Make snapshot projections self-diagnosing

internal/store/hosts.go

@@ -44,7 +44,7 @@ func (s *Store) LookupHostByAgentToken(ctx context.Context, tokenHash string) (*
 			repo_size_bytes, snapshot_count, open_alert_count,
 			applied_schedule_version, bandwidth_up_kbps, bandwidth_down_kbps,
 			pre_hook_default, post_hook_default,
-			repo_status, repo_status_error, always_on
+			repo_status, repo_status_error, always_on, snapshot_refreshed_at
 		 FROM hosts WHERE agent_token_hash = ?`,
 		tokenHash)
 	return scanHost(row)
@@ -59,7 +59,7 @@ func (s *Store) GetHost(ctx context.Context, id string) (*Host, error) {
 			repo_size_bytes, snapshot_count, open_alert_count,
 			applied_schedule_version, bandwidth_up_kbps, bandwidth_down_kbps,
 			pre_hook_default, post_hook_default,
-			repo_status, repo_status_error, always_on
+			repo_status, repo_status_error, always_on, snapshot_refreshed_at
 		 FROM hosts WHERE id = ?`, id)
 	return scanHost(row)
 }
@@ -227,7 +227,7 @@ func (s *Store) ListHosts(ctx context.Context) ([]Host, error) {
 			repo_size_bytes, snapshot_count, open_alert_count,
 			applied_schedule_version, bandwidth_up_kbps, bandwidth_down_kbps,
 			pre_hook_default, post_hook_default,
-			repo_status, repo_status_error, always_on
+			repo_status, repo_status_error, always_on, snapshot_refreshed_at
 		 FROM hosts ORDER BY name`)
 	if err != nil {
 		return nil, fmt.Errorf("store: list hosts: %w", err)
@@ -268,6 +268,7 @@ func scanHostRow(s hostScanner) (*Host, error) {
 		bwUp, bwDown                 sql.NullInt64
 		preHook, postHook            sql.NullString
 		alwaysOn                     int
+		snapshotRefreshedAt          sql.NullString
 	)
 	err := s.Scan(&h.ID, &h.Name, &h.OS, &h.Arch,
 		&h.AgentVersion, &h.ResticVersion, &h.ProtocolVersion,
@@ -276,7 +277,7 @@ func scanHostRow(s hostScanner) (*Host, error) {
 		&h.RepoSizeBytes, &h.SnapshotCount, &h.OpenAlertCount,
 		&h.AppliedScheduleVersion, &bwUp, &bwDown,
 		&preHook, &postHook,
-		&h.RepoStatus, &h.RepoStatusError, &alwaysOn)
+		&h.RepoStatus, &h.RepoStatusError, &alwaysOn, &snapshotRefreshedAt)
 	if err != nil {
 		if errors.Is(err, sql.ErrNoRows) {
 			return nil, ErrNotFound
@@ -332,6 +333,13 @@ func scanHostRow(s hostScanner) (*Host, error) {
 		h.PostHookDefault = postHook.String
 	}
 	h.AlwaysOn = alwaysOn != 0
+	if snapshotRefreshedAt.Valid {
+		t, err := time.Parse(time.RFC3339Nano, snapshotRefreshedAt.String)
+		if err != nil {
+			return nil, fmt.Errorf("store: parse snapshot_refreshed_at: %w", err)
+		}
+		h.SnapshotRefreshedAt = &t
+	}
 	return &h, nil
 }
 

internal/store/jobs.go

@@ -39,6 +39,27 @@ type RunningJobActivity struct {
 	LastActivity time.Time
 }
 
+// LatestSuccessfulRepoMutation returns the newest completion time for a job
+// that can change the repository's snapshot projection.
+func (s *Store) LatestSuccessfulRepoMutation(ctx context.Context, hostID string) (*time.Time, error) {
+	var raw sql.NullString
+	err := s.db.QueryRowContext(ctx, `
+		SELECT MAX(finished_at) FROM jobs
+		WHERE host_id = ? AND status = 'succeeded'
+		  AND kind IN ('backup', 'forget', 'prune')`, hostID).Scan(&raw)
+	if err != nil {
+		return nil, fmt.Errorf("store: latest successful repo mutation: %w", err)
+	}
+	if !raw.Valid {
+		return nil, nil
+	}
+	t, err := time.Parse(time.RFC3339Nano, raw.String)
+	if err != nil {
+		return nil, fmt.Errorf("store: parse latest repo mutation: %w", err)
+	}
+	return &t, nil
+}
+
 // ListRunningJobActivity returns every running job with its latest persisted
 // activity. Jobs without started_at are excluded because they have not actually
 // entered the running state coherently and cannot be aged safely here.

internal/store/migrations/0025_hosts_snapshot_refreshed_at.sql

@@ -0,0 +1 @@
+ALTER TABLE hosts ADD COLUMN snapshot_refreshed_at TEXT;

internal/store/snapshots.go

@@ -69,8 +69,8 @@ func (s *Store) ReplaceHostSnapshots(ctx context.Context, hostID string, snaps [
 	}
 
 	if _, err := tx.ExecContext(ctx,
-		`UPDATE hosts SET snapshot_count = ? WHERE id = ?`,
-		len(snaps), hostID); err != nil {
+		`UPDATE hosts SET snapshot_count = ?, snapshot_refreshed_at = ? WHERE id = ?`,
+		len(snaps), when.UTC().Format(time.RFC3339Nano), hostID); err != nil {
 		return fmt.Errorf("store: update host snapshot_count: %w", err)
 	}
 

internal/store/snapshots_test.go

@@ -138,7 +138,8 @@ func TestReplaceHostSnapshotsEmpty(t *testing.T) {
 		t.Fatalf("replace 1: %v", err)
 	}
 	// Then empty — host has been wiped.
-	if err := s.ReplaceHostSnapshots(ctx, hostID, nil, time.Now().UTC()); err != nil {
+	refreshedAt := time.Now().UTC().Truncate(time.Millisecond)
+	if err := s.ReplaceHostSnapshots(ctx, hostID, nil, refreshedAt); err != nil {
 		t.Fatalf("replace empty: %v", err)
 	}
 	out, err := s.ListSnapshotsByHost(ctx, hostID)
@@ -152,4 +153,7 @@ func TestReplaceHostSnapshotsEmpty(t *testing.T) {
 	if h.SnapshotCount != 0 {
 		t.Errorf("snapshot_count should reset to 0, got %d", h.SnapshotCount)
 	}
+	if h.SnapshotRefreshedAt == nil || !h.SnapshotRefreshedAt.Equal(refreshedAt) {
+		t.Errorf("empty projection refresh time = %v, want %s", h.SnapshotRefreshedAt, refreshedAt)
+	}
 }

internal/store/types.go

@@ -77,6 +77,7 @@ type Host struct {
 	LastBackupStatus       *string
 	RepoSizeBytes          int64
 	SnapshotCount          int
+	SnapshotRefreshedAt    *time.Time
 	OpenAlertCount         int
 	AppliedScheduleVersion int64
 	// Host-wide bandwidth caps applied to every restic invocation