The local whole-guest tier keeps one restore point per day for 3 days: keep-daily instead of keep-last (R-1149, operator D17, unreleased)

internal/localapi/backup_cap_test.go

@@ -12,6 +12,7 @@ import (
 	"time"
 
 	"gitea.dooplex.hu/admin/felhom-agent/internal/backup"
+	"gitea.dooplex.hu/admin/felhom-agent/internal/config"
 	"gitea.dooplex.hu/admin/felhom-agent/internal/hub"
 )
 
@@ -25,11 +26,46 @@ type keepLastTier struct {
 	mu       sync.Mutex
 	clock    *testClock
 	keepLast int
+	// spec, when set, is the `--prune-backups` value the agent passes (R-1149) and the tier prunes the way Proxmox
+	// reads it: keep-last=N → the newest N; keep-daily=N → the newest copy of each of the N newest days with one.
+	spec     string
+	t        *testing.T
 	archives []time.Time
 	fail     bool
 	runs     int
 }
 
+// pruneBySpec applies a Proxmox `--prune-backups` spec to archives (oldest first) and returns what is kept.
+func pruneBySpec(t *testing.T, archives []time.Time, spec string) []time.Time {
+	var n int
+	var mode string
+	if _, err := fmt.Sscanf(spec, "keep-last=%d", &n); err == nil {
+		mode = "last"
+	} else if _, err := fmt.Sscanf(spec, "keep-daily=%d", &n); err == nil {
+		mode = "daily"
+	} else {
+		t.Fatalf("the model does not know the prune spec %q", spec)
+	}
+	var kept []time.Time
+	days := map[string]bool{}
+	for i := len(archives) - 1; i >= 0; i-- { // newest first, as Proxmox walks them
+		a := archives[i]
+		switch mode {
+		case "last":
+			if len(kept) < n {
+				kept = append([]time.Time{a}, kept...)
+			}
+		case "daily":
+			d := a.Format("2006-01-02")
+			if !days[d] && len(days) < n {
+				days[d] = true
+				kept = append([]time.Time{a}, kept...)
+			}
+		}
+	}
+	return kept
+}
+
 func (k *keepLastTier) BackupWithSnapshotHook(_ context.Context, vmid int, _ func()) (hub.Backup, error) {
 	k.mu.Lock()
 	defer k.mu.Unlock()
@@ -39,7 +75,9 @@ func (k *keepLastTier) BackupWithSnapshotHook(_ context.Context, vmid int, _ fun
 		return hub.Backup{VMID: vmid, StartedAt: now.Format(time.RFC3339)}, fmt.Errorf("vzdump failed")
 	}
 	k.archives = append(k.archives, now)
-	if len(k.archives) > k.keepLast {
+	if k.spec != "" {
+		k.archives = pruneBySpec(k.t, k.archives, k.spec)
+	} else if len(k.archives) > k.keepLast {
 		k.archives = k.archives[len(k.archives)-k.keepLast:]
 	}
 	return hub.Backup{VMID: vmid, Success: true, Archive: "x", StartedAt: now.Format(time.RFC3339)}, nil
@@ -291,3 +329,49 @@ func TestBackupCadenceCap(t *testing.T) {
 		}
 	}
 }
+
+// R-1149 — the consequence of keep-daily on the local tier, with the agent's REAL prune spec (config.BackupTiers →
+// PruneSpec, what runDaemon hands the runner). Three honest nights, then a compromise at 01:00 that floods
+// POST /backup for the rest of the day. The cap lets 3 runs through inside the rolling 24 h (two at once, one when
+// last night's success ages out). With keep-last 3 those three were the only copies left; with keep-daily 3 the two
+// earlier nights survive the whole day.
+//
+// COMPANION RED-PROOF (observed 2026-10-19): make BackupTier.PruneSpec return keep-last for the primary → this fails
+// with "after a day of flooding the local tier holds 0 copies made before the compromise (kept […]) — want the 2
+// earlier nights". Restored.
+func TestBackupCap_R1149_LocalFloodKeepsOlderDays(t *testing.T) {
+	tiers, _ := config.BackupConfig{LocalBackupTarget: "local", LocalBackupRetention: 3}.BackupTiers()
+	spec := tiers[0].PruneSpec()
+	day := time.Date(2026, 6, 10, 0, 0, 0, 0, time.UTC)
+	clock := &testClock{t: day}
+	local := &keepLastTier{clock: clock, keepLast: 3, spec: spec, t: t}
+	srv := capServer(t, filepath.Join(t.TempDir(), "s.json"), clock, local, &keepLastTier{clock: clock, keepLast: 2})
+	for d := 3; d >= 1; d-- { // the nights of D-3, D-2, D-1 at 03:00
+		clock.t = day.Add(-time.Duration(d)*24*time.Hour + 3*time.Hour)
+		if code := runBackup(t, srv, "local"); code != http.StatusAccepted {
+			t.Fatalf("seeding night D-%d: %d", d, code)
+		}
+	}
+	compromise := day.Add(time.Hour)
+	ran := 0
+	for clock.t = compromise; clock.t.Before(day.Add(24 * time.Hour)); clock.Advance(10 * time.Minute) {
+		if runBackup(t, srv, "local") == http.StatusAccepted {
+			ran++
+		}
+	}
+	if ran != 3 {
+		t.Fatalf("the flood ran %d backups on day D — want 3 (the cap)", ran)
+	}
+	local.mu.Lock()
+	kept := append([]time.Time(nil), local.archives...)
+	local.mu.Unlock()
+	pre := 0
+	for _, a := range kept {
+		if a.Before(compromise) {
+			pre++
+		}
+	}
+	if pre != 2 {
+		t.Fatalf("after a day of flooding the local tier holds %d copies made before the compromise (kept %v) — want the 2 earlier nights", pre, kept)
+	}
+}