juicedata/juicefs · error

SMembers %s: %s

Error message

SMembers %s: %s

What it means

When fetching session detail (detail=true), getSession reads the set of sustained (held-open but not yet committed) chunk slices via SMembers on the session's `sustained` key. If Redis returns a non-Nil error for that read, it is wrapped and returned as `SMembers <sid>: <err>`. This is a Redis-level failure (connection, cluster downscaled, key type mismatch, NOAUTH, READONLY, etc.), not a data corruption issue.

Source

Thrown at pkg/meta/redis.go:552

}

func (m *redisMeta) getSession(sid string, detail bool) (*Session, error) {
	ctx := Background()
	info, err := m.rdb.HGet(ctx, m.sessionInfos(), sid).Bytes()
	if err == redis.Nil { // legacy client has no info
		info = []byte("{}")
	} else if err != nil {
		return nil, fmt.Errorf("HGet sessionInfos %s: %s", sid, err)
	}
	var s Session
	if err := json.Unmarshal(info, &s); err != nil {
		return nil, fmt.Errorf("corrupted session info; json error: %s", err)
	}
	s.Sid, _ = strconv.ParseUint(sid, 10, 64)
	if detail {
		inodes, err := m.rdb.SMembers(ctx, m.sustained(s.Sid)).Result()
		if err != nil {
			return nil, fmt.Errorf("SMembers %s: %s", sid, err)
		}
		s.Sustained = make([]Ino, 0, len(inodes))
		for _, sinode := range inodes {
			inode, _ := strconv.ParseUint(sinode, 10, 64)
			s.Sustained = append(s.Sustained, Ino(inode))
		}

		locks, err := m.rdb.SMembers(ctx, m.lockedKey(s.Sid)).Result()
		if err != nil {
			return nil, fmt.Errorf("SMembers %s: %s", sid, err)
		}
		s.Flocks = make([]Flock, 0, len(locks)) // greedy
		s.Plocks = make([]Plock, 0, len(locks))
		for _, lock := range locks {
			owners, err := m.rdb.HGetAll(ctx, lock).Result()
			if err != nil {
				return nil, fmt.Errorf("HGetAll %s: %s", lock, err)
			}

View on GitHub (pinned to c9a67b23e8)

Solutions

  1. Check Redis connectivity/auth from the client: run `redis-cli -u <meta-url> SMEMBERS <anykey>` and resolve NOAUTH/WRONGPASS/READONLY errors first.
  2. If cluster MOVED/resharding, point juicefs at the correct cluster endpoints or retry after resharding completes; ensure `juicefs` connects to the primary, not a replica.
  3. Verify no external process overwrote the `sustained*` keys with wrong types (TYPE check); fix or delete the wrongly typed key.
  4. If the session is stale/crashed, simply remove the session record instead of fetching details.
  5. Retry the status command on transient network errors; add --no-detail by listing without detail to skip SMembers entirely where supported.

Example fix

// before: status fails with SMembers <sid>: READONLY You can't write against a read only replica.
// after: point the client at the primary endpoint
// before
juicefs status redis://replica-host:6379/1 --session 42
// after
juicefs status redis://primary-host:6379/1 --session 42
Defensive patterns

Strategy: try-catch

Validate before calling

// Precheck Redis health and auth from the same endpoint used by juicefs
$ redis-cli -u "$META_URL" PING
$ redis-cli -u "$META_URL" SMEMBERS <prefix>sustained<sid>

Try / catch

sess, err := metaCli.GetSession(sid, true)
if err != nil {
	var netErr net.Error
	if errors.As(err, &netErr) || strings.Contains(err.Error(), "SMembers") {
		// transient Redis failure: back off and retry
		time.Sleep(2 * time.Second)
		sess, err = metaCli.GetSession(sid, true)
	}
	if err != nil { return err }
}

Prevention

When it happens

Trigger: Calling `juicefs status <meta-url> --session <sid>` (GetSession/ListSessions with detail) while the Redis connection fails mid-request, the sustained key has been replaced with a non-set type, Redis is in READONLY replica mode, or auth/ACL denies access to the key.

Common situations: 1) Redis failover/timeout during the status command (MOVED, cluster resharding). 2) Wrong password/ACL after a config change (NOAUTH/WRONGPASS). 3) Querying a read-only replica or a proxy that rejects SMembers. 4) External tooling wrote a wrong-typed key at the sustained prefix.

Understand the failure class

Background: Database query failed: Internal Server Error 500s wrapping SQL, Prisma, and connection failures — what to check first — this error's family across 16 libraries.

Related errors

AI-assisted analysis of juicedata/juicefs@c9a67b23e8 (2026-09-06). Data as JSON: /api/errors/65132c1d9b41100e. Report an issue: GitHub.