fix: checkpoint initialization gap on restarts

Entire-Checkpoint: 3f4d366b037c

pkg/metricstore/checkpoint.go

@@ -86,14 +86,16 @@ var (
 
 // Checkpointing starts a background worker that periodically saves metric data to disk.
 //
-// Checkpoints are written every 12 hours (hardcoded).
+// restoreFrom is the earliest timestamp of data loaded from checkpoint files at startup.
+// The first periodic checkpoint after restart will cover [restoreFrom, now], ensuring that
+// loaded data is re-persisted before old checkpoint files are cleaned up.
 //
 // Format behaviour:
 //   - "json": Periodic checkpointing every checkpointInterval
 //   - "wal":  Periodic binary snapshots + WAL rotation every checkpointInterval
-func Checkpointing(wg *sync.WaitGroup, ctx context.Context) {
+func Checkpointing(wg *sync.WaitGroup, ctx context.Context, restoreFrom time.Time) {
 	lastCheckpointMu.Lock()
-	lastCheckpoint = time.Now()
+	lastCheckpoint = restoreFrom
 	lastCheckpointMu.Unlock()
 
 	ms := GetMemoryStore()
@@ -337,25 +339,35 @@ func (l *Level) toCheckpoint(dir string, from, to int64, m *MemoryStore) error {
 		return ErrNoNewArchiveData
 	}
 
-	filepath := path.Join(dir, fmt.Sprintf("%d.json", from))
-	f, err := os.OpenFile(filepath, os.O_CREATE|os.O_WRONLY|os.O_TRUNC, CheckpointFilePerms)
+	finalPath := path.Join(dir, fmt.Sprintf("%d.json", from))
+	tmpPath := finalPath + ".tmp"
+
+	f, err := os.OpenFile(tmpPath, os.O_CREATE|os.O_WRONLY|os.O_TRUNC, CheckpointFilePerms)
 	if err != nil && os.IsNotExist(err) {
 		err = os.MkdirAll(dir, CheckpointDirPerms)
 		if err == nil {
-			f, err = os.OpenFile(filepath, os.O_CREATE|os.O_WRONLY|os.O_TRUNC, CheckpointFilePerms)
+			f, err = os.OpenFile(tmpPath, os.O_CREATE|os.O_WRONLY|os.O_TRUNC, CheckpointFilePerms)
 		}
 	}
 	if err != nil {
 		return err
 	}
-	defer f.Close()
 
 	bw := bufio.NewWriter(f)
 	if err = json.NewEncoder(bw).Encode(cf); err != nil {
+		f.Close()
+		os.Remove(tmpPath)
 		return err
 	}
 
-	return bw.Flush()
+	if err = bw.Flush(); err != nil {
+		f.Close()
+		os.Remove(tmpPath)
+		return err
+	}
+	f.Close()
+
+	return os.Rename(tmpPath, finalPath)
 }
 
 // enqueueCheckpointHosts traverses checkpoint directory and enqueues cluster/host pairs.
@@ -470,7 +482,7 @@ func (l *Level) loadFile(cf *CheckpointFile, m *MemoryStore) error {
 			data:      metric.Data[0:n:n],
 			prev:      nil,
 			next:      nil,
-			archived:  true,
+			archived:  false,
 		}
 
 		minfo, ok := m.Metrics[name]