Intermittent stalled jobs with Bull 4.x when using concurrency > 1, related to ioredis minor version drift

[andressalro]

Description

We experienced an intermittent issue with Bull 4.x where jobs are marked as stalled without transitioning to failed, and workers appear to stop processing jobs when concurrency > 1.
The same codebase works correctly with concurrency = 1.

After extensive comparison, the issue was traced to a minor version difference in ioredis, even though the Bull version and Node.js version were identical.

This issue may be relevant for other users running Bull 4.x in production environments (Kubernetes / Docker / Redis clusters).

---

Environment

Bull: 4.16.5
Node.js: v20.19.0

---

Observed Behavior

With concurrency = 1
Jobs are processed normally
No stalled events

With concurrency = 3
Jobs enter active
After some time, jobs are marked as stalled
No failed event is emitted
The worker stops processing further jobs ("silent stall")
No application-level errors are logged

This behavior is intermittent but reproducible under load.

---

Key Finding

Two independent workers, built from the same monorepo and running the same Bull and Node.js versions, exhibited different runtime behavior due to a minor version difference in the Redis client (ioredis).

Worker Bull ioredis
Worker A (stable) 4.16.5 5.8.2
Worker B (unstable) 4.16.5 5.9.0

The issue was not related to application logic, Redis infrastructure, or Bull configuration, but rather to dependency resolution drift.

[blackalkane]

We encountered the same issue, while ioredis v5.9.2 seems to fix it, could you please retry to confirm @andressalro ? Thanks

[SiebelsTim]

We encounter a similar issue, however we cannot pinpoint it to different versions, they are the same.

We observe that a job is marked as stalled without ever entering the active state.

"time":14:12:09,"message":"Job with id \"3227\" failed: job stalled more than allowable limit"
"time":14:10:38,"message":"Job with id \"3227\" has stalled.",
"time":14:09:11,"message":"Job with id \"3227\" is waiting to be processed."
"time":14:09:10,"message":"Created Job with id \"3227\"

In this example, lockDuration is configured to be 120_000ms.

edit: I've had more commands in the redis queue than it could handle. That might've been the issue.