Stop snapshot purge when a snapshot cannot leave the local index (closes #337)

snapshot purge, and the purge step of snapshot create --prune, logged a
failed local index delete, then deleted that snapshot's metadata from
the destination store anyway and exited 0, leaving a local record whose
metadata was gone. It now stops at that snapshot with an error, before
touching its metadata, so the command exits non-zero; snapshots already
purged stay purged. This matches how the same loop already handles a
failed metadata delete.

Model: opus-5-5

README.md

@@ -339,7 +339,9 @@ local index alone, and still exits zero.
 
 **`snapshot purge`**: Remove old snapshots based on criteria. Retention is
 per-snapshot-name (`--keep-latest` keeps the latest of each name, not the
-latest globally).
+latest globally). If a snapshot cannot be removed from the local index, the
+command stops there, leaves that snapshot's metadata on the destination store
+and exits non-zero; `snapshot create --prune` fails the same way.
 * `--keep-latest`: Keep only the most recent snapshot of each name
 * `--older-than <duration>`: Remove snapshots older than duration (e.g. `30d`,
   `4w`, `6mo`, `1y`; `m` is minutes, `mo` is months)

internal/cli/entry_snapshot_purge_test.go

@@ -0,0 +1,90 @@
+package cli //nolint:testpackage // shares runEntry and the hermetic config helpers
+
+import (
+	"context"
+	"database/sql"
+	"os"
+	"path/filepath"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/vaultik/internal/database"
+	"sneak.berlin/go/vaultik/internal/snapshot"
+	"sneak.berlin/go/vaultik/internal/types"
+)
+
+// The `snapshot purge` command word and flags used below.
+const (
+	cmdPurge       = "purge"
+	flagKeepLatest = "--keep-latest"
+	flagForce      = "--force"
+)
+
+// TestEntrySnapshotPurgeKeepsMetadataWhenLocalIndexDeleteFails makes
+// removing a snapshot from the local index fail and checks that
+// `snapshot purge` exits non-zero and leaves that snapshot's metadata on
+// the destination store. It used to log the error, delete the metadata
+// anyway and exit 0.
+//
+//nolint:paralleltest // replaces os.Args, os.Stdout, os.Stderr and the xdg globals
+func TestEntrySnapshotPurgeKeepsMetadataWhenLocalIndexDeleteFails(t *testing.T) {
+	ctx := context.Background()
+	configPath := writeHermeticPruneConfig(t, false)
+
+	// writeHermeticPruneConfig puts the index and the store directory
+	// beside the config file.
+	dir := filepath.Dir(configPath)
+
+	// Two snapshots of the same name: --keep-latest purges the older.
+	olderStart := time.Date(2026, 5, 1, 9, 0, 0, 0, time.UTC)
+	newerStart := olderStart.Add(time.Hour)
+	olderID := "test-host_test_" + olderStart.Format(time.RFC3339)
+	newerID := "test-host_test_" + newerStart.Format(time.RFC3339)
+
+	db, err := database.New(ctx, filepath.Join(dir, "index.sqlite"))
+	require.NoError(t, err)
+
+	repos := database.NewRepositories(db)
+
+	manifestPaths := make(map[string]string)
+
+	for id, startedAt := range map[string]time.Time{
+		olderID: olderStart,
+		newerID: newerStart,
+	} {
+		completedAt := startedAt.Add(time.Minute)
+		snap := &database.Snapshot{
+			ID:             types.SnapshotID(id),
+			Hostname:       "test-host",
+			VaultikVersion: "test",
+			StartedAt:      startedAt,
+			CompletedAt:    &completedAt,
+		}
+
+		err = repos.WithTx(ctx, func(ctx context.Context, tx *sql.Tx) error {
+			return repos.Snapshots.Create(ctx, tx, snap)
+		})
+		require.NoError(t, err)
+
+		manifestPaths[id] = filepath.Join(dir, "store", "metadata",
+			snapshot.RemoteSnapshotKey(id), "manifest.json.zst")
+		require.NoError(t, os.MkdirAll(filepath.Dir(manifestPaths[id]), 0o750))
+		require.NoError(t, os.WriteFile(manifestPaths[id], []byte("stub"), 0o600))
+	}
+
+	_, err = db.Conn().ExecContext(ctx, `
+		CREATE TRIGGER refuse_snapshot_delete BEFORE DELETE ON snapshots
+		BEGIN SELECT RAISE(ABORT, 'simulated index error'); END`)
+	require.NoError(t, err)
+	require.NoError(t, db.Close())
+
+	code, _, stderr := runEntry(t,
+		flagConfig, configPath, cmdSnapshot, cmdPurge, flagKeepLatest, flagForce)
+
+	assert.Equal(t, 1, code)
+	assert.Contains(t, stderr, "simulated index error")
+	assert.FileExists(t, manifestPaths[olderID],
+		"a snapshot still in the local index keeps its metadata")
+}

internal/vaultik/deep_verify_test.go

@@ -1,7 +1,9 @@
 package vaultik_test
 
 import (
+	"bytes"
 	"context"
+	"crypto/rand"
 	"io"
 	"os"
 	"path/filepath"
@@ -9,7 +11,9 @@ import (
 
 	"github.com/spf13/afero"
 	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/vaultik/internal/config"
 	"sneak.berlin/go/vaultik/internal/log"
+	"sneak.berlin/go/vaultik/internal/storage"
 	"sneak.berlin/go/vaultik/internal/ui"
 	"sneak.berlin/go/vaultik/internal/vaultik"
 )
@@ -77,6 +81,174 @@ func TestDeepVerifyAcceptsHealthyAndRejectsCorruptBlob(t *testing.T) {
 		"deep verify should fail on a corrupted blob")
 }
 
+// TestDeepVerifyJSONCountsBlobs checks the verified and missing counts
+// that deep verify reports under --json: on a healthy snapshot, with a
+// damaged blob, and with missing blobs. Deep verify stops at the first
+// damaged blob, so the blobs read before it are the ones verified; when
+// blobs are missing it reads no blob contents, but counts every missing
+// blob.
+func TestDeepVerifyJSONCountsBlobs(t *testing.T) {
+	log.Initialize(log.Config{})
+	t.Parallel()
+
+	fs := afero.NewOsFs()
+	ctx := context.Background()
+	storeDir := filepath.Join(t.TempDir(), "remote")
+
+	cfg, storer, snapshotID, keys := backUpSeveralBlobs(ctx, t, storeDir)
+
+	healthy, err := deepVerifyJSON(ctx, t, cfg, storer, snapshotID)
+	require.NoError(t, err)
+	require.Equal(t, "ok", healthy.Status)
+	require.Len(t, keys, healthy.BlobCount)
+	require.Equal(t, healthy.BlobCount, healthy.Verified)
+	require.Zero(t, healthy.Missing)
+
+	// Damage the last blob read, keeping its length, so every blob before
+	// it is read and passes first.
+	lastPath := filepath.Join(storeDir, keys[len(keys)-1])
+	blob, err := afero.ReadFile(fs, lastPath)
+	require.NoError(t, err)
+
+	blob[len(blob)/2] ^= 0xff
+	require.NoError(t, afero.WriteFile(fs, lastPath, blob, 0o600))
+
+	got, err := deepVerifyJSON(ctx, t, cfg, storer, snapshotID)
+	require.Error(t, err)
+	require.Equal(t, "failed", got.Status)
+	require.Equal(t, healthy.BlobCount-1, got.Verified)
+	require.Zero(t, got.Missing)
+
+	// Delete the first blob and the damaged last one: both must be
+	// counted as missing, not only the first one found.
+	var deletedSize int64
+
+	for _, key := range []string{keys[0], keys[len(keys)-1]} {
+		info, statErr := storer.Stat(ctx, key)
+		require.NoError(t, statErr)
+
+		deletedSize += info.Size
+
+		require.NoError(t, storer.Delete(ctx, key))
+	}
+
+	got, err = deepVerifyJSON(ctx, t, cfg, storer, snapshotID)
+	require.Error(t, err)
+	require.Equal(t, "failed", got.Status)
+	require.Equal(t, 2, got.Missing)
+	require.Equal(t, deletedSize, got.MissingSize)
+	require.Zero(t, got.Verified)
+}
+
+// TestDeepVerifyJSONCountsWrongSizeBlob checks that under --deep --json a
+// blob of the wrong size is counted in mismatched and fails the run, and
+// that the check goes on past it to count a missing blob after it.
+func TestDeepVerifyJSONCountsWrongSizeBlob(t *testing.T) {
+	log.Initialize(log.Config{})
+	t.Parallel()
+
+	fs := afero.NewOsFs()
+	ctx := context.Background()
+	storeDir := filepath.Join(t.TempDir(), "remote")
+
+	cfg, storer, snapshotID, keys := backUpSeveralBlobs(ctx, t, storeDir)
+
+	// Make the first blob checked longer, and delete the last one.
+	firstPath := filepath.Join(storeDir, keys[0])
+	blob, err := afero.ReadFile(fs, firstPath)
+	require.NoError(t, err)
+
+	blob = append(blob, []byte("extra")...)
+	require.NoError(t, afero.WriteFile(fs, firstPath, blob, 0o600))
+	require.NoError(t, storer.Delete(ctx, keys[len(keys)-1]))
+
+	got, err := deepVerifyJSON(ctx, t, cfg, storer, snapshotID)
+	require.Error(t, err)
+	require.Equal(t, "failed", got.Status)
+	require.Equal(t, 1, got.Mismatched)
+	require.Equal(t, 1, got.Missing)
+	require.Zero(t, got.Verified)
+}
+
+// TestDeepVerifyStopsOnBlobThatCannotBeChecked checks that a blob that is
+// present in storage but cannot be checked is not counted as missing:
+// deep verify stops and fails with the storage error.
+func TestDeepVerifyStopsOnBlobThatCannotBeChecked(t *testing.T) {
+	log.Initialize(log.Config{})
+	t.Parallel()
+
+	ctx := context.Background()
+	storeDir := filepath.Join(t.TempDir(), "remote")
+
+	cfg, storer, snapshotID, keys := backUpSeveralBlobs(ctx, t, storeDir)
+	failing := statFailingStorer{Storer: storer, failKeys: []string{keys[0]}}
+
+	got, err := deepVerifyJSON(ctx, t, cfg, failing, snapshotID)
+	require.Error(t, err)
+	require.Equal(t, "failed", got.Status)
+	require.Zero(t, got.Missing)
+	require.Contains(t, got.ErrorMessage, errStatDenied.Error())
+}
+
+// backUpSeveralBlobs backs up random data, which does not compress or
+// deduplicate, so it fills at least three blobs in a store on disk at
+// storeDir. It returns the blobs' storage keys sorted, which is hash
+// order, the order deep verify reads them in.
+func backUpSeveralBlobs(
+	ctx context.Context, t *testing.T, storeDir string,
+) (*config.Config, *storage.FileStorer, string, []string) {
+	t.Helper()
+
+	fs := afero.NewOsFs()
+	tempDir := t.TempDir()
+	dataDir := filepath.Join(tempDir, "source")
+	dbPath := filepath.Join(tempDir, "index.sqlite")
+
+	chunkSize := int64(32 * 1024)
+	maxBlobSize := int64(128 * 1024)
+
+	data := make([]byte, maxBlobSize*4)
+	_, err := rand.Read(data)
+	require.NoError(t, err)
+	require.NoError(t, fs.MkdirAll(dataDir, 0o755))
+	require.NoError(t, afero.WriteFile(fs,
+		filepath.Join(dataDir, "data.bin"), data, 0o644))
+
+	cfg, storer, snapshotID := runFileStorageBackup(
+		ctx, t, fs, dataDir, storeDir, dbPath, chunkSize, maxBlobSize)
+
+	keys := storedBlobKeys(t, storeDir)
+	require.GreaterOrEqual(t, len(keys), 3)
+
+	return cfg, storer, snapshotID, keys
+}
+
+// deepVerifyJSON runs deep verify with --json on snapshotID and returns
+// the decoded result and the error RunDeepVerify returned.
+func deepVerifyJSON(
+	ctx context.Context, t *testing.T,
+	cfg *config.Config, storer storage.Storer, snapshotID string,
+) (vaultik.VerifyResult, error) {
+	t.Helper()
+
+	var out bytes.Buffer
+
+	v := &vaultik.Vaultik{
+		Config:  cfg,
+		Storage: storer,
+		Fs:      afero.NewOsFs(),
+		Stdout:  &out,
+		Stderr:  io.Discard,
+		UI:      ui.NewWithColor(io.Discard, false),
+	}
+	v.SetContext(ctx)
+
+	err := v.RunDeepVerify(snapshotID,
+		&vaultik.VerifyOptions{Deep: true, JSON: true})
+
+	return decodeVerifyResult(t, out.Bytes()), err
+}
+
 // corruptOneBlob flips a middle byte of the first blob file found under
 // blobsDir, leaving the file length unchanged.
 func corruptOneBlob(t *testing.T, fs afero.Fs, blobsDir string) {

internal/vaultik/snapshot.go

@@ -627,7 +627,8 @@ func selectSnapshotsToPurge(
 }
 
 // confirmAndExecutePurge shows deletion candidates, confirms with the
-// user, and deletes snapshots.
+// user, and deletes snapshots. It stops with an error at the first
+// snapshot it cannot delete.
 func (v *Vaultik) confirmAndExecutePurge(
 	toDelete []SnapshotInfo, force, quiet bool,
 ) error {
@@ -669,10 +670,11 @@ func (v *Vaultik) confirmAndExecutePurge(
 		snapshotID := snap.ID.String()
 		log.Info("Deleting snapshot", "id", snapshotID)
 
+		// Stop before the remote delete: a snapshot still in the local
+		// index must keep its metadata on the destination store.
 		err := v.deleteSnapshotFromLocalDB(snapshotID)
 		if err != nil {
-			log.Error("Failed to delete from local database",
-				"snapshot_id", snapshotID, "error", err)
+			return fmt.Errorf("deleting snapshot %s from local index: %w", snapshotID, err)
 		}
 
 		_, err = v.deleteRemoteSnapshotByKey(snapshot.RemoteSnapshotKey(snapshotID))

internal/vaultik/verify.go

@@ -17,6 +17,7 @@ import (
 	"sneak.berlin/go/vaultik/internal/database"
 	"sneak.berlin/go/vaultik/internal/log"
 	"sneak.berlin/go/vaultik/internal/snapshot"
+	"sneak.berlin/go/vaultik/internal/storage"
 )
 
 // Sentinel errors for snapshot verification failures.
@@ -32,7 +33,9 @@ var (
 	errManifestExtraBlob   = errors.New("manifest contains blob not in database")
 	errManifestMissingBlob = errors.New(
 		"manifest omits blob present in database")
-	errBlobSizeMismatch = errors.New("blob size mismatch")
+	errBlobSizeMismatch  = errors.New("blob size mismatch")
+	errBlobsNotInStorage = errors.New(
+		"blobs missing from storage or stored with the wrong size")
 )
 
 // verifyStatusFailed is the JSON status value for a failed verification.
@@ -149,7 +152,6 @@ func (v *Vaultik) RunDeepVerify(snapshotID string, opts *VerifyOptions) error {
 	}
 
 	result.Status = "ok"
-	result.Verified = len(dbBlobs)
 
 	if opts.JSON {
 		return v.outputVerifyJSON(result)
@@ -306,7 +308,7 @@ func (v *Vaultik) runVerificationSteps(
 		v.stdoutf("Checking blob existence in remote storage...\n")
 	}
 
-	err = v.verifyBlobExistenceFromDB(dbBlobs)
+	err = v.verifyBlobExistenceFromDB(dbBlobs, result)
 	if err != nil {
 		return v.deepVerifyFailure(result, opts, err.Error(), err)
 	}
@@ -317,7 +319,8 @@ func (v *Vaultik) runVerificationSteps(
 			len(dbBlobs), ubytes(totalSize))
 	}
 
-	err = v.performDeepVerificationFromDB(dbBlobs, tdb.db.Conn(), opts, identities)
+	err = v.performDeepVerificationFromDB(
+		dbBlobs, tdb.db.Conn(), opts, result, identities)
 	if err != nil {
 		return v.deepVerifyFailure(result, opts, err.Error(), err)
 	}
@@ -669,8 +672,14 @@ func (v *Vaultik) verifyManifestAgainstDatabase(
 	return nil
 }
 
-// verifyBlobExistenceFromDB checks that all blobs from database exist in S3
-func (v *Vaultik) verifyBlobExistenceFromDB(blobs []snapshot.BlobInfo) error {
+// verifyBlobExistenceFromDB checks that every blob from the database is
+// in storage with the size the database records. It checks them all,
+// counting the missing and wrong-size blobs in result, and fails if it
+// found any. Any storage error other than not found stops the check and
+// fails with that error.
+func (v *Vaultik) verifyBlobExistenceFromDB(
+	blobs []snapshot.BlobInfo, result *VerifyResult,
+) error {
 	log.Info("Verifying blob existence in S3", "blob_count", len(blobs))
 
 	for i, blob := range blobs {
@@ -683,17 +692,23 @@ func (v *Vaultik) verifyBlobExistenceFromDB(blobs []snapshot.BlobInfo) error {
 		// Construct blob path
 		blobPath := fmt.Sprintf("blobs/%s/%s/%s", blob.Hash[:2], blob.Hash[2:4], blob.Hash)
 
-		// Check blob exists
 		stat, err := v.Storage.Stat(v.ctx, blobPath)
-		if err != nil {
-			return fmt.Errorf("blob %s missing from storage: %w", blob.Hash, err)
-		}
+		switch {
+		case errors.Is(err, storage.ErrNotFound):
+			log.Warn("Blob missing from storage", "hash", blob.Hash)
 
-		// Verify size matches
-		if stat.Size != blob.CompressedSize {
-			return fmt.Errorf(
-				"%w: blob %s: S3 has %d bytes, database has %d bytes",
-				errBlobSizeMismatch, blob.Hash, stat.Size, blob.CompressedSize)
+			result.Missing++
+			result.MissingSize += blob.CompressedSize
+		case err != nil:
+			return fmt.Errorf("checking blob %s in storage: %w",
+				blob.Hash, err)
+		case stat.Size != blob.CompressedSize:
+			log.Warn("Blob has the wrong size in storage",
+				"hash", blob.Hash,
+				"storage_size", stat.Size,
+				"database_size", blob.CompressedSize)
+
+			result.Mismatched++
 		}
 
 		// Progress update every 100 blobs
@@ -707,16 +722,22 @@ func (v *Vaultik) verifyBlobExistenceFromDB(blobs []snapshot.BlobInfo) error {
 		}
 	}
 
+	if result.Missing > 0 || result.Mismatched > 0 {
+		return fmt.Errorf("%w: %d missing, %d with the wrong size",
+			errBlobsNotInStorage, result.Missing, result.Mismatched)
+	}
+
 	log.Info("✓ All blobs exist in storage")
 
 	return nil
 }
 
 // performDeepVerificationFromDB downloads and verifies the content of
-// each blob using the database as source.
+// each blob using the database as source, stopping at the first blob that
+// fails. Each blob that passes is counted in result.Verified.
 func (v *Vaultik) performDeepVerificationFromDB(
 	blobs []snapshot.BlobInfo, db *sql.DB, opts *VerifyOptions,
-	identities []age.Identity,
+	result *VerifyResult, identities []age.Identity,
 ) error {
 	// Calculate total bytes for ETA
 	var totalBytesExpected int64
@@ -739,6 +760,7 @@ func (v *Vaultik) performDeepVerificationFromDB(
 			return fmt.Errorf("blob %s verification failed: %w", blobInfo.Hash, err)
 		}
 
+		result.Verified++
 		bytesProcessed += blobInfo.CompressedSize
 		elapsed := time.Since(startTime)
 		remaining := len(blobs) - (i + 1)