fix(dbbackup): 备份新鲜度按最新的 daily 包算，手动或异地包不再掩盖停掉的定时任务

internal/watchdog/probes.go

@@ -341,8 +341,12 @@ func PostgresDown(err error) Finding {
 	}
 }
 
-// BackupFinding reports a control-plane database backup older than a day, or
-// none at all, in dir, and a fresh one that would restore no servers.
+// BackupFinding reports a daily control-plane database backup older than a
+// day, or none at all, in dir, and a fresh one that would restore no servers.
+// The age goes by the daily bundles alone: a manual or off-site bundle taken
+// since would keep a stopped timer quiet until it too is a day old. The
+// servers check reads the newest bundle of any label, the one a lost host
+// restores from.
 func BackupFinding(dir string, now time.Time) *Finding {
 	bundles, err := dbbackup.List(dir)
 	if err != nil {
@@ -353,19 +357,28 @@ func BackupFinding(dir string, now time.Time) *Finding {
 			Hint:      "docs/troubleshooting.md §16",
 		}
 	}
-	if len(bundles) > 0 && now.Sub(bundles[0].Created) <= maxBackupAge {
+	daily, ok := dbbackup.Newest(bundles, dbbackup.LabelDaily)
+	if ok && now.Sub(daily.Created) <= maxBackupAge {
 		return serversFinding(bundles[0])
 	}
 	f := &Finding{
 		Key: "db-backup", Severity: Critical, For: backupFor,
 		Summary:   fmt.Sprintf("%s 里没有任何控制面数据库备份", dir),
 		SummaryEN: fmt.Sprintf("no control-plane database backup in %s", dir),
-		Hint:      "journalctl -u felis-db-backup -n 50; take one now with `sudo felis db backup` (docs/troubleshooting.md §16)",
+		Hint:      "journalctl -u felis-db-backup -n 50; once fixed, run it now with `sudo systemctl start felis-db-backup` (docs/troubleshooting.md §16)",
 	}
-	if len(bundles) > 0 {
-		age := roundHours(now.Sub(bundles[0].Created))
-		f.Summary = fmt.Sprintf("最新的控制面数据库备份已是 %s 前（%s）", age, bundles[0].Name)
-		f.SummaryEN = fmt.Sprintf("the newest control-plane database backup is %s old (%s)", age, bundles[0].Name)
+	switch {
+	case ok:
+		age := roundHours(now.Sub(daily.Created))
+		f.Summary = fmt.Sprintf("最新的每日控制面数据库备份已是 %s 前（%s）", age, daily.Name)
+		f.SummaryEN = fmt.Sprintf("the newest daily control-plane database backup is %s old (%s)", age, daily.Name)
+	case len(bundles) > 0:
+		f.Summary = fmt.Sprintf("%s 里没有每日定时的控制面数据库备份", dir)
+		f.SummaryEN = fmt.Sprintf("no daily control-plane database backup in %s", dir)
+	}
+	if len(bundles) > 0 && bundles[0].Label != dbbackup.LabelDaily {
+		f.Summary += fmt.Sprintf("；更新的 %s 来自手动或其他任务，定时任务修好之前它会一天天变旧", bundles[0].Name)
+		f.SummaryEN += fmt.Sprintf("; the newer %s came from a manual run or another job and ages while the timer stays broken", bundles[0].Name)
 	}
 	return f
 }

internal/watchdog/probes_test.go

@@ -149,16 +149,34 @@ func TestBackupFinding(t *testing.T) {
 	if f := BackupFinding(dir, t0); f == nil || !strings.Contains(f.SummaryEN, "no control-plane database backup") {
 		t.Fatalf("empty dir: %+v", f)
 	}
-	touch := func(at time.Time) {
-		if err := os.WriteFile(filepath.Join(dir, dbbackup.BundleName(at, "daily")), []byte("x"), 0o600); err != nil {
+	touch := func(at time.Time, label string) {
+		if err := os.WriteFile(filepath.Join(dir, dbbackup.BundleName(at, label)), []byte("x"), 0o600); err != nil {
 			t.Fatal(err)
 		}
 	}
-	touch(t0.Add(-30 * time.Hour))
-	if f := BackupFinding(dir, t0); f == nil || !strings.Contains(f.SummaryEN, "30h old") {
-		t.Fatalf("stale: %+v", f)
+	// A manual bundle alone: the timer has never written one.
+	touch(t0.Add(-time.Hour), "manual")
+	if f := BackupFinding(dir, t0); f == nil || f.Key != "db-backup" ||
+		f.SummaryEN != "no daily control-plane database backup in "+dir+"; the newer felis-db-20260924T110000Z-manual.tar came from a manual run or another job and ages while the timer stays broken" ||
+		f.Summary != dir+" 里没有每日定时的控制面数据库备份；更新的 felis-db-20260924T110000Z-manual.tar 来自手动或其他任务，定时任务修好之前它会一天天变旧" {
+		t.Fatalf("manual only: %+v", f)
 	}
-	touch(t0.Add(-2 * time.Hour))
+	// A stale daily bundle under that fresh manual one: still the timer's alarm.
+	touch(t0.Add(-30*time.Hour), "daily")
+	if f := BackupFinding(dir, t0); f == nil || f.Key != "db-backup" ||
+		f.SummaryEN != "the newest daily control-plane database backup is 30h old (felis-db-20260923T060000Z-daily.tar); the newer felis-db-20260924T110000Z-manual.tar came from a manual run or another job and ages while the timer stays broken" ||
+		!strings.Contains(f.Hint, "sudo systemctl start felis-db-backup") {
+		t.Fatalf("stale daily under a fresh manual: %+v", f)
+	}
+	stale := t.TempDir()
+	if err := os.WriteFile(filepath.Join(stale, dbbackup.BundleName(t0.Add(-30*time.Hour), "daily")), []byte("x"), 0o600); err != nil {
+		t.Fatal(err)
+	}
+	if f := BackupFinding(stale, t0); f == nil || f.SummaryEN != "the newest daily control-plane database backup is 30h old (felis-db-20260923T060000Z-daily.tar)" ||
+		f.Summary != "最新的每日控制面数据库备份已是 30h 前（felis-db-20260923T060000Z-daily.tar）" {
+		t.Fatalf("stale daily alone: %+v", f)
+	}
+	touch(t0.Add(-2*time.Hour), "daily")
 	if f := BackupFinding(dir, t0); f != nil {
 		t.Fatalf("fresh backup reported: %+v", f)
 	}