mgr.CloseAll())
-}
-
 // TestEventDBCreateOmitsAssociations covers the connection-level
 // guard directly: a Delivery carrying its Event and Target in
 // memory, written through the manager's handle, must store only the
@@ -393,19 +17,19 @@ func TestOpenSucceedsWithoutTargetsTable(t *testing.T) {
 func TestEventDBCreateOmitsAssociations(t *testing.T) {
 	t.Parallel()
 
-	dataDir := eventDBDataDir(t)
 	webhookID := uuid.New().String()
 
-	mgr := database.NewTestWebhookDBManager(dataDir)
+	mgr := databasetest.NewWebhookDBManager(t, t.TempDir())
 
 	db, err := mgr.GetDB(webhookID)
 	require.NoError(t, err)
 
 	target := database.Target{
-		WebhookID: webhookID,
-		Name:      "leaky-target",
-		Type:      database.TargetTypeSlack,
-		Config:    `{"webhookUrl":"https://hooks.example/secret"}`,
+		WebhookID:   webhookID,
+		Name:        "leaky-target",
+		Type:        database.TargetTypeSlack,
+		URL:         "https://hooks.example/secret",
+		MaxAttempts: 1,
 	}
 	target.ID = uuid.New().String()
 
@@ -431,8 +55,9 @@ func TestEventDBCreateOmitsAssociations(t *testing.T) {
 	require.NoError(t, db.Model(d).
 		Update("status", database.DeliveryStatusDelivered).
 		Error)
-	require.NoError(t, mgr.CloseAll())
 
-	check := openRawEventDB(t, dataDir, webhookID)
-	assert.Zero(t, countTargetRows(t, check))
+	// The file has no targets table, so the create and update above
+	// would have failed had GORM tried to save the target.
+	assert.False(t, db.Migrator().HasTable(&database.Target{}))
+	require.NoError(t, mgr.CloseAll())
 }

internal/database/event_tier_indexes_test.go

@@ -13,10 +13,8 @@ import (
 	"sneak.berlin/go/webhooker/internal/database"
 )
 
-// TestWebhookDBManager_OpenAddsEventTierIndexes verifies that opening a
-// per-webhook database that predates these indexes creates them. It
-// stands in for an older database file by dropping the indexes
-// AutoMigrate just created, then reopening the same file.
+// TestWebhookDBManager_OpenAddsEventTierIndexes verifies that a new
+// per-webhook database has these indexes.
 func TestWebhookDBManager_OpenAddsEventTierIndexes(t *testing.T) {
 	t.Parallel()
 
@@ -42,27 +40,9 @@ func TestWebhookDBManager_OpenAddsEventTierIndexes(t *testing.T) {
 	db, err := mgr.GetDB(webhookID)
 	require.NoError(t, err)
 
-	// A fresh database has them.
-	for _, ix := range indexes {
-		require.True(t, db.Migrator().HasIndex(ix.model, ix.name))
-	}
-
-	// Stand in for a database file created before the indexes existed.
-	for _, ix := range indexes {
-		require.NoError(t, db.Migrator().DropIndex(ix.model, ix.name))
-		require.False(t, db.Migrator().HasIndex(ix.model, ix.name))
-	}
-
-	// Drop the cached connection so the next open reopens the file and
-	// runs AutoMigrate against it, as a restart would.
-	require.NoError(t, mgr.CloseAll())
-
-	db, err = mgr.GetDB(webhookID)
-	require.NoError(t, err)
-
 	for _, ix := range indexes {
 		assert.True(t, db.Migrator().HasIndex(ix.model, ix.name),
-			"opening the existing database should create %s", ix.name)
+			"a new database should have %s", ix.name)
 	}
 }
 
@@ -93,14 +73,15 @@ func TestEventTierQueriesUseTheirIndexes(t *testing.T) {
 		deliveries []database.Delivery
 		results    []database.DeliveryResult
 		depths     []struct{ Depth int }
+		removed    []database.TargetTotals
 	)
 
 	byStatus := "idx_deliveries_status (status=? AND deleted_at=?)"
 	byEvent := "idx_deliveries_event_id (event_id=? AND deleted_at=?)"
-	byAge := "idx_events_deleted_at_created_at (deleted_at=? AND created_at<?)"
 
-	// The delivery engine: recovery and the retry sweep, the sweep for
-	// stranded pending deliveries, and the queue depth count.
+	// The delivery engine: the recovery routine, which reads the
+	// retrying deliveries and, on a tick of the sweep, the pending ones
+	// untouched for a while; and the queue depth count.
 	assertPlanUses(t, db, dry.Where(
 		"status = ?", database.DeliveryStatusRetrying,
 	).Find(&deliveries), byStatus)
@@ -123,25 +104,207 @@ func TestEventTierQueriesUseTheirIndexes(t *testing.T) {
 		Order("attempt_num ASC").Find(&results),
 		"idx_delivery_results_delivery_id (delivery_id=? AND deleted_at=?)")
 
-	// Retention's three deletes (reapExpired), whose subqueries are built
-	// afresh for each statement as it builds them.
-	expiredEventIDs := func() *gorm.DB {
-		return dry.Model(&database.Event{}).Select("id").
-			Where("created_at < ?", cutoff)
+	// Retention (reapExpired, deleteEvents): one batch of expired
+	// events, then their attempts, deliveries and the events.
+	var expired []string
+
+	assertPlanUses(t, db, dry.Unscoped().Model(&database.Event{}).
+		Where("created_at < ?", cutoff).
+		Limit(database.ExportReapBatchSize).Pluck("id", &expired),
+		"idx_events_created_at (created_at<?)")
+	assertPlanUses(t, db, dry.Unscoped().Where(
+		"delivery_id IN (?)", dry.Unscoped().Model(&database.Delivery{}).
+			Select("id").Where("event_id IN ?", ids),
+	).Delete(&database.DeliveryResult{}),
+		"idx_delivery_results_delivery_id (delivery_id=?)",
+		"idx_deliveries_event_id (event_id=?)")
+	assertPlanUses(t, db, dry.Unscoped().Model(&database.Delivery{}).
+		Select("target_id, count(*) AS deliveries_removed, "+
+			"count(CASE WHEN status = ? THEN 1 END) AS failed_removed",
+			database.DeliveryStatusFailed).
+		Where("event_id IN ?", ids).Group("target_id").Find(&removed),
+		"idx_deliveries_event_id (event_id=?)")
+	assertPlanUses(t, db, dry.Unscoped().Where("event_id IN ?", ids).
+		Delete(&database.Delivery{}), "idx_deliveries_event_id (event_id=?)")
+	assertPlanUses(t, db, dry.Unscoped().Where("id IN ?", ids).
+		Delete(&database.Event{}), "sqlite_autoindex_events_1 (id=?)")
+}
+
+// TestEventLogFiltersUseTheStatusIndex does the same for the event log's
+// Failed and Pending lists, of the newest events with a delivery in
+// given statuses, and for their counts (eventsWithStatus and
+// countEventsWithStatus in the handlers). The lists must also reach
+// the events table only by ID: from the matching deliveries, then from
+// the newest of those events.
+func TestEventLogFiltersUseTheStatusIndex(t *testing.T) {
+	t.Parallel()
+
+	mgr, lc := setupTestWebhookDBManager(t)
+	ctx := context.Background()
+	require.NoError(t, lc.Start(ctx))
+
+	defer func() { require.NoError(t, lc.Stop(ctx)) }()
+
+	webhookID := uuid.New().String()
+
+	db, err := mgr.GetDB(webhookID)
+	require.NoError(t, err)
+
+	dry := db.Session(&gorm.Session{DryRun: true})
+	byStatus := "idx_deliveries_status (status=? AND deleted_at=?)"
+	pending := []database.DeliveryStatus{
+		database.DeliveryStatusPending,
+		database.DeliveryStatusRetrying,
 	}
 
-	assertPlanUses(t, db, dry.Unscoped().Where(
-		"delivery_id IN (?)", dry.Model(&database.Delivery{}).
-			Select("id").Where("event_id IN (?)", expiredEventIDs()),
-	).Delete(&database.DeliveryResult{}),
-		"idx_delivery_results_delivery_id (delivery_id=?)", byEvent, byAge)
-	assertPlanUses(t, db, dry.Unscoped().Where(
-		"event_id IN (?)", expiredEventIDs(),
-	).Delete(&database.Delivery{}),
-		"idx_deliveries_event_id (event_id=?)", byAge)
-	assertPlanUses(t, db, dry.Unscoped().Where(
-		"created_at < ?", cutoff,
-	).Delete(&database.Event{}), "idx_events_created_at (created_at<?)")
+	var (
+		rows  []struct{ ID string }
+		count int64
+	)
+
+	matching := dry.Model(&database.Delivery{}).
+		Distinct("event_id").Where("status IN ?", pending)
+	newest := dry.Table("(?) AS matching", matching).
+		Joins("CROSS JOIN events ON events.id = matching.event_id").
+		Where(
+			"events.webhook_id = ? AND events.deleted_at IS NULL",
+			webhookID,
+		).
+		Order("events.created_at DESC").Limit(50).
+		Select("events.id AS event_id")
+
+	// Each step of the plan is printed in braces, so these name the
+	// lookup that follows each scan.
+	byID := "{SEARCH events USING INDEX sqlite_autoindex_events_1 (id=?)}"
+
+	assertPlanUses(t, db, dry.Table("(?) AS newest", newest).
+		Joins("CROSS JOIN events ON events.id = newest.event_id").
+		Select("id").Order("created_at DESC").Limit(50).Find(&rows),
+		byStatus, "{SCAN matching} "+byID, "{SCAN newest} "+byID)
+	assertPlanUses(t, db, dry.Model(&database.Delivery{}).
+		Distinct("event_id").Where("status IN ?", pending).Count(&count),
+		byStatus)
+}
+
+// TestStatisticsQueriesUseTheirIndexes does the same for the webhook
+// page's statistics (readEventStats in the handlers): deliveries in
+// progress, each target's deliveries finished since a time, which must
+// come from the index alone, and events received since a time.
+func TestStatisticsQueriesUseTheirIndexes(t *testing.T) {
+	t.Parallel()
+
+	mgr, lc := setupTestWebhookDBManager(t)
+	ctx := context.Background()
+	require.NoError(t, lc.Start(ctx))
+
+	defer func() { require.NoError(t, lc.Stop(ctx)) }()
+
+	db, err := mgr.GetDB(uuid.New().String())
+	require.NoError(t, err)
+
+	dry := db.Session(&gorm.Session{DryRun: true})
+	since := time.Now()
+
+	var (
+		count    int64
+		byTarget []struct{ TargetID string }
+	)
+
+	assertPlanUses(t, db, dry.Model(&database.Delivery{}).
+		Where("status IN ?", []database.DeliveryStatus{
+			database.DeliveryStatusPending,
+			database.DeliveryStatusRetrying,
+		}).Count(&count),
+		"idx_deliveries_status (status=? AND deleted_at=?)")
+	assertPlanUses(t, db, dry.Model(&database.Delivery{}).
+		Select("target_id, "+
+			"count(CASE WHEN status = ? THEN 1 END) AS delivered, "+
+			"count(CASE WHEN status = ? THEN 1 END) AS failed",
+			database.DeliveryStatusDelivered,
+			database.DeliveryStatusFailed).
+		Where("status IN ? AND finished_at >= ?",
+			[]database.DeliveryStatus{
+				database.DeliveryStatusDelivered,
+				database.DeliveryStatusFailed,
+			}, since).
+		Group("target_id").Find(&byTarget),
+		"COVERING INDEX idx_deliveries_status "+
+			"(status=? AND deleted_at=? AND finished_at>?)")
+	assertPlanUses(t, db, dry.Model(&database.Event{}).
+		Where("created_at >= ?", since).Count(&count),
+		"idx_events_deleted_at_created_at "+
+			"(deleted_at=? AND created_at>?)")
+}
+
+// TestResubmitCountUsesItsIndex does the same for the event log's count
+// of the events resubmitted from each of a page's events (resubmitCounts
+// in the handlers). It passes a full page of 25 ids: with an index on
+// resubmitted_from_id alone, SQLite uses it for three ids and turns to
+// the deleted_at index from five.
+func TestResubmitCountUsesItsIndex(t *testing.T) {
+	t.Parallel()
+
+	mgr, lc := setupTestWebhookDBManager(t)
+	ctx := context.Background()
+	require.NoError(t, lc.Start(ctx))
+
+	defer func() { require.NoError(t, lc.Stop(ctx)) }()
+
+	db, err := mgr.GetDB(uuid.New().String())
+	require.NoError(t, err)
+
+	dry := db.Session(&gorm.Session{DryRun: true})
+
+	page := make([]string, 25)
+	for i := range page {
+		page[i] = uuid.New().String()
+	}
+
+	var counts []struct{ Total int }
+
+	assertPlanUses(t, db, dry.Model(&database.Event{}).
+		Select("resubmitted_from_id, count(*) AS total").
+		Where("resubmitted_from_id IN ?", page).
+		Group("resubmitted_from_id").Find(&counts),
+		"idx_events_resubmitted_from_id "+
+			"(resubmitted_from_id=? AND deleted_at=?)")
+}
+
+// TestEntrypointEventsUseTheirIndex does the same for the webhook
+// page's count, for each entrypoint, of the events that arrived on its
+// URL since the retention cutoff (addEntrypointEvents in the
+// handlers), which must come from the index alone. It passes 25
+// entrypoints, as TestResubmitCountUsesItsIndex passes 25 events.
+func TestEntrypointEventsUseTheirIndex(t *testing.T) {
+	t.Parallel()
+
+	mgr, lc := setupTestWebhookDBManager(t)
+	ctx := context.Background()
+	require.NoError(t, lc.Start(ctx))
+
+	defer func() { require.NoError(t, lc.Stop(ctx)) }()
+
+	db, err := mgr.GetDB(uuid.New().String())
+	require.NoError(t, err)
+
+	dry := db.Session(&gorm.Session{DryRun: true})
+
+	entrypoints := make([]string, 25)
+	for i := range entrypoints {
+		entrypoints[i] = uuid.New().String()
+	}
+
+	var rows []struct{ Events int }
+
+	assertPlanUses(t, db, dry.Model(&database.Event{}).
+		Select("entrypoint_id, count(*) AS events").
+		Where("entrypoint_id IN ? AND resubmitted_from_id IS NULL",
+			entrypoints).
+		Where("created_at >= ?", time.Now()).
+		Group("entrypoint_id").Find(&rows),
+		"COVERING INDEX idx_events_entrypoint_id "+
+			"(entrypoint_id=? AND deleted_at=? AND "+
+			"resubmitted_from_id=? AND created_at>?)")
 }
 
 // assertPlanUses asserts that SQLite's plan for a statement GORM built

internal/database/export_test.go

@@ -5,6 +5,7 @@ import (
 	"io"
 	"log/slog"
 	"os"
+	"testing"
 	"time"
 
 	"go.uber.org/fx"
@@ -28,6 +29,10 @@ func NewTestRetentionReaper(
 	}
 }
 
+// ExportReapBatchSize exposes how many expired events one retention
+// transaction deletes.
+const ExportReapBatchSize = reapBatchSize
+
 // ExportSweep runs a single retention sweep synchronously for tests.
 func (r *RetentionReaper) ExportSweep(ctx context.Context) {
 	r.sweep(ctx)
@@ -79,3 +84,14 @@ func (d *Database) ExportSetBannerOut(w io.Writer) {
 func DummyPasswordHashForTest() string {
 	return dummyPasswordHash()
 }
+
+// HashAtShippedCostForTest makes HashPassword hash at the shipped
+// memory cost until t ends. t must not run in parallel with other
+// tests, which would hash at that cost alongside it.
+func HashAtShippedCostForTest(t *testing.T) {
+	t.Helper()
+
+	hashAtShippedCostInTest = true
+
+	t.Cleanup(func() { hashAtShippedCostInTest = false })
+}

internal/database/model_apikey.go

@@ -15,6 +15,7 @@ type APIKey struct {
 	Description string     `json:"description"`
 	LastUsedAt  *time.Time `json:"lastUsedAt,omitempty"`
 
-	// Relations
-	User User `json:"user,omitzero"`
+	// Relations. No model marshals the record it belongs to:
+	// User.APIKeys leads back here, and the JSON could loop.
+	User User `json:"-"`
 }

internal/database/model_delivery.go

@@ -1,6 +1,10 @@
 package database
 
-import "gorm.io/gorm"
+import (
+	"time"
+
+	"gorm.io/gorm"
+)
 
 // DeliveryStatus represents the status of a delivery
 type DeliveryStatus string
@@ -37,7 +41,7 @@ type Delivery struct {
 	BaseModel
 
 	EventID  string         `gorm:"type:uuid;not null;index:idx_deliveries_event_id,priority:1"       json:"eventId"`
-	TargetID string         `gorm:"type:uuid;not null"                                                json:"targetId"`
+	TargetID string         `gorm:"type:uuid;not null;index:idx_deliveries_status,priority:4"         json:"targetId"`
 	Status   DeliveryStatus `gorm:"not null;default:'pending';index:idx_deliveries_status,priority:1" json:"status"`
 
 	// DeletedAt repeats the BaseModel field only to be the second column
@@ -45,8 +49,21 @@ type Delivery struct {
 	// gives.
 	DeletedAt gorm.DeletedAt `gorm:"index:idx_deliveries_event_id,priority:2;index:idx_deliveries_status,priority:2" json:"deletedAt,omitzero"`
 
-	// Relations
-	Event           Event            `json:"event,omitzero"`
-	Target          Target           `json:"target,omitzero"`
+	// FinishedAt is when the delivery became delivered or failed, and
+	// nil while it is pending or retrying. It and then TargetID end the
+	// status index, so the webhook page counts each target's deliveries
+	// that finished in a recent window by reading just that window from
+	// the index.
+	FinishedAt *time.Time `gorm:"index:idx_deliveries_status,priority:3" json:"finishedAt,omitempty"`
+
+	// Replay is set on a delivery created by the event log's Replay
+	// action, so the pages can tell it from the delivery it repeats.
+	Replay bool `gorm:"not null;default:false" json:"replay"`
+
+	// Relations. No model marshals the record it belongs to:
+	// Event.Deliveries and Target.Deliveries lead back here, and the
+	// JSON could loop.
+	Event           Event            `json:"-"`
+	Target          Target           `json:"-"`
 	DeliveryResults []DeliveryResult `json:"deliveryResults,omitempty"`
 }

internal/database/model_delivery_result.go

@@ -23,6 +23,7 @@ type DeliveryResult struct {
 	Error        string `json:"error,omitempty"`
 	Duration     int64  `json:"durationMs"` // Duration in milliseconds
 
-	// Relations
-	Delivery Delivery `json:"delivery,omitzero"`
+	// Relations. No model marshals the record it belongs to:
+	// Delivery.DeliveryResults leads back here, and the JSON could loop.
+	Delivery Delivery `json:"-"`
 }

internal/database/model_entrypoint.go

@@ -15,6 +15,7 @@ type Entrypoint struct {
 	Description string `json:"description"`
 	Active      bool   `gorm:"default:true" json:"active"`
 
-	// Relations
-	Webhook Webhook `json:"webhook,omitzero"`
+	// Relations. No model marshals the record it belongs to:
+	// Webhook.Entrypoints leads back here, and the JSON could loop.
+	Webhook Webhook `json:"-"`
 }

internal/database/model_event.go

@@ -19,28 +19,40 @@ type Event struct {
 	// narrows by a < only on the last column it uses. Its final delete
 	// has no deleted_at condition and uses the index on created_at
 	// alone. The other tables keep the unindexed BaseModel created_at.
-	CreatedAt time.Time      `gorm:"index;index:idx_events_deleted_at_created_at,priority:2" json:"createdAt"`
-	DeletedAt gorm.DeletedAt `gorm:"index:idx_events_deleted_at_created_at,priority:1"       json:"deletedAt,omitzero"`
+	// DeletedAt is also the second column of the resubmitted_from_id
+	// index, for the reason DeliveryResult gives. The entrypoint_id
+	// index, for the webhook page's entrypoint list, has it second too,
+	// resubmitted_from_id third, and created_at last, which the list
+	// compares with a range.
+	CreatedAt time.Time      `gorm:"index;index:idx_events_deleted_at_created_at,priority:2;index:idx_events_entrypoint_id,priority:4"                                           json:"createdAt"`
+	DeletedAt gorm.DeletedAt `gorm:"index:idx_events_deleted_at_created_at,priority:1;index:idx_events_resubmitted_from_id,priority:2;index:idx_events_entrypoint_id,priority:2" json:"deletedAt,omitzero"`
 
-	WebhookID    string `gorm:"type:uuid;not null" json:"webhookId"`
-	EntrypointID string `gorm:"type:uuid;not null" json:"entrypointId"`
+	WebhookID    string `gorm:"type:uuid;not null"                                           json:"webhookId"`
+	EntrypointID string `gorm:"type:uuid;not null;index:idx_events_entrypoint_id,priority:1" json:"entrypointId"`
 
-	// Request data
+	// Request data. RawQuery is the receiving request's query string
+	// as sent, without the leading "?".
 	Method      string `gorm:"not null"    json:"method"`
+	RawQuery    string `gorm:"type:text"   json:"rawQuery"`
 	Headers     string `gorm:"type:text"   json:"headers"` // JSON
 	Body        string `gorm:"type:text"   json:"body"`
 	ContentType string `json:"contentType"`
 
+	// BodyBytes is the size of Body in bytes, recorded when the event
+	// is stored, so that the recent events list, which reads only the
+	// start of each body, knows the whole body's size.
+	BodyBytes int64 `gorm:"not null" json:"bodyBytes"`
+
 	// ResubmittedFromID names the event this one was copied from by
 	// an operator resubmit. It is nil for an event that arrived on
-	// the receiver, which is every event created before the column
-	// existed. It is not a foreign key: the source event can be
+	// the receiver. It is not a foreign key: the source event can be
 	// reaped by retention while its copies remain, and the id is
 	// kept as the record of where the copy came from either way.
-	ResubmittedFromID *string `gorm:"type:uuid;index" json:"resubmittedFromId,omitempty"`
+	ResubmittedFromID *string `gorm:"type:uuid;index:idx_events_resubmitted_from_id,priority:1;index:idx_events_entrypoint_id,priority:3" json:"resubmittedFromId,omitempty"`
 
-	// Relations
-	Webhook    Webhook    `json:"webhook,omitzero"`
-	Entrypoint Entrypoint `json:"entrypoint,omitzero"`
+	// Relations. No model marshals the record it belongs to, so
+	// Webhook and Entrypoint are left out of the JSON.
+	Webhook    Webhook    `json:"-"`
+	Entrypoint Entrypoint `json:"-"`
 	Deliveries []Delivery `json:"deliveries,omitempty"`
 }

internal/database/model_parent_test.go

@@ -0,0 +1,127 @@
+package database_test
+
+import (
+	"testing"
+
+	"github.com/google/uuid"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/webhooker/internal/database"
+)
+
+// TestPreloadedModelsMarshalWithoutTheirParent pins that a child's
+// reference to the record it belongs to is left out of the JSON, so a
+// webhook and its targets cannot marshal each other in a loop, and that
+// GORM still preloads that reference, since it ignores json tags.
+func TestPreloadedModelsMarshalWithoutTheirParent(t *testing.T) {
+	t.Parallel()
+
+	db := startedTestDB(t)
+
+	stored := database.Webhook{
+		UserID:      uuid.New().String(),
+		Name:        testWebhookName,
+		Entrypoints: []database.Entrypoint{{Path: uuid.New().String()}},
+		Targets: []database.Target{{
+			Name:        "log",
+			Type:        database.TargetTypeLog,
+			MaxAttempts: 1,
+		}},
+	}
+	require.NoError(t, db.Create(&stored).Error)
+
+	entrypointID := stored.Entrypoints[0].ID
+	targetID := stored.Targets[0].ID
+
+	var webhook database.Webhook
+
+	require.NoError(t, db.
+		Preload("Entrypoints.Webhook").
+		Preload("Targets.Webhook").
+		First(&webhook, "id = ?", stored.ID).Error)
+
+	require.Len(t, webhook.Entrypoints, 1)
+	require.Len(t, webhook.Targets, 1)
+	assert.Equal(t, stored.ID, webhook.Entrypoints[0].Webhook.ID)
+	assert.Equal(t, stored.ID, webhook.Targets[0].Webhook.ID)
+
+	encoded := marshalModel(t, webhook)
+
+	assert.Contains(t, encoded, entrypointID)
+	assert.Contains(t, encoded, targetID)
+
+	// Each child holds the parent's id as its webhookId, so the parent
+	// is looked for by its own id field.
+	parentIDField := `"id":"` + stored.ID + `"`
+
+	assert.NotContains(t, marshalModel(t, webhook.Entrypoints[0]), parentIDField)
+	assert.NotContains(t, marshalModel(t, webhook.Targets[0]), parentIDField)
+
+	var target database.Target
+
+	require.NoError(t, db.
+		Preload("Webhook").
+		First(&target, "id = ?", targetID).Error)
+
+	assert.Equal(t, stored.ID, target.Webhook.ID)
+
+	encoded = marshalModel(t, target)
+
+	assert.Contains(t, encoded, stored.ID)
+	assert.NotContains(t, encoded, parentIDField)
+}
+
+// TestModelsMarshalWithoutTheirParent covers the other references to a
+// parent: each model is built with its parent set, and the parent's id
+// must not appear in the JSON.
+func TestModelsMarshalWithoutTheirParent(t *testing.T) {
+	t.Parallel()
+
+	parent := database.BaseModel{ID: uuid.New().String()}
+
+	cases := []struct {
+		name  string
+		model any
+	}{
+		{
+			name:  "Webhook.User",
+			model: database.Webhook{User: database.User{BaseModel: parent}},
+		},
+		{
+			name:  "APIKey.User",
+			model: database.APIKey{User: database.User{BaseModel: parent}},
+		},
+		{
+			name:  "Delivery.Event",
+			model: database.Delivery{Event: database.Event{BaseModel: parent}},
+		},
+		{
+			name:  "Delivery.Target",
+			model: database.Delivery{Target: database.Target{BaseModel: parent}},
+		},
+		{
+			name: "DeliveryResult.Delivery",
+			model: database.DeliveryResult{
+				Delivery: database.Delivery{BaseModel: parent},
+			},
+		},
+		{
+			name:  "Event.Webhook",
+			model: database.Event{Webhook: database.Webhook{BaseModel: parent}},
+		},
+		{
+			name: "Event.Entrypoint",
+			model: database.Event{
+				Entrypoint: database.Entrypoint{BaseModel: parent},
+			},
+		},
+	}
+
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+
+			assert.NotContains(t, marshalModel(t, tc.model), parent.ID)
+		})
+	}
+}

internal/database/model_secrets_test.go

@@ -29,8 +29,10 @@ func marshalModel(t *testing.T, v any) string {
 // these tags the first handler that marshals a model serialises the
 // secret with it. Each field below is a live credential:
 //
-//   - Target.Config holds an incoming-webhook URL whose path segments
-//     are the bearer token.
+//   - Target.URL is an incoming-webhook URL whose path segments are
+//     the bearer token.
+//   - Target.Headers holds header values, routinely authorization
+//     tokens.
 //   - APIKey.Key is a bearer token outright.
 //   - Setting.Value holds the session encryption key.
 //   - User.Password holds the Argon2 hash, and was already tagged.
@@ -44,11 +46,21 @@ func TestModelsDoNotMarshalTheirSecrets(t *testing.T) {
 		model any
 	}{
 		{
-			name: "target config",
+			name: "target url",
 			model: database.Target{
-				Name:   keptField,
-				Type:   database.TargetTypeSlack,
-				Config: `{"webhookUrl":"https://h/s/` + marker + `"}`,
+				Name:        keptField,
+				Type:        database.TargetTypeSlack,
+				URL:         "https://h/s/" + marker,
+				MaxAttempts: 1,
+			},
+		},
+		{
+			name: "target headers",
+			model: database.Target{
+				Name:        keptField,
+				Type:        database.TargetTypeHTTP,
+				Headers:     map[string]string{"Authorization": marker},
+				MaxAttempts: 1,
 			},
 		},
 		{
@@ -86,10 +98,10 @@ func TestModelsDoNotMarshalTheirSecrets(t *testing.T) {
 	}
 }
 
-// TestWebhookMarshalsNoTargetConfig covers the nested case: a webhook
+// TestWebhookMarshalsNoTargetURL covers the nested case: a webhook
 // marshalled with its targets preloaded must not carry the credential
 // through the association either.
-func TestWebhookMarshalsNoTargetConfig(t *testing.T) {
+func TestWebhookMarshalsNoTargetURL(t *testing.T) {
 	t.Parallel()
 
 	const marker = "QQNESTEDMARKERQQ"
@@ -97,8 +109,9 @@ func TestWebhookMarshalsNoTargetConfig(t *testing.T) {
 	encoded := marshalModel(t, database.Webhook{
 		Name: keptField,
 		Targets: []database.Target{{
-			Name:   "slack",
-			Config: `{"webhookUrl":"https://h/s/` + marker + `"}`,
+			Name:        "slack",
+			URL:         "https://h/s/" + marker,
+			MaxAttempts: 1,
 		}},
 	})
 

internal/database/model_target.go

@@ -1,5 +1,7 @@
 package database
 
+import "time"
+
 // TargetType represents the type of delivery target
 type TargetType string
 
@@ -11,7 +13,9 @@ const (
 	TargetTypeSlack    TargetType = "slack"
 )
 
-// Target represents a delivery target for a webhook
+// Target represents a delivery target for a webhook. Its settings
+// are the fields from URL to ArchiveRotation; each type uses some of
+// them, and the ones a type does not use stay empty.
 type Target struct {
 	BaseModel
 
@@ -20,21 +24,45 @@ type Target struct {
 	Type      TargetType `gorm:"not null"           json:"type"`
 	Active    bool       `gorm:"default:true"       json:"active"`
 
-	// Configuration fields (JSON stored based on type).
+	// URL is where an http or slack target sends each event.
 	//
-	// json:"-" because the blob holds the target's credential — a
-	// Slack incoming-webhook URL, or an http destination whose path
-	// segments are the secret. delivery.TargetView is the masking
-	// barrier for the HTML path; this tag is the barrier for any
-	// handler that marshals the model itself.
-	Config string `gorm:"type:text" json:"-"` // JSON configuration
+	// json:"-" because the URL is the target's credential: a Slack
+	// incoming-webhook URL, or an http destination whose path segments
+	// are the secret. delivery.TargetView is the masking barrier for
+	// the HTML path; this tag is the barrier for any handler that
+	// marshals the model itself.
+	URL string `json:"-"`
 
-	// For HTTP targets (max_retries=0 means fire-and-forget,
-	// >0 enables retries with backoff)
-	MaxRetries   int `json:"maxRetries,omitempty"`
-	MaxQueueSize int `json:"maxQueueSize,omitempty"`
+	// Headers is an http target's extra request headers, name to
+	// value, stored as JSON text. json:"-" for the same reason as URL:
+	// a header value is routinely an authorization token.
+	Headers map[string]string `gorm:"serializer:json" json:"-"`
 
-	// Relations
-	Webhook    Webhook    `json:"webhook,omitzero"`
+	// TimeoutSeconds is an http target's per-request timeout; 0 uses
+	// the delivery engine's shared client timeout.
+	TimeoutSeconds int `json:"timeoutSeconds,omitempty"`
+
+	// ForwardQuery passes each event's query string on to an http
+	// target, appended to URL. Off, URL is sent exactly as stored.
+	ForwardQuery bool `json:"forwardQuery,omitempty"`
+
+	// ArchiveExpiry is how long a database target keeps each archived
+	// event; 0 keeps them forever.
+	ArchiveExpiry time.Duration `json:"archiveExpiry,omitempty"`
+
+	// ArchiveRotation is how often a database target starts a new
+	// archive file: none, monthly, daily or hourly.
+	ArchiveRotation string `json:"archiveRotation,omitempty"`
+
+	// MaxAttempts is how many times an HTTP or Slack target attempts
+	// each delivery in all, at least 1. 1 is a single attempt with no
+	// circuit breaker; above 1, a failed attempt is retried with
+	// backoff behind the breaker. The table's check constraint refuses
+	// a target saved with less than 1, whatever its type.
+	MaxAttempts int `gorm:"not null;check:max_attempts >= 1" json:"maxAttempts,omitempty"`
+
+	// Relations. No model marshals the record it belongs to:
+	// Webhook.Targets leads back here, and the JSON could loop.
+	Webhook    Webhook    `json:"-"`
 	Deliveries []Delivery `json:"deliveries,omitempty"`
 }

internal/database/model_target_test.go

@@ -0,0 +1,63 @@
+package database_test
+
+import (
+	"testing"
+
+	"github.com/google/uuid"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/webhooker/internal/database"
+)
+
+// newHTTPTarget returns an http target that makes maxAttempts attempts
+// at each delivery.
+func newHTTPTarget(maxAttempts int) *database.Target {
+	return &database.Target{
+		WebhookID:   uuid.New().String(),
+		Name:        "target",
+		Type:        database.TargetTypeHTTP,
+		URL:         "https://example.com/",
+		MaxAttempts: maxAttempts,
+	}
+}
+
+// TestTargetCreate_RefusesZeroAttempts relies on the targets table
+// alone: nothing in the model checks the number, so the check
+// constraint is what keeps a target with 0 attempts out of the
+// database.
+func TestTargetCreate_RefusesZeroAttempts(t *testing.T) {
+	t.Parallel()
+
+	db := startedTestDB(t)
+
+	err := db.Create(newHTTPTarget(0)).Error
+	require.Error(t, err)
+	assert.Contains(t, err.Error(), "CHECK constraint failed")
+}
+
+func TestTargetCreate_AcceptsOneAttempt(t *testing.T) {
+	t.Parallel()
+
+	db := startedTestDB(t)
+
+	require.NoError(t, db.Create(newHTTPTarget(1)).Error)
+}
+
+// TestTargetHeaders_StoredAsJSON reads the headers column back as
+// text, to show it holds the header map as JSON.
+func TestTargetHeaders_StoredAsJSON(t *testing.T) {
+	t.Parallel()
+
+	db := startedTestDB(t)
+
+	target := newHTTPTarget(1)
+	target.Headers = map[string]string{"Authorization": "Bearer x"}
+	require.NoError(t, db.Create(target).Error)
+
+	var headers string
+
+	require.NoError(t, db.Raw(
+		"SELECT headers FROM targets WHERE id = ?", target.ID,
+	).Scan(&headers).Error)
+	assert.JSONEq(t, `{"Authorization":"Bearer x"}`, headers)
+}

internal/database/model_totals.go

@@ -0,0 +1,136 @@
+package database
+
+import (
+	"fmt"
+	"time"
+
+	"gorm.io/gorm"
+)
+
+// The running totals in a webhook's event database keep the webhook
+// page's lifetime figures right after retention has removed the rows
+// they count, and let the page show them without counting every row.
+// Each total changes in the transaction that writes or deletes the
+// rows it counts.
+
+// EventTotals is the single row counting a webhook's events: every
+// event ever stored, how many of them retention has deleted, and when
+// the newest arrived, which retention leaves as it is.
+type EventTotals struct {
+	ID int64 `gorm:"primaryKey"`
+
+	Events        int64 `gorm:"not null"`
+	EventsRemoved int64 `gorm:"not null"`
+
+	// LastEventAt is when the newest event arrived, or nil before the
+	// first.
+	LastEventAt *time.Time
+}
+
+// TableName names the table AddEventTotals updates.
+func (EventTotals) TableName() string {
+	return "event_totals"
+}
+
+// TargetTotals is one row per target counting its deliveries: every
+// delivery ever created, how many became delivered and how many
+// failed, and how many deliveries and failed deliveries retention has
+// deleted. The webhook's delivery figures are these rows summed.
+type TargetTotals struct {
+	TargetID string `gorm:"type:uuid;primaryKey"`
+
+	Deliveries int64 `gorm:"not null"`
+	Delivered  int64 `gorm:"not null"`
+	Failed     int64 `gorm:"not null"`
+
+	DeliveriesRemoved int64 `gorm:"not null"`
+	FailedRemoved     int64 `gorm:"not null"`
+}
+
+// TableName names the table AddTargetTotals updates.
+func (TargetTotals) TableName() string {
+	return "target_totals"
+}
+
+// EntrypointTotals is one row per entrypoint, created by the first
+// event that arrives on its URL: when the newest such event arrived,
+// which retention leaves as it is. A resubmitted copy did not arrive
+// on the URL and does not change it.
+type EntrypointTotals struct {
+	EntrypointID string `gorm:"type:uuid;primaryKey"`
+
+	LastEventAt time.Time `gorm:"not null"`
+}
+
+// TableName names the table AddEntrypointTotals updates.
+func (EntrypointTotals) TableName() string {
+	return "entrypoint_totals"
+}
+
+// AddEventTotals adds each count in add to the webhook's event totals,
+// and records add.LastEventAt as when the newest event arrived if it is
+// set. Call it on the transaction that writes or deletes the events it
+// counts.
+func AddEventTotals(tx *gorm.DB, add EventTotals) error {
+	err := tx.Exec(
+		`UPDATE event_totals SET
+			events = events + ?,
+			events_removed = events_removed + ?,
+			last_event_at = coalesce(?, last_event_at)`,
+		add.Events, add.EventsRemoved, add.LastEventAt,
+	).Error
+	if err != nil {
+		return fmt.Errorf("adding to event totals: %w", err)
+	}
+
+	return nil
+}
+
+// AddTargetTotals adds each count in add to the totals of the target
+// add.TargetID names, creating its row the first time. Call it on the
+// transaction that writes or deletes the deliveries it counts.
+func AddTargetTotals(tx *gorm.DB, add TargetTotals) error {
+	err := tx.Exec(
+		`INSERT INTO target_totals (target_id, deliveries, delivered,
+			failed, deliveries_removed, failed_removed)
+		VALUES (?, ?, ?, ?, ?, ?)
+		ON CONFLICT (target_id) DO UPDATE SET
+			deliveries = deliveries + excluded.deliveries,
+			delivered = delivered + excluded.delivered,
+			failed = failed + excluded.failed,
+			deliveries_removed =
+				deliveries_removed + excluded.deliveries_removed,
+			failed_removed = failed_removed + excluded.failed_removed`,
+		add.TargetID, add.Deliveries, add.Delivered,
+		add.Failed, add.DeliveriesRemoved, add.FailedRemoved,
+	).Error
+	if err != nil {
+		return fmt.Errorf(
+			"adding to totals of target %s: %w", add.TargetID, err,
+		)
+	}
+
+	return nil
+}
+
+// AddEntrypointTotals records add.LastEventAt as when the newest event
+// arrived on the URL of the entrypoint add.EntrypointID names, creating
+// its row the first time. Call it on the transaction that stores the
+// event.
+func AddEntrypointTotals(tx *gorm.DB, add EntrypointTotals) error {
+	err := tx.Exec(
+		`INSERT INTO entrypoint_totals (entrypoint_id, last_event_at)
+		VALUES (?, ?)
+		ON CONFLICT (entrypoint_id) DO UPDATE SET
+			last_event_at = excluded.last_event_at`,
+		add.EntrypointID, add.LastEventAt,
+	).Error
+	if err != nil {
+		return fmt.Errorf(
+			"adding to totals of entrypoint %s: %w",
+			add.EntrypointID, err,
+		)
+	}
+
+	return nil
+}

internal/database/model_user.go

@@ -1,13 +1,58 @@
 package database
 
+import (
+	"errors"
+	"fmt"
+
+	"gorm.io/gorm"
+)
+
+// MaxUsernameBytes is the longest username, in bytes, that a user may
+// have. The same number appears in the check constraint on
+// User.Username, because a struct tag cannot reference a constant.
+//
+// A login stores the username in the session cookie, and both
+// securecookie and browsers refuse a cookie value past about 4096
+// bytes. That value is the session base64-encoded twice, so it holds
+// 4096 × 3/4 × 3/4 = 2304 bytes of session, and the signature,
+// timestamp and the session's other values take about 270 of those: a
+// username longer than about 2030 bytes can never log in. The limit is
+// about half that, so the session can carry more values later without
+// locking out an account whose username is already at the limit.
+const MaxUsernameBytes = 1024
+
+// ErrUsernameTooLong is returned when a user is saved with a username
+// longer than MaxUsernameBytes.
+var ErrUsernameTooLong = errors.New("username is too long")
+
 // User represents a user of the webhooker service
+//
+//nolint:lll // a struct tag cannot wrap
 type User struct {
 	BaseModel
 
-	Username string `gorm:"uniqueIndex;not null" json:"username"`
-	Password string `gorm:"not null"             json:"-"` // Argon2 hashed
+	Username string `gorm:"uniqueIndex;not null;check:length(CAST(username AS BLOB)) <= 1024" json:"username"`
+	Password string `gorm:"not null"                                                          json:"-"` // Argon2 hashed
 
 	// Relations
 	Webhooks []Webhook `json:"webhooks,omitempty"`
 	APIKeys  []APIKey  `json:"apiKeys,omitempty"`
 }
+
+// BeforeSave rejects a username longer than MaxUsernameBytes when a whole
+// User is created or saved, so those calls get ErrUsernameTooLong rather
+// than the database's constraint error. A column update such as
+// Update("username", ...) is caught only by the check constraint, as is
+// any path that writes the table without this model.
+func (u *User) BeforeSave(_ *gorm.DB) error {
+	if len(u.Username) > MaxUsernameBytes {
+		return fmt.Errorf(
+			"%w: %d bytes, limit is %d",
+			ErrUsernameTooLong,
+			len(u.Username),
+			MaxUsernameBytes,
+		)
+	}
+
+	return nil
+}

internal/database/model_user_test.go

@@ -0,0 +1,65 @@
+package database_test
+
+import (
+	"strings"
+	"testing"
+
+	"github.com/google/uuid"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/webhooker/internal/database"
+)
+
+// usernameAtLimit is exactly MaxUsernameBytes long, built from a
+// two-byte character. A check that counted characters rather than bytes
+// would see half the length and let the one-byte-longer name through.
+func usernameAtLimit() string {
+	return strings.Repeat("é", database.MaxUsernameBytes/2)
+}
+
+func TestUserCreate_RejectsOverlongUsername(t *testing.T) {
+	t.Parallel()
+
+	db := startedTestDB(t)
+
+	err := db.Create(&database.User{
+		Username: usernameAtLimit() + "x",
+		Password: "hash",
+	}).Error
+
+	require.ErrorIs(t, err, database.ErrUsernameTooLong)
+}
+
+func TestUserCreate_AcceptsUsernameAtLimit(t *testing.T) {
+	t.Parallel()
+
+	db := startedTestDB(t)
+
+	require.NoError(t, db.Create(&database.User{
+		Username: usernameAtLimit(),
+		Password: "hash",
+	}).Error)
+}
+
+// TestUsersTable_EnforcesUsernameLimitWithoutTheModel inserts with raw
+// SQL, as a path that bypassed User.BeforeSave would, so only the
+// table's check constraint stands between it and an over-long
+// username. Accepting the name at the limit and refusing the next byte
+// also pins the constraint's number to MaxUsernameBytes.
+func TestUsersTable_EnforcesUsernameLimitWithoutTheModel(t *testing.T) {
+	t.Parallel()
+
+	db := startedTestDB(t)
+
+	insert := "INSERT INTO users (id, username, password) VALUES (?, ?, ?)"
+
+	require.NoError(t, db.Exec(
+		insert, uuid.New().String(), usernameAtLimit(), "hash",
+	).Error)
+
+	err := db.Exec(
+		insert, uuid.New().String(), usernameAtLimit()+"x", "hash",
+	).Error
+	require.Error(t, err)
+	assert.Contains(t, err.Error(), "CHECK constraint failed")
+}

internal/database/model_webhook.go

@@ -4,27 +4,13 @@ import (
 	"math"
 	"strconv"
 	"time"
-
-	"gorm.io/gorm"
 )
 
 const (
-	// DefaultRetentionDays is the event retention period applied to a
-	// webhook created without an explicit retention value. It is the
-	// single source of truth for that policy and must stay in sync
-	// with the `gorm:"default:30"` column default on
-	// Webhook.RetentionDays below; a struct tag cannot reference a
-	// constant, so a test asserts the two agree.
+	// DefaultRetentionDays is the event retention period the new
+	// webhook form applies when its retention field is left blank.
 	DefaultRetentionDays = 30
 
-	// RetentionForeverDays is the sentinel RetentionDays value meaning
-	// "retain events forever". Users express that intent as 0, which
-	// Webhook.BeforeSave rewrites to this value: the column default
-	// substitutes DefaultRetentionDays for a zero value at insert
-	// time, so a zero can never survive a round trip to the database.
-	// Nothing outside this file may hardcode the number.
-	RetentionForeverDays = 365 * 1000
-
 	// MaxFiniteRetentionDays is the largest finite retention period the
 	// reaper's cutoff arithmetic can represent, and therefore the
 	// largest one a caller may request. It is derived from that
@@ -38,9 +24,7 @@ const (
 	// turns the cutoff into a timestamp in the far future that matches
 	// every row in the webhook's database. That is why this bound is
 	// enforced on input and why retentionCutoff saturates underneath
-	// it. Note that RetentionForeverDays deliberately sits above this
-	// ceiling: such webhooks are skipped before any cutoff is
-	// computed, and never reach the arithmetic at all.
+	// it.
 	MaxFiniteRetentionDays = int(
 		math.MaxInt64 / int64(time.Hour) / hoursPerDay,
 	)
@@ -48,12 +32,10 @@ const (
 
 // Webhook represents a webhook processing unit that groups entrypoints and targets
 //
-// Every method below takes a pointer receiver. BeforeSave has to,
-// because it mutates the record and GORM only invokes hooks declared
-// that way; the display helpers follow suit so the receiver kinds do
-// not mix. Handlers therefore put a *Webhook into template data:
-// html/template cannot call a pointer method on a value held in a map,
-// because a map element is not addressable.
+// Every method below takes a pointer receiver, so handlers put a
+// *Webhook into template data: html/template cannot call a pointer
+// method on a value held in a map, because a map element is not
+// addressable.
 type Webhook struct {
 	BaseModel
 
@@ -61,47 +43,28 @@ type Webhook struct {
 	Name        string `gorm:"not null"           json:"name"`
 	Description string `json:"description"`
 
-	// RetentionDays is the number of days to retain events. A value of
-	// RetentionForeverDays means retain forever. The column default
-	// must equal DefaultRetentionDays.
-	RetentionDays int `gorm:"default:30" json:"retentionDays"`
+	// RetentionDays is the number of days to retain events, and 0
+	// means retain them forever.
+	RetentionDays int `json:"retentionDays"`
 
-	// Relations
-	User        User         `json:"user,omitzero"`
+	// Relations. No model marshals the record it belongs to:
+	// User.Webhooks leads back here, and the JSON could loop.
+	User        User         `json:"-"`
 	Entrypoints []Entrypoint `json:"entrypoints,omitempty"`
 	Targets     []Target     `json:"targets,omitempty"`
 }
 
-// BeforeSave normalises RetentionDays on every insert and update. A
-// non-positive value is the user's way of asking for "retain forever",
-// which is stored as the RetentionForeverDays sentinel.
-//
-// This has to happen in a hook rather than at the call sites. GORM
-// substitutes the column default (DefaultRetentionDays) for a zero
-// value while building the insert statement, which runs after
-// BeforeSave; rewriting any later than this loses that race and the
-// row lands at 30 days. Living on the model also means a future call
-// site — a REST API, a fixture, a migration — cannot bypass it.
-func (w *Webhook) BeforeSave(_ *gorm.DB) error {
-	if w.RetentionDays <= 0 {
-		w.RetentionDays = RetentionForeverDays
-	}
-
-	return nil
-}
-
-// retainsForever reports whether a stored RetentionDays value means
-// "keep events indefinitely". It is the single definition of that
-// question, shared by Webhook.RetainsForever and by the reaper's
+// retainsForever reports whether a RetentionDays value means "keep
+// events indefinitely", which 0 does. It is the single definition of
+// that question, shared by Webhook.RetainsForever and by the reaper's
 // cutoff computation so the two cannot disagree about which webhooks
 // are exempt from reaping.
 //
-// It accepts the RetentionForeverDays sentinel written by BeforeSave
-// and, defensively, the non-positive values that rows written before
-// the sentinel existed may still carry.
+// A negative value, which the forms refuse, is read the same way: as a
+// day count it would put the cutoff in the future, where it matches
+// every event the webhook has.
 func retainsForever(retentionDays int) bool {
-	return retentionDays <= 0 ||
-		retentionDays >= RetentionForeverDays
+	return retentionDays <= 0
 }
 
 // RetainsForever reports whether this webhook's events are kept
@@ -110,8 +73,16 @@ func (w *Webhook) RetainsForever() bool {
 	return retainsForever(w.RetentionDays)
 }
 
+// RetentionCutoff returns the time before which this webhook's events
+// have expired, as the reaper computes it, and false when the webhook
+// retains them forever.
+func (w *Webhook) RetentionCutoff(now time.Time) (time.Time, bool) {
+	return retentionCutoff(now, w.RetentionDays)
+}
+
 // RetentionLabel returns the webhook's retention policy as display
-// text, so that no template has to know about the sentinel value.
+// text: "forever" rather than "0 days" for a webhook that keeps its
+// events.
 func (w *Webhook) RetentionLabel() string {
 	if w.RetainsForever() {
 		return "forever"

internal/database/model_webhook_test.go

@@ -2,8 +2,6 @@ package database_test
 
 import (
 	"context"
-	"reflect"
-	"strconv"
 	"testing"
 	"time"
 
@@ -47,8 +45,7 @@ func storedRetention(t *testing.T, db *gorm.DB, id string) int {
 }
 
 // newWebhookWithRetention creates a webhook through the ordinary Create
-// path, so the BeforeSave hook and the GORM column default both apply
-// exactly as they do in production.
+// path, as the new webhook form does.
 func newWebhookWithRetention(
 	t *testing.T,
 	db *gorm.DB,
@@ -67,7 +64,7 @@ func newWebhookWithRetention(
 	return wh.ID
 }
 
-func TestWebhookBeforeSave_ZeroBecomesForeverSentinel(t *testing.T) {
+func TestWebhookRetentionDays_ZeroIsStoredAsZero(t *testing.T) {
 	t.Parallel()
 
 	db := startedTestDB(t)
@@ -75,31 +72,10 @@ func TestWebhookBeforeSave_ZeroBecomesForeverSentinel(t *testing.T) {
 	wh := &database.Webhook{RetentionDays: 0}
 	id := newWebhookWithRetention(t, db, wh)
 
-	assert.Equal(
-		t,
-		database.RetentionForeverDays,
-		storedRetention(t, db, id),
-		"a zero retention must be stored as the sentinel, "+
-			"not replaced by the column default",
-	)
+	assert.Equal(t, 0, storedRetention(t, db, id))
 }
 
-func TestWebhookBeforeSave_NegativeBecomesForeverSentinel(t *testing.T) {
-	t.Parallel()
-
-	db := startedTestDB(t)
-
-	wh := &database.Webhook{RetentionDays: -5}
-	id := newWebhookWithRetention(t, db, wh)
-
-	assert.Equal(
-		t,
-		database.RetentionForeverDays,
-		storedRetention(t, db, id),
-	)
-}
-
-func TestWebhookBeforeSave_PositiveIsPreserved(t *testing.T) {
+func TestWebhookRetentionDays_DayCountIsStored(t *testing.T) {
 	t.Parallel()
 
 	db := startedTestDB(t)
@@ -110,10 +86,9 @@ func TestWebhookBeforeSave_PositiveIsPreserved(t *testing.T) {
 	assert.Equal(t, 7, storedRetention(t, db, id))
 }
 
-// TestWebhookBeforeSave_UpdateToZeroBecomesSentinel proves the hook
-// fires on update as well as insert, via the same Save call the edit
-// handler makes.
-func TestWebhookBeforeSave_UpdateToZeroBecomesSentinel(t *testing.T) {
+// TestWebhookRetentionDays_UpdateToZeroIsStoredAsZero changes a day
+// count to 0 through the same Save call the edit handler makes.
+func TestWebhookRetentionDays_UpdateToZeroIsStoredAsZero(t *testing.T) {
 	t.Parallel()
 
 	db := startedTestDB(t)
@@ -125,28 +100,7 @@ func TestWebhookBeforeSave_UpdateToZeroBecomesSentinel(t *testing.T) {
 	wh.RetentionDays = 0
 	require.NoError(t, db.Omit(clause.Associations).Save(wh).Error)
 
-	assert.Equal(
-		t,
-		database.RetentionForeverDays,
-		storedRetention(t, db, id),
-	)
-}
-
-// TestWebhookRetentionColumnDefaultMatchesConstant guards the one place
-// the default lives twice: a struct tag cannot reference a constant, so
-// this asserts the tag and DefaultRetentionDays agree.
-func TestWebhookRetentionColumnDefaultMatchesConstant(t *testing.T) {
-	t.Parallel()
-
-	field, ok := reflect.TypeFor[database.Webhook]().
-		FieldByName("RetentionDays")
-	require.True(t, ok, "Webhook.RetentionDays must exist")
-
-	assert.Equal(
-		t,
-		"default:"+strconv.Itoa(database.DefaultRetentionDays),
-		field.Tag.Get("gorm"),
-	)
+	assert.Equal(t, 0, storedRetention(t, db, id))
 }
 
 // TestMaxFiniteRetentionDaysIsTheOverflowCeiling asserts that the
@@ -176,14 +130,6 @@ func TestMaxFiniteRetentionDaysIsTheOverflowCeiling(t *testing.T) {
 		time.Duration(overCeiling*hoursPerDay)*time.Hour,
 		"one day past the ceiling must overflow",
 	)
-
-	assert.Less(
-		t,
-		database.MaxFiniteRetentionDays,
-		database.RetentionForeverDays,
-		"the sentinel sits above the ceiling and is only safe "+
-			"because retain-forever webhooks skip the arithmetic",
-	)
 }
 
 func TestWebhookRetainsForeverAndLabel(t *testing.T) {
@@ -195,16 +141,8 @@ func TestWebhookRetainsForeverAndLabel(t *testing.T) {
 		forever bool
 		label   string
 	}{
-		{
-			"sentinel",
-			database.RetentionForeverDays, true, testForeverLabel,
-		},
-		{
-			"above sentinel",
-			database.RetentionForeverDays + 1, true, testForeverLabel,
-		},
-		{"legacy zero", 0, true, testForeverLabel},
-		{"legacy negative", -1, true, testForeverLabel},
+		{"zero", 0, true, testForeverLabel},
+		{"negative", -1, true, testForeverLabel},
 		{"default", database.DefaultRetentionDays, false, "30 days"},
 		{"one day", 1, false, "1 day"},
 	}

internal/database/models.go

@@ -1,16 +0,0 @@
-package database
-
-// Migrate runs database migrations for the main application database.
-// Only configuration-tier models are stored in the main database.
-// Event-tier models (Event, Delivery, DeliveryResult) live in
-// per-webhook dedicated databases managed by WebhookDBManager.
-func (d *Database) Migrate() error {
-	return d.db.AutoMigrate(
-		&Setting{},
-		&User{},
-		&APIKey{},
-		&Webhook{},
-		&Entrypoint{},
-		&Target{},
-	)
-}

internal/database/password.go

@@ -9,6 +9,7 @@ import (
 	"math/big"
 	"strings"
 	"sync"
+	"testing"
 
 	"golang.org/x/crypto/argon2"
 )
@@ -63,10 +64,31 @@ func DefaultPasswordConfig() *PasswordConfig {
 	}
 }
 
-// HashPassword generates an Argon2id hash of the password
+// testArgon2Memory is the Argon2id memory cost, in KiB, that a test
+// binary hashes with: 1 MB instead of the shipped 64 MB. Every test
+// that starts a database hashes the bootstrap admin password, dozens
+// of them run in parallel, and each 64 MB hash holds at least 64 MB
+// while it runs (about 150 MB when the suite ran under the race
+// detector). VerifyPassword reads the cost from the hash it checks, so
+// verification follows.
+const testArgon2Memory = 1024
+
+// hashAtShippedCostInTest makes a test binary hash at the shipped
+// memory cost. Only TestHashPassword_ShippedParameters sets it.
+//
+//nolint:gochecknoglobals // set by one test, see above
+var hashAtShippedCostInTest bool
+
+// HashPassword generates an Argon2id hash of the password. A binary
+// built by go test hashes at testArgon2Memory; one built by go build
+// always hashes at the defaults.
 func HashPassword(password string) (string, error) {
 	config := DefaultPasswordConfig()
 
+	if testing.Testing() && !hashAtShippedCostInTest {
+		config.Memory = testArgon2Memory
+	}
+
 	// Generate a salt
 	salt := make([]byte, config.SaltLen)
 

internal/database/password_test.go

@@ -192,6 +192,39 @@ func TestHashPasswordUniqueness(t *testing.T) {
 	}
 }
 
+// TestHashPassword_ShippedParameters hashes and verifies through
+// HashPassword at the shipped Argon2id parameters. Every other test
+// hashes at the lower memory cost a test binary uses, so this is the
+// one that keeps production hashing covered. One hash and one
+// verification: each costs 64 MB.
+//
+//nolint:paralleltest // changes the hashing cost for the whole binary
+func TestHashPassword_ShippedParameters(t *testing.T) {
+	database.HashAtShippedCostForTest(t)
+
+	password := "correct horse battery staple"
+
+	hash, err := database.HashPassword(password)
+	if err != nil {
+		t.Fatalf("hashing with the shipped parameters: %v", err)
+	}
+
+	const shipped = "$argon2id$v=19$m=65536,t=1,p=4$"
+
+	if !strings.HasPrefix(hash, shipped) {
+		t.Errorf("hash = %q, want prefix %q", hash, shipped)
+	}
+
+	valid, err := database.VerifyPassword(password, hash)
+	if err != nil {
+		t.Fatalf("VerifyPassword() error = %v", err)
+	}
+
+	if !valid {
+		t.Error("VerifyPassword() returned false for correct password")
+	}
+}
+
 // TestVerifyDummyPassword_DoesRealWork covers the anti-enumeration
 // path. Login charges an unknown username a verification against a
 // dummy hash so that a nonexistent account is not answered in

internal/database/retention.go

@@ -18,6 +18,19 @@ import (
 // computation.
 const hoursPerDay = 24
 
+// reapBatchSize is how many expired events one retention transaction
+// deletes. A transaction holds the event database's write lock, which
+// the receiver and the delivery workers wait for, so a large prune is
+// split into transactions each short enough to finish well inside the
+// busy timeout.
+const reapBatchSize = 1000
+
+// reapBatchPause is how long retention waits after one batch before
+// starting the next. A writer waiting for the write lock checks for it
+// again after at most 100 ms, so a longer pause lets it in between two
+// batches instead of only after the whole prune.
+const reapBatchPause = 200 * time.Millisecond
+
 // RetentionReaperParams holds the fx dependencies for the
 // RetentionReaper.
 type RetentionReaperParams struct {
@@ -171,32 +184,32 @@ func (r *RetentionReaper) sweep(ctx context.Context) {
 
 		wh := webhooks[i]
 
-		// Skip retain-forever webhooks before building any query.
-		// RetainsForever covers both the RetentionForeverDays
-		// sentinel and the non-positive values that predate it: the
-		// sentinel is a positive number, so without this the reaper
-		// would compute a cutoff a thousand years in the past and
-		// issue a DELETE matching nothing on every single sweep.
-		if wh.RetainsForever() {
-			continue
-		}
-
-		// Nothing to reap if the per-webhook database has never
-		// been created.
+		// A missing database has nothing to reap. The delivery
+		// engine's recovery routine reports a lost one at start (see
+		// WebhookDBManager.GetDB).
 		if !r.dbManager.DBExists(wh.ID) {
 			continue
 		}
 
-		r.reapWebhook(wh.ID, wh.RetentionDays)
+		r.reapWebhook(ctx, wh.ID, wh.RetentionDays)
 	}
 }
 
 // reapWebhook removes every expired event (and its dependents) from a
-// single webhook's database.
+// single webhook's database, or as many as it reaches before ctx is
+// cancelled.
 func (r *RetentionReaper) reapWebhook(
+	ctx context.Context,
 	webhookID string,
 	retentionDays int,
 ) {
+	// A retain-forever webhook has no cutoff, so its database is not
+	// even opened.
+	cutoff, ok := retentionCutoff(time.Now(), retentionDays)
+	if !ok {
+		return
+	}
+
 	db, err := r.dbManager.GetDB(webhookID)
 	if err != nil {
 		r.log.Error(
@@ -208,12 +221,7 @@ func (r *RetentionReaper) reapWebhook(
 		return
 	}
 
-	cutoff, ok := retentionCutoff(time.Now(), retentionDays)
-	if !ok {
-		return
-	}
-
-	deleted, err := reapExpired(db, cutoff)
+	deleted, err := reapExpired(ctx, db, cutoff)
 	if err != nil {
 		r.log.Error(
 			"retention sweep: failed to reap expired events",
@@ -236,7 +244,8 @@ func (r *RetentionReaper) reapWebhook(
 
 // retentionCutoff returns the timestamp before which a webhook's
 // events have expired, and whether any cutoff applies at all. It
-// reports false for a retain-forever policy, so no DELETE is issued.
+// reports false for a retention of 0, which means retain forever, so
+// no DELETE is issued.
 //
 // The day count is clamped to MaxFiniteRetentionDays first. This is
 // defense in depth rather than decoration: a time.Duration is an int64
@@ -246,7 +255,7 @@ func (r *RetentionReaper) reapWebhook(
 // row in the database: the sweep then deletes every event, delivery,
 // and delivery result, including ones created seconds ago. Rejecting
 // out-of-range input at the form is the primary guard; saturating here
-// means an old row, a migration, or a future call site cannot turn a
+// means a value that reaches the row by any other path cannot turn a
 // too-large retention into total data loss.
 func retentionCutoff(
 	now time.Time,
@@ -265,57 +274,107 @@ func retentionCutoff(
 	), true
 }
 
-// reapExpired hard-deletes, in foreign-key-safe order, the delivery
-// results, deliveries, and events associated with events older than
-// cutoff. Deletes are unscoped so rows are physically removed rather
-// than soft-deleted, reclaiming disk. It returns the number of events
-// deleted.
-func reapExpired(db *gorm.DB, cutoff time.Time) (int64, error) {
-	// Fresh subqueries are built per statement to avoid reusing a
-	// mutated builder across executions.
-	expiredEventIDs := func() *gorm.DB {
-		return db.Model(&Event{}).
-			Select("id").
-			Where("created_at < ?", cutoff)
-	}
-	expiredDeliveryIDs := func() *gorm.DB {
-		return db.Model(&Delivery{}).
-			Select("id").
-			Where("event_id IN (?)", expiredEventIDs())
-	}
+// reapExpired hard-deletes the events older than cutoff, with their
+// deliveries and delivery results, reapBatchSize events per
+// transaction with reapBatchPause between transactions, until none is
+// left. Once ctx is cancelled it returns after the batch in hand,
+// leaving the rest to the next sweep, so stopping the app does not
+// wait for a long prune. It returns the number of events deleted.
+func reapExpired(
+	ctx context.Context, db *gorm.DB, cutoff time.Time,
+) (int64, error) {
+	var total int64
 
-	// 1. Delivery results whose delivery belongs to an expired event.
-	res := db.Unscoped().
-		Where("delivery_id IN (?)", expiredDeliveryIDs()).
-		Delete(&DeliveryResult{})
-	if res.Error != nil {
-		return 0, fmt.Errorf(
-			"deleting expired delivery results: %w",
-			res.Error,
-		)
-	}
+	for {
+		var eventIDs []string
 
-	// 2. Deliveries belonging to an expired event.
-	del := db.Unscoped().
-		Where("event_id IN (?)", expiredEventIDs()).
-		Delete(&Delivery{})
-	if del.Error != nil {
-		return 0, fmt.Errorf(
-			"deleting expired deliveries: %w",
-			del.Error,
-		)
-	}
+		err := db.Transaction(func(tx *gorm.DB) error {
+			err := tx.Unscoped().Model(&Event{}).
+				Where("created_at < ?", cutoff).
+				Limit(reapBatchSize).
+				Pluck("id", &eventIDs).Error
+			if err != nil {
+				return fmt.Errorf("selecting expired events: %w", err)
+			}
 
-	// 3. The expired events themselves.
-	ev := db.Unscoped().
-		Where("created_at < ?", cutoff).
-		Delete(&Event{})
-	if ev.Error != nil {
-		return 0, fmt.Errorf(
-			"deleting expired events: %w",
-			ev.Error,
-		)
-	}
+			if len(eventIDs) == 0 {
+				return nil
+			}
 
-	return ev.RowsAffected, nil
+			return deleteEvents(tx, eventIDs)
+		})
+		if err != nil {
+			return total, err
+		}
+
+		total += int64(len(eventIDs))
+
+		if len(eventIDs) < reapBatchSize {
+			return total, nil
+		}
+
+		select {
+		case <-ctx.Done():
+			return total, nil
+		case <-time.After(reapBatchPause):
+		}
+	}
+}
+
+// deleteEvents hard-deletes the given events and, in foreign-key-safe
+// order before them, their delivery results and deliveries, then adds
+// what it deleted to the running totals. It runs on reapExpired's
+// transaction, so the totals change exactly when the rows do. Deletes
+// are unscoped so rows are physically removed rather than
+// soft-deleted, reclaiming disk.
+func deleteEvents(tx *gorm.DB, eventIDs []string) error {
+	// 1. The delivery results of the events' deliveries.
+	err := tx.Unscoped().
+		Where("delivery_id IN (?)", tx.Unscoped().Model(&Delivery{}).
+			Select("id").
+			Where("event_id IN ?", eventIDs)).
+		Delete(&DeliveryResult{}).Error
+	if err != nil {
+		return fmt.Errorf("deleting expired delivery results: %w", err)
+	}
+
+	// 2. The events' deliveries, after counting them, and the failed
+	// ones among them, per target. The status is tested in the select
+	// list rather than the WHERE clause: there, SQLite would read every
+	// failed delivery the webhook has through the status index,
+	// instead of only these through the event_id index.
+	var removed []TargetTotals
+
+	err = tx.Unscoped().Model(&Delivery{}).
+		Select("target_id, count(*) AS deliveries_removed, "+
+			"count(CASE WHEN status = ? THEN 1 END) AS failed_removed",
+			DeliveryStatusFailed).
+		Where("event_id IN ?", eventIDs).
+		Group("target_id").
+		Find(&removed).Error
+	if err != nil {
+		return fmt.Errorf("counting expired deliveries: %w", err)
+	}
+
+	err = tx.Unscoped().
+		Where("event_id IN ?", eventIDs).
+		Delete(&Delivery{}).Error
+	if err != nil {
+		return fmt.Errorf("deleting expired deliveries: %w", err)
+	}
+
+	// 3. The events themselves.
+	ev := tx.Unscoped().Where("id IN ?", eventIDs).Delete(&Event{})
+	if ev.Error != nil {
+		return fmt.Errorf("deleting expired events: %w", ev.Error)
+	}
+
+	for i := range removed {
+		err = AddTargetTotals(tx, removed[i])
+		if err != nil {
+			return err
+		}
+	}
+
+	return AddEventTotals(tx, EventTotals{EventsRemoved: ev.RowsAffected})
 }

internal/database/retention_test.go

@@ -51,7 +51,7 @@ func setupRetentionTest(t *testing.T) *retentionTestEnv {
 
 	mgr, err := database.NewWebhookDBManager(
 		lc,
-		database.WebhookDBManagerParams{Config: cfg, Logger: l},
+		database.WebhookDBManagerParams{Config: cfg, Logger: l.Get()},
 	)
 	require.NoError(t, err)
 
@@ -85,41 +85,6 @@ func createWebhook(
 		db.Omit(clause.Associations).Create(wh).Error,
 	)
 
-	// Webhook.BeforeSave rewrites a non-positive RetentionDays to the
-	// retain-forever sentinel, and the column's GORM default would
-	// otherwise substitute 30. Force the requested value with a
-	// column-level update so tests can plant legacy rows that predate
-	// the sentinel and still carry a literal 0 or negative value.
-	require.NoError(
-		t,
-		db.Model(wh).
-			Update("retention_days", retentionDays).Error,
-	)
-
-	return wh.ID
-}
-
-// createWebhookNormally inserts a webhook through the ordinary Create
-// path, with no column-level forcing, so Webhook.BeforeSave applies
-// exactly as it does in production. Passing 0 therefore yields a row
-// holding the RetentionForeverDays sentinel.
-func createWebhookNormally(
-	t *testing.T,
-	db *gorm.DB,
-	retentionDays int,
-) string {
-	t.Helper()
-
-	wh := &database.Webhook{
-		UserID:        uuid.New().String(),
-		Name:          testWebhookName,
-		RetentionDays: retentionDays,
-	}
-	require.NoError(
-		t,
-		db.Omit(clause.Associations).Create(wh).Error,
-	)
-
 	return wh.ID
 }
 
@@ -281,19 +246,19 @@ func TestRetentionReaper_ReapsExpiredKeepsRecent(t *testing.T) {
 	assertChainPresent(t, db, recent)
 }
 
-// TestRetentionReaper_SkipsSentinelReapsFiniteInSameSweep covers the
-// end-to-end retain-forever path: a webhook created the normal way with
-// a requested retention of 0 lands on the RetentionForeverDays
-// sentinel, and the reaper leaves its ancient events alone while still
-// reaping a finite-retention webhook in the very same sweep.
-func TestRetentionReaper_SkipsSentinelReapsFiniteInSameSweep(
+// TestRetentionReaper_SkipsZeroReapsFiniteInSameSweep covers the
+// end-to-end retain-forever path: a webhook created with a retention
+// of 0 reads back 0, and the reaper leaves its ancient events alone
+// while still reaping a finite-retention webhook in the very same
+// sweep.
+func TestRetentionReaper_SkipsZeroReapsFiniteInSameSweep(
 	t *testing.T,
 ) {
 	t.Parallel()
 
 	env := setupRetentionTest(t)
 
-	foreverID := createWebhookNormally(t, env.mainDB.DB(), 0)
+	foreverID := createWebhook(t, env.mainDB.DB(), 0)
 
 	var stored database.Webhook
 
@@ -302,14 +267,12 @@ func TestRetentionReaper_SkipsSentinelReapsFiniteInSameSweep(
 		env.mainDB.DB().Where("id = ?", foreverID).
 			First(&stored).Error,
 	)
-	require.Equal(
-		t,
-		database.RetentionForeverDays,
-		stored.RetentionDays,
-		"a requested retention of 0 must persist as the sentinel",
+	require.Zero(
+		t, stored.RetentionDays,
+		"a retention of 0 must be stored as 0",
 	)
 
-	finiteID := createWebhookNormally(t, env.mainDB.DB(), 30)
+	finiteID := createWebhook(t, env.mainDB.DB(), 30)
 
 	foreverDB, err := env.mgr.GetDB(foreverID)
 	require.NoError(t, err)
@@ -335,14 +298,10 @@ func TestRetentionReaper_SkipsSentinelReapsFiniteInSameSweep(
 // negative, so subtracting it moves the cutoff into the far future,
 // where "created_at < cutoff" matches every row: an event created a
 // moment ago, and its delivery and delivery result, were all deleted
-// on the first sweep. 200000 is inside that band and below the
-// retain-forever sentinel, so it is treated as a finite policy and
-// really does reach the arithmetic.
+// on the first sweep. 200000 is inside that band.
 //
-// The row is planted at the column level because such a value can no
-// longer be submitted through the form; the point of the test is that
-// a row from an older version, or a future call site, still cannot
-// trigger the wipe.
+// The form refuses such a value; the point of the test is that a row
+// written by any other path still cannot trigger the wipe.
 func TestRetentionReaper_HugeFiniteRetentionRetainsRecentEvents(
 	t *testing.T,
 ) {
@@ -358,12 +317,6 @@ func TestRetentionReaper_HugeFiniteRetentionRetainsRecentEvents(
 		database.MaxFiniteRetentionDays,
 		"the test value must exceed what the cutoff can represent",
 	)
-	require.Less(
-		t,
-		overflowingRetentionDays,
-		database.RetentionForeverDays,
-		"the test value must not be rescued by the forever skip",
-	)
 
 	webhookID := createWebhook(
 		t, env.mainDB.DB(), overflowingRetentionDays,
@@ -379,14 +332,15 @@ func TestRetentionReaper_HugeFiniteRetentionRetainsRecentEvents(
 	assertChainPresent(t, db, fresh)
 }
 
-func TestRetentionReaper_RetainsForeverWhenNonPositive(t *testing.T) {
+// TestRetentionReaper_RetainsForeverWhenNegative checks that a negative
+// retention, which the forms refuse, keeps every event rather than
+// putting the cutoff in the future, where it would match them all.
+func TestRetentionReaper_RetainsForeverWhenNegative(t *testing.T) {
 	t.Parallel()
 
 	env := setupRetentionTest(t)
 
-	// A legacy row written before the sentinel existed still carries a
-	// literal 0; the <= 0 guard must keep honouring it.
-	webhookID := createWebhook(t, env.mainDB.DB(), 0)
+	webhookID := createWebhook(t, env.mainDB.DB(), -1)
 
 	db, err := env.mgr.GetDB(webhookID)
 	require.NoError(t, err)

internal/database/schema_test.go

@@ -0,0 +1,173 @@
+package database_test
+
+import (
+	"log/slog"
+	"net/http"
+	"testing"
+	"time"
+
+	"github.com/google/uuid"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"gorm.io/gorm"
+	"sneak.berlin/go/webhooker/internal/database"
+	"sneak.berlin/go/webhooker/internal/database/databasetest"
+)
+
+// withFixedNow returns db with GORM's clock stopped at a time in UTC,
+// so that the created and updated times it writes come back from
+// SQLite equal to the row written, which holds them in UTC.
+func withFixedNow(db *gorm.DB) (*gorm.DB, time.Time) {
+	at := time.Date(2026, 10, 13, 12, 30, 45, 123456789, time.UTC)
+
+	return db.Session(&gorm.Session{
+		NowFunc: func() time.Time { return at },
+	}), at
+}
+
+// newBaseModel returns a BaseModel with a new ID.
+func newBaseModel() database.BaseModel {
+	return database.BaseModel{ID: uuid.New().String()}
+}
+
+// assertRoundTrip writes row, which must be the only row of its
+// table, and reads back the same row.
+func assertRoundTrip[T any](t *testing.T, db *gorm.DB, row *T) {
+	t.Helper()
+
+	require.NoError(t, db.Create(row).Error)
+
+	var got T
+
+	require.NoError(t, db.First(&got).Error)
+	assert.Equal(t, *row, got)
+}
+
+// TestMainSchemaHoldsEveryModel writes a row of each main database
+// model, every field set, into a database made from its SQL files.
+func TestMainSchemaHoldsEveryModel(t *testing.T) {
+	t.Parallel()
+
+	mainDB, err := database.Open(
+		t.TempDir(), slog.New(slog.DiscardHandler),
+	)
+	require.NoError(t, err)
+
+	t.Cleanup(func() { _ = mainDB.Close() })
+
+	db, at := withFixedNow(mainDB.DB())
+
+	assertRoundTrip(t, db, &database.Setting{
+		Key: "session_key", Value: "value",
+	})
+	assertRoundTrip(t, db, &database.User{
+		BaseModel: newBaseModel(),
+		Username:  "user",
+		Password:  "password hash",
+	})
+	assertRoundTrip(t, db, &database.APIKey{
+		BaseModel:   newBaseModel(),
+		UserID:      uuid.New().String(),
+		Key:         "key",
+		Description: "API key description",
+		LastUsedAt:  &at,
+	})
+	assertRoundTrip(t, db, &database.Webhook{
+		BaseModel:     newBaseModel(),
+		UserID:        uuid.New().String(),
+		Name:          "webhook",
+		Description:   "webhook description",
+		RetentionDays: 7,
+	})
+	assertRoundTrip(t, db, &database.Entrypoint{
+		BaseModel:   newBaseModel(),
+		WebhookID:   uuid.New().String(),
+		Path:        uuid.New().String(),
+		Description: "entrypoint description",
+		Active:      true,
+	})
+	assertRoundTrip(t, db, &database.Target{
+		BaseModel:       newBaseModel(),
+		WebhookID:       uuid.New().String(),
+		Name:            "target",
+		Type:            database.TargetTypeHTTP,
+		Active:          true,
+		URL:             "https://example.com/",
+		Headers:         map[string]string{"X-Api-Key": "key"},
+		TimeoutSeconds:  30,
+		ForwardQuery:    true,
+		ArchiveExpiry:   720 * time.Hour,
+		ArchiveRotation: "daily",
+		MaxAttempts:     3,
+	})
+}
+
+// TestEventSchemaHoldsEveryModel does the same for a webhook's event
+// database, which the manager makes from its SQL files.
+func TestEventSchemaHoldsEveryModel(t *testing.T) {
+	t.Parallel()
+
+	mgr := databasetest.NewWebhookDBManager(t, t.TempDir())
+
+	t.Cleanup(func() { _ = mgr.CloseAll() })
+
+	events, err := mgr.GetDB(uuid.New().String())
+	require.NoError(t, err)
+
+	db, at := withFixedNow(events)
+	source := uuid.New().String()
+
+	assertRoundTrip(t, db, &database.Event{
+		BaseModel:         newBaseModel(),
+		WebhookID:         uuid.New().String(),
+		EntrypointID:      uuid.New().String(),
+		Method:            http.MethodPost,
+		RawQuery:          "a=1",
+		Headers:           `{"X":"Y"}`,
+		Body:              "body",
+		ContentType:       "text/plain",
+		BodyBytes:         4,
+		ResubmittedFromID: &source,
+	})
+	assertRoundTrip(t, db, &database.Delivery{
+		BaseModel:  newBaseModel(),
+		EventID:    uuid.New().String(),
+		TargetID:   uuid.New().String(),
+		Status:     database.DeliveryStatusDelivered,
+		FinishedAt: &at,
+		Replay:     true,
+	})
+	assertRoundTrip(t, db, &database.DeliveryResult{
+		BaseModel:    newBaseModel(),
+		DeliveryID:   uuid.New().String(),
+		AttemptNum:   2,
+		Success:      true,
+		StatusCode:   200,
+		ResponseBody: "ok",
+		Error:        "error",
+		Duration:     15,
+	})
+	assertRoundTrip(t, db, &database.TargetTotals{
+		TargetID:          uuid.New().String(),
+		Deliveries:        5,
+		Delivered:         3,
+		Failed:            1,
+		DeliveriesRemoved: 2,
+		FailedRemoved:     1,
+	})
+	assertRoundTrip(t, db, &database.EntrypointTotals{
+		EntrypointID: uuid.New().String(), LastEventAt: at,
+	})
+
+	// A new event database has its one row of event totals already.
+	require.NoError(t, database.AddEventTotals(db, database.EventTotals{
+		Events: 3, EventsRemoved: 1, LastEventAt: &at,
+	}))
+
+	var totals []database.EventTotals
+
+	require.NoError(t, db.Find(&totals).Error)
+	assert.Equal(t, []database.EventTotals{{
+		ID: 1, Events: 3, EventsRemoved: 1, LastEventAt: &at,
+	}}, totals)
+}

internal/database/sqlite_mode_test.go

@@ -137,7 +137,7 @@ func TestPerWebhookEventDatabaseFilesAreOwnerOnly(t *testing.T) {
 }
 
 // TestArchiveDatabaseFilesAreOwnerOnly covers the archive-*.db tier.
-// internal/delivery builds that path and opens it through OpenSQLite,
+// internal/archive builds that path and opens it through OpenSQLite,
 // the same single open path exercised here, so the mode is settled for
 // all three tiers in one place.
 func TestArchiveDatabaseFilesAreOwnerOnly(t *testing.T) {
@@ -182,6 +182,29 @@ func TestOpenSQLiteTightensFilesLeftWorldReadable(t *testing.T) {
 	requireDatabaseSetOwnerOnly(t, path)
 }
 
+// TestOpenSQLiteRefusesADirectorySidecar covers a directory in place
+// of -wal or -shm. Beside a -shm directory SQLite opens the database
+// read-only without a word, and every write then fails naming no file,
+// so the open must stop instead, naming the directory.
+func TestOpenSQLiteRefusesADirectorySidecar(t *testing.T) {
+	t.Parallel()
+
+	for _, suffix := range []string{"-wal", "-shm"} {
+		t.Run(suffix, func(t *testing.T) {
+			t.Parallel()
+
+			path := filepath.Join(t.TempDir(), database.MainDBFileName)
+			require.NoError(t, os.Mkdir(path+suffix, 0o700))
+
+			_, err := database.OpenSQLite(
+				path, database.SQLiteModeCreate,
+			)
+			require.Error(t, err)
+			assert.Contains(t, err.Error(), path+suffix)
+		})
+	}
+}
+
 // TestOpenSQLiteExistingModeDoesNotCreateTheFile guards the mechanism
 // the fix uses: OpenSQLite now creates the database file itself, and
 // must not do so for a caller that asked for an existing database. An

internal/database/sqlite_open.go

@@ -7,6 +7,7 @@ import (
 	"io/fs"
 	"net/url"
 	"os"
+	"syscall"
 	"time"
 
 	_ "modernc.org/sqlite" // Pure Go SQLite driver
@@ -93,7 +94,8 @@ const (
 const SQLiteFilePerm fs.FileMode = 0o600
 
 // reserveSQLiteFile puts path at SQLiteFilePerm before the driver ever
-// touches it, and tightens any sidecar already on disk.
+// touches it, and tightens any sidecar already on disk. A directory in
+// place of any of them is an error naming it.
 //
 // The mode has to be settled here rather than by a chmod after opening,
 // because SQLite picks it: robust_open substitutes
@@ -116,9 +118,8 @@ const SQLiteFilePerm fs.FileMode = 0o600
 // SQLite still reports the absence rather than this function
 // materializing an empty database the caller asked not to create.
 //
-// Chmod of a file that already exists is what tightens a data
-// directory an earlier build left at 0644 — including a developer's
-// own scratch directory — without any migration machinery.
+// The mode is set on every open, whoever created the file, so a
+// database or sidecar found at a wider mode is tightened too.
 func reserveSQLiteFile(path string, create bool) error {
 	if create {
 		// gosec G304: the path is the database file the caller asked
@@ -143,7 +144,15 @@ func reserveSQLiteFile(path string, create bool) error {
 	for _, p := range append(
 		[]string{path}, sqliteSidecarPaths(path)...,
 	) {
-		err := os.Chmod(p, SQLiteFilePerm)
+		// Chmod accepts a directory, and SQLite opens a database whose
+		// -shm is one read-only, without a word: every write then
+		// fails naming no file.
+		info, err := os.Stat(p)
+		if err == nil && info.IsDir() {
+			return fmt.Errorf("securing %s: %w", p, syscall.EISDIR)
+		}
+
+		err = os.Chmod(p, SQLiteFilePerm)
 		if err != nil && !errors.Is(err, fs.ErrNotExist) {
 			return fmt.Errorf("securing %s: %w", p, err)
 		}
@@ -152,6 +161,20 @@ func reserveSQLiteFile(path string, create bool) error {
 	return nil
 }
 
+// missingOrEmpty reports whether opening path in SQLiteModeCreate
+// would start a new, empty database: the file is not there, or it is
+// zero-length, which SQLite opens as an empty database. A file left at
+// zero length by an interrupted first start or a truncated copy holds
+// as little as a missing one, and must be reported the same way.
+func missingOrEmpty(path string) bool {
+	info, err := os.Stat(path)
+	if errors.Is(err, fs.ErrNotExist) {
+		return true
+	}
+
+	return err == nil && info.Size() == 0
+}
+
 // sqliteSidecarPaths returns the files SQLite maintains beside a
 // database under WAL. They carry the same rows as the database itself,
 // so a fix that tightens only the main file has fixed nothing.
@@ -218,7 +241,19 @@ func SQLiteDSN(path, mode string) string {
 	)
 	q.Add("_pragma", "journal_mode(WAL)")
 
-	return "file:" + path + "?" + q.Encode()
+	// SQLite reads a file: URI's path only up to the first ? or #, and
+	// decodes %XX escapes in it, so url.URL escapes the path. OmitHost
+	// keeps a relative path as "file:data/x.db"; written
+	// "file://data/x.db", SQLite would reject "data" as the URI's
+	// authority.
+	dsn := url.URL{
+		Scheme:   "file",
+		Path:     path,
+		OmitHost: true,
+		RawQuery: q.Encode(),
+	}
+
+	return dsn.String()
 }
 
 // OpenSQLite opens the SQLite file at path with the service's

internal/database/sqlite_open_test.go

@@ -2,6 +2,8 @@ package database_test
 
 import (
 	"context"
+	"database/sql"
+	"os"
 	"path/filepath"
 	"strings"
 	"testing"
@@ -112,6 +114,79 @@ func TestMainDBAppliesPragmasOnALiveHandle(t *testing.T) {
 	assert.Equal(t, "10000", busy)
 }
 
+// TestOpenSQLiteOpensThePathAskedFor uses a data directory whose name
+// holds ?, # and %. SQLite reads a file: URI's path only up to a ? or
+// #, and decodes %41 to A, so an unescaped path sent every database in
+// that directory to one file outside it.
+func TestOpenSQLiteOpensThePathAskedFor(t *testing.T) {
+	t.Parallel()
+
+	dir := filepath.Join(t.TempDir(), "hooks#1?x%41")
+	require.NoError(t, os.Mkdir(dir, 0o700))
+
+	for _, name := range []string{
+		database.MainDBFileName,
+		"events-" + uuid.New().String() + ".db",
+	} {
+		path := filepath.Join(dir, name)
+
+		sqlDB, err := database.OpenSQLite(
+			path, database.SQLiteModeCreate,
+		)
+		require.NoError(t, err)
+
+		t.Cleanup(func() { assert.NoError(t, sqlDB.Close()) })
+
+		assertOpenedFile(t, sqlDB, path)
+	}
+}
+
+// TestOpenSQLiteOpensARelativePath opens a database at a path relative
+// to the working directory, which is what a relative DATA_DIR such as
+// the README's DATA_DIR=../webhooker-data gives OpenSQLite.
+//
+//nolint:paralleltest // t.Chdir moves the whole process.
+func TestOpenSQLiteOpensARelativePath(t *testing.T) {
+	t.Chdir(t.TempDir())
+	require.NoError(t, os.Mkdir("data", 0o700))
+
+	path := filepath.Join("data", database.MainDBFileName)
+
+	sqlDB, err := database.OpenSQLite(path, database.SQLiteModeCreate)
+	require.NoError(t, err)
+
+	defer func() { require.NoError(t, sqlDB.Close()) }()
+
+	assertOpenedFile(t, sqlDB, path)
+}
+
+// assertOpenedFile checks that sqlDB is the SQLite file at path.
+// database_list names the file SQLite actually opened, as an absolute
+// path with symlinks resolved, so the files are compared, not the names.
+func assertOpenedFile(t *testing.T, sqlDB *sql.DB, path string) {
+	t.Helper()
+
+	var (
+		seq          int
+		schema, file string
+	)
+
+	require.NoError(t, sqlDB.
+		QueryRowContext(t.Context(), "pragma database_list").
+		Scan(&seq, &schema, &file))
+
+	asked, err := os.Stat(path)
+	require.NoError(t, err)
+
+	opened, err := os.Stat(file)
+	require.NoError(t, err)
+
+	assert.True(
+		t, os.SameFile(asked, opened),
+		"asked for %s, SQLite opened %s", path, file,
+	)
+}
+
 // TestConcurrentReaderDoesNotBlockWrites is the unit-scale form of the
 // reproduction in
 // https://git.eeqj.de/sneak/webhooker/issues/256: an operator's

internal/database/testing.go

@@ -1,47 +0,0 @@
-package database
-
-import (
-	"log/slog"
-	"os"
-
-	"gorm.io/gorm"
-)
-
-// NewTestDatabase creates a Database wrapper around a pre-opened *gorm.DB.
-// Intended for use in tests that need a *database.Database without the
-// full fx lifecycle. The caller is responsible for closing the underlying
-// sql.DB connection.
-func NewTestDatabase(db *gorm.DB) *Database {
-	return &Database{
-		db: db,
-		log: slog.New(slog.NewTextHandler(
-			os.Stderr,
-			&slog.HandlerOptions{Level: slog.LevelDebug},
-		)),
-	}
-}
-
-// NewTestWebhookDBManager creates a WebhookDBManager backed by the given
-// data directory. Intended for use in tests without the fx lifecycle.
-func NewTestWebhookDBManager(dataDir string) *WebhookDBManager {
-	return NewTestWebhookDBManagerWithLogger(
-		dataDir,
-		slog.New(slog.NewTextHandler(
-			os.Stderr,
-			&slog.HandlerOptions{Level: slog.LevelDebug},
-		)),
-	)
-}
-
-// NewTestWebhookDBManagerWithLogger is NewTestWebhookDBManager with the
-// logger supplied by the caller. The per-webhook databases this manager
-// opens hand that logger to gormlog, so a test that needs to see the SQL
-// the service emits can capture it.
-func NewTestWebhookDBManagerWithLogger(
-	dataDir string, log *slog.Logger,
-) *WebhookDBManager {
-	return &WebhookDBManager{
-		dataDir: dataDir,
-		log:     log,
-	}
-}

internal/database/totals_test.go

@@ -0,0 +1,417 @@
+package database_test
+
+import (
+	"context"
+	"net/http"
+	"testing"
+	"time"
+
+	"github.com/google/uuid"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"gorm.io/gorm"
+	"sneak.berlin/go/webhooker/internal/database"
+)
+
+// readEventTotals reads a webhook database's row of event totals,
+// asserting that it has exactly one.
+func readEventTotals(t *testing.T, db *gorm.DB) database.EventTotals {
+	t.Helper()
+
+	var rows []database.EventTotals
+
+	require.NoError(t, db.Find(&rows).Error)
+	require.Len(t, rows, 1)
+
+	return rows[0]
+}
+
+// readTargetTotals reads a webhook database's target totals, keyed by
+// target.
+func readTargetTotals(
+	t *testing.T, db *gorm.DB,
+) map[string]database.TargetTotals {
+	t.Helper()
+
+	var rows []database.TargetTotals
+
+	require.NoError(t, db.Find(&rows).Error)
+
+	byTarget := make(map[string]database.TargetTotals, len(rows))
+	for _, row := range rows {
+		byTarget[row.TargetID] = row
+	}
+
+	return byTarget
+}
+
+// TestWebhookDBManager_TotalsSurviveReopen verifies that a new event
+// database starts with one row of zero event totals and no target
+// totals, that adding to a target twice adds to the one row, and that
+// opening the database again keeps everything added.
+func TestWebhookDBManager_TotalsSurviveReopen(t *testing.T) {
+	t.Parallel()
+
+	mgr, lc := setupTestWebhookDBManager(t)
+	ctx := context.Background()
+	require.NoError(t, lc.Start(ctx))
+
+	defer func() { require.NoError(t, lc.Stop(ctx)) }()
+
+	webhookID := uuid.New().String()
+
+	db, err := mgr.GetDB(webhookID)
+	require.NoError(t, err)
+
+	fresh := readEventTotals(t, db)
+	assert.Equal(t, database.EventTotals{ID: fresh.ID}, fresh)
+	assert.Empty(t, readTargetTotals(t, db))
+
+	first, second := uuid.New().String(), uuid.New().String()
+
+	require.NoError(t, database.AddEventTotals(db, database.EventTotals{
+		Events: 2,
+	}))
+	require.NoError(t, database.AddTargetTotals(db, database.TargetTotals{
+		TargetID: first, Deliveries: 2, Delivered: 1,
+	}))
+	require.NoError(t, database.AddTargetTotals(db, database.TargetTotals{
+		TargetID: first, Failed: 1,
+	}))
+	require.NoError(t, database.AddTargetTotals(db, database.TargetTotals{
+		TargetID: second, Deliveries: 1,
+	}))
+
+	// Drop the cached connection so the next open reopens the file,
+	// as a restart would.
+	require.NoError(t, mgr.CloseAll())
+
+	db, err = mgr.GetDB(webhookID)
+	require.NoError(t, err)
+
+	assert.Equal(t, database.EventTotals{ID: fresh.ID, Events: 2},
+		readEventTotals(t, db))
+	assert.Equal(t, map[string]database.TargetTotals{
+		first: {
+			TargetID: first, Deliveries: 2, Delivered: 1, Failed: 1,
+		},
+		second: {TargetID: second, Deliveries: 1},
+	}, readTargetTotals(t, db))
+}
+
+// seedExpiredEvents stores count events created at the given time,
+// each with a delivered delivery to one target and a failed delivery
+// to the other, and one attempt for each delivery.
+//
+// It and seedBareEvents insert 50 rows per statement, not more. The
+// SQLite driver looks up each parameter's value by scanning the
+// statement's arguments from the first until it reaches that
+// parameter's, so the time to bind a statement grows with the square of
+// its parameter count: at 500 rows, several thousand parameters, the
+// seeding took most of these tests' time under -race.
+func seedExpiredEvents(
+	t *testing.T,
+	db *gorm.DB,
+	webhookID string,
+	count int,
+	createdAt time.Time,
+	delivered, failed string,
+) {
+	t.Helper()
+
+	events := make([]database.Event, count)
+	deliveries := make([]database.Delivery, 0, 2*count)
+
+	for i := range events {
+		events[i] = database.Event{
+			WebhookID:    webhookID,
+			EntrypointID: uuid.New().String(),
+			Method:       http.MethodPost,
+		}
+		events[i].ID = uuid.New().String()
+		events[i].CreatedAt = createdAt
+
+		deliveries = append(deliveries,
+			database.Delivery{
+				EventID:  events[i].ID,
+				TargetID: delivered,
+				Status:   database.DeliveryStatusDelivered,
+			},
+			database.Delivery{
+				EventID:  events[i].ID,
+				TargetID: failed,
+				Status:   database.DeliveryStatusFailed,
+			},
+		)
+	}
+
+	require.NoError(t, db.CreateInBatches(events, 50).Error)
+	require.NoError(t, db.CreateInBatches(deliveries, 50).Error)
+
+	results := make([]database.DeliveryResult, len(deliveries))
+	for i := range deliveries {
+		results[i] = database.DeliveryResult{
+			DeliveryID: deliveries[i].ID, AttemptNum: 1,
+		}
+	}
+
+	require.NoError(t, db.CreateInBatches(results, 50).Error)
+}
+
+// seedBareEvents stores count events created at the given time, with
+// no deliveries.
+func seedBareEvents(
+	t *testing.T,
+	db *gorm.DB,
+	webhookID string,
+	count int,
+	createdAt time.Time,
+) {
+	t.Helper()
+
+	events := make([]database.Event, count)
+	for i := range events {
+		events[i] = database.Event{
+			WebhookID:    webhookID,
+			EntrypointID: uuid.New().String(),
+			Method:       http.MethodPost,
+		}
+		events[i].CreatedAt = createdAt
+	}
+
+	require.NoError(t, db.CreateInBatches(events, 50).Error)
+}
+
+// TestRetentionReaper_PrunesMoreThanOneBatch verifies that a prune
+// larger than one transaction's batch removes every expired event with
+// its deliveries and delivery results, keeps the recent event, and
+// adds what it removed to the event and target totals, so the totals
+// within retention match the rows still stored.
+func TestRetentionReaper_PrunesMoreThanOneBatch(t *testing.T) {
+	t.Parallel()
+
+	env := setupRetentionTest(t)
+
+	webhookID := createWebhook(t, env.mainDB.DB(), 30)
+
+	db, err := env.mgr.GetDB(webhookID)
+	require.NoError(t, err)
+
+	expired := database.ExportReapBatchSize + 1
+	delivered, failed := uuid.New().String(), uuid.New().String()
+	seedExpiredEvents(t, db, webhookID, expired,
+		time.Now().Add(-40*24*time.Hour), delivered, failed)
+
+	// One recent event, delivered to the first target.
+	recent := seedEventChain(t, db, webhookID, time.Now())
+	require.NoError(t, db.Model(&database.Delivery{}).
+		Where("id = ?", recent.deliveryID).
+		Update("target_id", delivered).Error)
+
+	// The totals storing those rows would have left.
+	n := int64(expired)
+	require.NoError(t, database.AddEventTotals(db, database.EventTotals{
+		Events: n + 1,
+	}))
+	require.NoError(t, database.AddTargetTotals(db, database.TargetTotals{
+		TargetID: delivered, Deliveries: n + 1, Delivered: n + 1,
+	}))
+	require.NoError(t, database.AddTargetTotals(db, database.TargetTotals{
+		TargetID: failed, Deliveries: n, Failed: n,
+	}))
+
+	env.reaper.ExportSweep(context.Background())
+
+	// Only the recent event's rows are left.
+	for _, model := range []any{
+		&database.Event{}, &database.Delivery{}, &database.DeliveryResult{},
+	} {
+		var count int64
+
+		require.NoError(t, db.Model(model).Count(&count).Error)
+		assert.Equal(t, int64(1), count, "%T rows left", model)
+	}
+
+	assertChainPresent(t, db, recent)
+
+	eventTotals := readEventTotals(t, db)
+	assert.Equal(t, database.EventTotals{
+		ID: eventTotals.ID, Events: n + 1, EventsRemoved: n,
+	}, eventTotals)
+
+	targetTotals := readTargetTotals(t, db)
+	assert.Equal(t, map[string]database.TargetTotals{
+		delivered: {
+			TargetID: delivered, Deliveries: n + 1, Delivered: n + 1,
+			DeliveriesRemoved: n,
+		},
+		failed: {
+			TargetID: failed, Deliveries: n, Failed: n,
+			DeliveriesRemoved: n, FailedRemoved: n,
+		},
+	}, targetTotals)
+
+	// A sweep with nothing left to remove changes nothing.
+	env.reaper.ExportSweep(context.Background())
+
+	assert.Equal(t, eventTotals, readEventTotals(t, db))
+	assert.Equal(t, targetTotals, readTargetTotals(t, db))
+}
+
+// TestRetentionReaper_WriteDuringPruneSucceeds verifies that a prune
+// of several batches lets other writers in between its batches: an
+// event stored once the first batch is deleted is stored while expired
+// events are still left, not only after the prune has finished.
+func TestRetentionReaper_WriteDuringPruneSucceeds(t *testing.T) {
+	t.Parallel()
+
+	env := setupRetentionTest(t)
+
+	webhookID := createWebhook(t, env.mainDB.DB(), 30)
+
+	db, err := env.mgr.GetDB(webhookID)
+	require.NoError(t, err)
+
+	// Three batches of expired events, with nothing else stored: only
+	// the number of batches matters here.
+	expired := 3 * database.ExportReapBatchSize
+	seedBareEvents(t, db, webhookID, expired,
+		time.Now().Add(-40*24*time.Hour))
+
+	cutoff := time.Now().Add(-30 * 24 * time.Hour)
+	countExpired := func() int64 {
+		var count int64
+
+		require.NoError(t, db.Model(&database.Event{}).
+			Where("created_at < ?", cutoff).
+			Count(&count).Error)
+
+		return count
+	}
+
+	pruned := make(chan struct{})
+
+	go func() {
+		defer close(pruned)
+
+		env.reaper.ExportSweep(context.Background())
+	}()
+
+	t.Cleanup(func() { <-pruned })
+
+	// Every stored event is expired until the write below.
+	require.Eventually(t, func() bool {
+		var count int64
+
+		err := db.Model(&database.Event{}).Count(&count).Error
+
+		return err == nil && count < int64(expired)
+	}, 10*time.Second, 10*time.Millisecond)
+
+	event := &database.Event{
+		WebhookID:    webhookID,
+		EntrypointID: uuid.New().String(),
+		Method:       http.MethodPost,
+	}
+	require.NoError(t, db.Create(event).Error)
+
+	assert.Positive(t, countExpired(),
+		"the event was stored only after the whole prune")
+
+	<-pruned
+
+	assert.Zero(t, countExpired())
+
+	var stored database.Event
+
+	require.NoError(t, db.First(&stored, "id = ?", event.ID).Error)
+}
+
+// TestRetentionReaper_StopDuringPruneLeavesTheRest verifies that
+// stopping the reaper during a prune of several batches returns
+// between two batches, well inside the stop timeout, leaving the
+// remaining expired events for the next sweep, and that the totals
+// match the rows left.
+func TestRetentionReaper_StopDuringPruneLeavesTheRest(t *testing.T) {
+	t.Parallel()
+
+	env := setupRetentionTest(t)
+
+	webhookID := createWebhook(t, env.mainDB.DB(), 30)
+
+	db, err := env.mgr.GetDB(webhookID)
+	require.NoError(t, err)
+
+	// Two batches and one more of expired events, a few of them with a
+	// delivered and a failed delivery for the target totals to count.
+	// Most carry nothing else, to keep the test quick.
+	const withDeliveries = 10
+
+	expiredAt := time.Now().Add(-40 * 24 * time.Hour)
+	delivered, failed := uuid.New().String(), uuid.New().String()
+	seedExpiredEvents(t, db, webhookID, withDeliveries, expiredAt,
+		delivered, failed)
+	seedBareEvents(t, db, webhookID,
+		2*database.ExportReapBatchSize+1-withDeliveries, expiredAt)
+
+	n := int64(2*database.ExportReapBatchSize + 1)
+	require.NoError(t, database.AddEventTotals(db, database.EventTotals{
+		Events: n,
+	}))
+	require.NoError(t, database.AddTargetTotals(db, database.TargetTotals{
+		TargetID: delivered, Deliveries: withDeliveries,
+		Delivered: withDeliveries,
+	}))
+	require.NoError(t, database.AddTargetTotals(db, database.TargetTotals{
+		TargetID: failed, Deliveries: withDeliveries,
+		Failed: withDeliveries,
+	}))
+
+	env.reaper.ExportSetInterval(time.Millisecond)
+	env.reaper.ExportStart()
+
+	// Stop once the first batch is deleted. The stop lands in the pause
+	// after it, or at worst during the second batch, so at least the
+	// last event is left.
+	require.Eventually(t, func() bool {
+		var count int64
+
+		err := db.Model(&database.Event{}).Count(&count).Error
+
+		return err == nil && count < n
+	}, 10*time.Second, 10*time.Millisecond)
+
+	// The app's stop timeout.
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+
+	require.NoError(t, env.reaper.ExportStop(ctx))
+
+	var events int64
+
+	require.NoError(t, db.Model(&database.Event{}).Count(&events).Error)
+	assert.Positive(t, events, "the stop waited for the whole prune")
+
+	eventTotals := readEventTotals(t, db)
+	assert.Equal(t, events, eventTotals.Events-eventTotals.EventsRemoved)
+
+	targetTotals := readTargetTotals(t, db)
+	require.Len(t, targetTotals, 2)
+
+	for target, totals := range targetTotals {
+		var deliveries, failures int64
+
+		require.NoError(t, db.Model(&database.Delivery{}).
+			Where("target_id = ?", target).
+			Count(&deliveries).Error)
+		require.NoError(t, db.Model(&database.Delivery{}).
+			Where("target_id = ? AND status = ?",
+				target, database.DeliveryStatusFailed).
+			Count(&failures).Error)
+
+		assert.Equal(t, deliveries,
+			totals.Deliveries-totals.DeliveriesRemoved, target)
+		assert.Equal(t, failures, totals.Failed-totals.FailedRemoved,
+			target)
+	}
+}

internal/database/webhook_db_manager.go

@@ -14,8 +14,8 @@ import (
 	"gorm.io/gorm"
 	"sneak.berlin/go/webhooker/internal/config"
 	"sneak.berlin/go/webhooker/internal/datadir"
+	"sneak.berlin/go/webhooker/internal/db/migrations"
 	"sneak.berlin/go/webhooker/internal/gormlog"
-	"sneak.berlin/go/webhooker/internal/logger"
 )
 
 // WebhookDBManagerParams holds the fx dependencies for
@@ -24,7 +24,7 @@ type WebhookDBManagerParams struct {
 	fx.In
 
 	Config *config.Config
-	Logger *logger.Logger
+	Logger *slog.Logger
 }
 
 // errInvalidCachedDBType indicates a type assertion failure
@@ -33,9 +33,23 @@ var errInvalidCachedDBType = errors.New(
 	"invalid cached database type",
 )
 
+// ErrEventDBNotRemoved is in DeleteDB's error when the event
+// database file itself could not be removed: it is still on disk.
+var ErrEventDBNotRemoved = errors.New(
+	"event database file not removed",
+)
+
+// ErrSidecarNotRemoved is in DeleteDB's error when the event
+// database file was removed, so its events are gone, but its -wal
+// or -shm sidecar could not be.
+var ErrSidecarNotRemoved = errors.New(
+	"event database file removed, but a -wal or -shm sidecar was not",
+)
+
 // WebhookDBManager manages per-webhook SQLite database files
 // for event storage. Each webhook gets its own dedicated
-// database containing Events, Deliveries, and DeliveryResults.
+// database containing Events, Deliveries, DeliveryResults and the
+// running totals of them (EventTotals, TargetTotals, EntrypointTotals).
 // Database connections are opened lazily and cached.
 type WebhookDBManager struct {
 	dataDir string
@@ -56,7 +70,7 @@ func NewWebhookDBManager(
 ) (*WebhookDBManager, error) {
 	m := &WebhookDBManager{
 		dataDir: params.Config.DataDir,
-		log:     params.Logger.Get(),
+		log:     params.Logger,
 	}
 
 	// Create data directory if it doesn't exist. datadir.DirPerm is the
@@ -84,34 +98,37 @@ func NewWebhookDBManager(
 	return m, nil
 }
 
-// GetDB returns the database connection for a webhook,
-// creating the database file lazily if it doesn't exist.
+// GetDB returns the database connection for a webhook, opening it on
+// first use.
+//
+// The file is made by CreateDB when the webhook is created. One that is
+// missing or zero-length here means the webhook's events and pending
+// deliveries are gone: an empty database is created in its place so
+// the webhook keeps receiving, and that is logged as a warning naming
+// the file, as a new main database is.
 func (m *WebhookDBManager) GetDB(
 	webhookID string,
 ) (*gorm.DB, error) {
-	// Fast path: already open
-	if val, ok := m.dbs.Load(webhookID); ok {
-		return asGormDB(val, webhookID)
-	}
+	return m.getDB(webhookID, false)
+}
 
-	// Slow path: open the database under the lock, looking in the
-	// cache again first. A caller that raced another one here then
-	// waits for its handle instead of opening a second one.
+// GetDBIf is GetDB, done only when check reports true. check runs under
+// the lock DeleteDB holds while it removes the files, so a caller can
+// confirm the webhook still exists and open its database with no delete
+// in between. The handle is nil when check reports false. check must
+// not call the manager.
+func (m *WebhookDBManager) GetDBIf(
+	webhookID string, check func() (bool, error),
+) (*gorm.DB, error) {
 	m.mu.Lock()
 	defer m.mu.Unlock()
 
-	if val, ok := m.dbs.Load(webhookID); ok {
-		return asGormDB(val, webhookID)
-	}
-
-	db, err := m.openDB(webhookID)
-	if err != nil {
+	ok, err := check()
+	if err != nil || !ok {
 		return nil, err
 	}
 
-	m.dbs.Store(webhookID, db)
-
-	return db, nil
+	return m.getDBLocked(webhookID, false)
 }
 
 // asGormDB returns a value read from the cache as the database
@@ -129,12 +146,12 @@ func asGormDB(val any, webhookID string) (*gorm.DB, error) {
 	return db, nil
 }
 
-// CreateDB explicitly creates a new per-webhook database file
-// and runs migrations.
+// CreateDB creates a new webhook's database file and runs
+// migrations.
 func (m *WebhookDBManager) CreateDB(
 	webhookID string,
 ) error {
-	_, err := m.GetDB(webhookID)
+	_, err := m.getDB(webhookID, true)
 
 	return err
 }
@@ -150,7 +167,10 @@ func (m *WebhookDBManager) DBExists(
 }
 
 // DeleteDB closes the connection and deletes the database file
-// for a webhook. The file is permanently removed.
+// for a webhook, with its -wal and -shm sidecars. The files are
+// permanently removed. Each file is tried even when another could
+// not be removed, and the error wraps ErrEventDBNotRemoved or
+// ErrSidecarNotRemoved to say which was left, naming each file.
 func (m *WebhookDBManager) DeleteDB(
 	webhookID string,
 ) error {
@@ -169,16 +189,23 @@ func (m *WebhookDBManager) DeleteDB(
 		}
 	}
 
-	// Delete the main DB file and WAL/SHM files
 	path := m.dbPath(webhookID)
-	for _, suffix := range []string{"", "-wal", "-shm"} {
-		err := os.Remove(path + suffix)
-		if err != nil && !os.IsNotExist(err) {
-			return fmt.Errorf(
-				"deleting webhook database file %s%s: %w",
-				path, suffix, err,
-			)
-		}
+
+	dbErr := removeFile(path)
+	sidecarErr := errors.Join(
+		removeFile(path+"-wal"),
+		removeFile(path+"-shm"),
+	)
+
+	if dbErr != nil {
+		return fmt.Errorf(
+			"%w: %w",
+			ErrEventDBNotRemoved, errors.Join(dbErr, sidecarErr),
+		)
+	}
+
+	if sidecarErr != nil {
+		return fmt.Errorf("%w: %w", ErrSidecarNotRemoved, sidecarErr)
 	}
 
 	m.log.Info(
@@ -189,6 +216,17 @@ func (m *WebhookDBManager) DeleteDB(
 	return nil
 }
 
+// removeFile removes path. A file that is already gone counts as
+// removed; the error from any other failure names the file.
+func removeFile(path string) error {
+	err := os.Remove(path)
+	if errors.Is(err, os.ErrNotExist) {
+		return nil
+	}
+
+	return err
+}
+
 // CloseAll closes all open per-webhook database connections.
 // Called during application shutdown.
 func (m *WebhookDBManager) CloseAll() error {
@@ -231,6 +269,54 @@ func (m *WebhookDBManager) DBPath(
 	return m.dbPath(webhookID)
 }
 
+// getDB is GetDB, and CreateDB when isNew is true: the webhook has just
+// been created, so a missing file is expected rather than lost.
+func (m *WebhookDBManager) getDB(
+	webhookID string, isNew bool,
+) (*gorm.DB, error) {
+	// Fast path: already open
+	if val, ok := m.dbs.Load(webhookID); ok {
+		return asGormDB(val, webhookID)
+	}
+
+	m.mu.Lock()
+	defer m.mu.Unlock()
+
+	return m.getDBLocked(webhookID, isNew)
+}
+
+// getDBLocked is getDB's slow path, run with m.mu held. It looks in the
+// cache again first: a caller that raced another one to the lock then
+// gets its handle instead of opening a second one.
+func (m *WebhookDBManager) getDBLocked(
+	webhookID string, isNew bool,
+) (*gorm.DB, error) {
+	if val, ok := m.dbs.Load(webhookID); ok {
+		return asGormDB(val, webhookID)
+	}
+
+	// Checked before opening, which creates the file. See GetDB.
+	path := m.dbPath(webhookID)
+	replaced := !isNew && missingOrEmpty(path)
+
+	db, err := m.openDB(webhookID)
+	if err != nil {
+		return nil, err
+	}
+
+	if replaced {
+		m.log.Warn(
+			"created a new, empty database",
+			"webhook_id", webhookID,
+			"path", path,
+		)
+	}
+
+	m.dbs.Store(webhookID, db)
+
+	return db, nil
+}
+
 func (m *WebhookDBManager) dbPath(
 	webhookID string,
 ) string {
@@ -285,17 +371,7 @@ func (m *WebhookDBManager) openDB(
 		)
 	}
 
-	err = purgeTargetRows(db, m.log, webhookID)
-	if err != nil {
-		_ = sqlDB.Close()
-
-		return nil, err
-	}
-
-	// Run migrations for event-tier models only
-	err = db.AutoMigrate(
-		&Event{}, &Delivery{}, &DeliveryResult{},
-	)
+	err = migrations.Apply(db, migrations.Events)
 	if err != nil {
 		_ = sqlDB.Close()
 
@@ -305,6 +381,19 @@ func (m *WebhookDBManager) openDB(
 		)
 	}
 
+	err = parseModels(
+		db, &Event{}, &Delivery{}, &DeliveryResult{},
+		&EventTotals{}, &TargetTotals{}, &EntrypointTotals{},
+	)
+	if err != nil {
+		_ = sqlDB.Close()
+
+		return nil, fmt.Errorf(
+			"connecting to webhook database %s: %w",
+			webhookID, err,
+		)
+	}
+
 	m.log.Info(
 		"opened per-webhook database",
 		"webhook_id", webhookID,

internal/database/webhook_db_manager_test.go

@@ -18,6 +18,7 @@ import (
 	"gorm.io/gorm"
 	"sneak.berlin/go/webhooker/internal/config"
 	"sneak.berlin/go/webhooker/internal/database"
+	"sneak.berlin/go/webhooker/internal/database/databasetest"
 	"sneak.berlin/go/webhooker/internal/globals"
 	"sneak.berlin/go/webhooker/internal/logger"
 )
@@ -50,7 +51,7 @@ func setupTestWebhookDBManager(
 		lc,
 		database.WebhookDBManagerParams{
 			Config: cfg,
-			Logger: l,
+			Logger: l.Get(),
 		},
 	)
 	require.NoError(t, err)
@@ -117,7 +118,8 @@ func TestWebhookDBManager_ConcurrentFirstTouchOpensOnce(t *testing.T) {
 
 	var logs bytes.Buffer
 
-	mgr := database.NewTestWebhookDBManagerWithLogger(
+	mgr := databasetest.NewWebhookDBManagerWithLogger(
+		t,
 		t.TempDir(),
 		slog.New(slog.NewTextHandler(&logs, nil)),
 	)
@@ -182,17 +184,91 @@ func TestWebhookDBManager_DeleteDB(t *testing.T) {
 	}
 	require.NoError(t, db.Create(event).Error)
 
+	// Under WAL, an open database that has been written to has both
+	// sidecars beside it.
+	dbPath := mgr.DBPath(webhookID)
+	require.FileExists(t, dbPath+"-wal")
+	require.FileExists(t, dbPath+"-shm")
+
 	// Delete the DB
 	require.NoError(t, mgr.DeleteDB(webhookID))
 
 	// File should no longer exist
 	assert.False(t, mgr.DBExists(webhookID))
 
-	// Verify the file is actually gone from disk
+	// Verify the files are actually gone from disk
+	assert.NoFileExists(t, dbPath)
+	assert.NoFileExists(t, dbPath+"-wal")
+	assert.NoFileExists(t, dbPath+"-shm")
+}
+
+// blockRemoval puts a non-empty directory at path, which os.Remove
+// cannot remove whoever runs the test, root included.
+func blockRemoval(t *testing.T, path string) {
+	t.Helper()
+
+	require.NoError(t, os.MkdirAll(filepath.Join(path, "keep"), 0o700))
+}
+
+// TestWebhookDBManager_DeleteDBKeepsDatabaseFile proves that when the
+// event database file cannot be removed, the error says so, and both
+// sidecars are still removed.
+func TestWebhookDBManager_DeleteDBKeepsDatabaseFile(t *testing.T) {
+	t.Parallel()
+
+	mgr, lc := setupTestWebhookDBManager(t)
+	ctx := context.Background()
+	require.NoError(t, lc.Start(ctx))
+
+	defer func() { require.NoError(t, lc.Stop(ctx)) }()
+
+	webhookID := uuid.New().String()
 	dbPath := mgr.DBPath(webhookID)
 
-	_, err = os.Stat(dbPath)
-	assert.True(t, os.IsNotExist(err))
+	blockRemoval(t, dbPath)
+	require.NoError(t, os.WriteFile(dbPath+"-wal", nil, 0o600))
+	require.NoError(t, os.WriteFile(dbPath+"-shm", nil, 0o600))
+
+	err := mgr.DeleteDB(webhookID)
+
+	require.ErrorIs(t, err, database.ErrEventDBNotRemoved)
+	require.NotErrorIs(t, err, database.ErrSidecarNotRemoved)
+	assert.Contains(t, err.Error(), dbPath)
+	assert.NoFileExists(t, dbPath+"-wal")
+	assert.NoFileExists(t, dbPath+"-shm")
+}
+
+// TestWebhookDBManager_DeleteDBKeepsSidecar proves that when the
+// event database file is removed but a sidecar is not, the error
+// says the database file is gone, and the other sidecar is still
+// removed.
+func TestWebhookDBManager_DeleteDBKeepsSidecar(t *testing.T) {
+	t.Parallel()
+
+	mgr, lc := setupTestWebhookDBManager(t)
+	ctx := context.Background()
+	require.NoError(t, lc.Start(ctx))
+
+	defer func() { require.NoError(t, lc.Stop(ctx)) }()
+
+	webhookID := uuid.New().String()
+	dbPath := mgr.DBPath(webhookID)
+
+	require.NoError(t, mgr.CreateDB(webhookID))
+	// Closing removes the sidecars, so the ones below are the only
+	// ones there.
+	require.NoError(t, mgr.CloseAll())
+
+	blockRemoval(t, dbPath+"-wal")
+	require.NoError(t, os.WriteFile(dbPath+"-shm", nil, 0o600))
+
+	err := mgr.DeleteDB(webhookID)
+
+	require.ErrorIs(t, err, database.ErrSidecarNotRemoved)
+	require.NotErrorIs(t, err, database.ErrEventDBNotRemoved)
+	assert.Contains(t, err.Error(), dbPath+"-wal")
+	assert.NoFileExists(t, dbPath)
+	assert.NoFileExists(t, dbPath+"-shm")
 }
 
 func TestWebhookDBManager_LazyCreation(t *testing.T) {
@@ -215,6 +291,76 @@ func TestWebhookDBManager_LazyCreation(t *testing.T) {
 	assert.True(t, mgr.DBExists(webhookID))
 }
 
+// A webhook's database is made by CreateDB along with the webhook. One
+// that GetDB finds missing or zero-length has lost the webhook's events
+// and pending deliveries, so the empty database made in its place is
+// logged as a warning naming the file
+// (https://git.eeqj.de/sneak/webhooker/issues/290). CreateDB, and
+// reopening a database that is there, log no such warning.
+func TestWebhookDBManager_LostDatabaseIsLogged(t *testing.T) {
+	t.Parallel()
+
+	const created = `level=WARN msg="created a new, empty database"`
+
+	open := func(
+		t *testing.T, prepare func(*database.WebhookDBManager, string),
+	) (string, string) {
+		t.Helper()
+
+		var logs bytes.Buffer
+
+		mgr := databasetest.NewWebhookDBManagerWithLogger(
+			t,
+			t.TempDir(),
+			slog.New(slog.NewTextHandler(&logs, nil)),
+		)
+
+		webhookID := uuid.New().String()
+		prepare(mgr, webhookID)
+
+		_, err := mgr.GetDB(webhookID)
+		require.NoError(t, err)
+		require.NoError(t, mgr.CloseAll())
+
+		return logs.String(),
+			" webhook_id=" + webhookID + " path=" + mgr.DBPath(webhookID)
+	}
+
+	t.Run("missing", func(t *testing.T) {
+		t.Parallel()
+
+		logs, fields := open(
+			t, func(*database.WebhookDBManager, string) {},
+		)
+		assert.Contains(t, logs, created+fields)
+	})
+
+	t.Run("zero-length", func(t *testing.T) {
+		t.Parallel()
+
+		logs, fields := open(
+			t, func(mgr *database.WebhookDBManager, webhookID string) {
+				require.NoError(t, os.WriteFile(
+					mgr.DBPath(webhookID), nil, database.SQLiteFilePerm,
+				))
+			},
+		)
+		assert.Contains(t, logs, created+fields)
+	})
+
+	t.Run("created with the webhook, then reopened", func(t *testing.T) {
+		t.Parallel()
+
+		logs, _ := open(
+			t, func(mgr *database.WebhookDBManager, webhookID string) {
+				require.NoError(t, mgr.CreateDB(webhookID))
+				require.NoError(t, mgr.CloseAll())
+			},
+		)
+		assert.NotContains(t, logs, created)
+	})
+}
+
 func TestWebhookDBManager_DeliveryWorkflow(t *testing.T) {
 	t.Parallel()