n8n queue mode job stuck – why jobs stay in Queued and how to fix it

n8n Queue Mode Errors / Leave a Comment
Step by Step Guide to solve n8n queue mode job stuck
Step by Step Guide to solve n8n queue mode job stuck

 

Who this is for: Ops engineers, DevOps teams, and n8n power‑users who run n8n in queue mode (Redis or PostgreSQL) and need a fast, repeatable way to un‑block a hung execution. We cover this in detail in the n8n Queue Mode Errors Guide.

Quick Diagnosis

Problem: An n8n workflow stays in “pending” (or “queued”) forever, never advancing to “running” or “finished”.

One‑line fix: Find the execution’s id, reset its status with the n8n CLI (or a guarded DB update), and clear any orphaned queue lock.

1. Why Jobs Get Stuck in Queue Mode ?

If you encounter any n8n queue mode duplicate execution resolve them before continuing with the setup.

Root cause UI / Log symptom Typical trigger
Orphaned Redis lock `Job is still in pending state` repeats Worker crash or container restart
Corrupt DB row `status = “pending”` but `startedAt` is `null` Manual DB edit or incomplete migration
Infinite loop in a node No progress after first node runs Custom JavaScript that never resolves
Resource throttling Growing queue backlog, new jobs never dequeued CPU/memory limits, OOM kills
Version mismatch (core vs. queue plugin) `Cannot read property ‘queue’ of undefined` Upgrade without rebuilding the plugin

Understanding the root cause tells you whether to reset the execution, clear the queue, or fix the workflow code.

2. Locate the Stuck Job

2.1 UI (read‑only)

  1. Open Execution List → filter Status = Pending.
  2. Hover the row and copy the Execution ID (e.g., 62f9c7b1-3e8a-4d2b-9f7a-9c4e5a2d1b3c).

EEFA: Do not click “Retry” on a stuck job; it will re‑queue the same corrupted entry.

2.2 CLI

# List the most recent pending executions (requires n8n ≥ 1.0)
n8n execution:list --status=pending --limit=20

The output shows id, workflowId, and createdAt.

2.3 Direct DB Query (PostgreSQL example)

SELECT id, workflow_id, status, created_at
FROM execution_entity
WHERE status = 'pending'
ORDER BY created_at ASC
LIMIT 20;

EEFA: Run the query on a read‑only replica first to verify rows before making changes. If you encounter any n8n queue mode missing worker process resolve them before continuing with the setup.

3. Safely Release / Reset the Stuck Job

3.1 Preferred: CLI “reset” command

# Replace  with the ID you located
n8n execution:reset

What it does

  • Sets status = 'failed' in the DB.
  • Deletes the Redis lock key n8n:queue:lock:<id>.
  • Emits an execution.finished event so downstream triggers (e.g., Webhooks) are notified.

EEFA: The CLI respects the executionTimeout setting, preventing accidental infinite loops.

3.2 Manual DB reset (when CLI unavailable)

Step 1 – Mark the execution as failed

BEGIN;

UPDATE execution_entity
SET status = 'failed',
    stopped_at = NOW(),
    error = 'Manually reset after queue stall'
WHERE id = '<EXECUTION_ID>';

Step 2 – Remove the Redis lock (if using Redis)

# Run this in a Redis client
redis-cli DEL "n8n:queue:lock:<EXECUTION_ID>"
COMMIT;

EEFA: Wrap the UPDATE in a transaction and take a DB snapshot before committing on production.

3.3 Flush the entire queue (last resort)

Redis backend

# List all queue keys and delete them
redis-cli KEYS "n8n:queue:*" | xargs -r redis-cli DEL

PostgreSQL‑based queue (pg‑boss)

DELETE FROM pgboss.job WHERE state = 'created';

EEFA: Flushing discards **all** pending jobs. Use only during a maintenance window when the backlog can be safely dropped.

4. Verify the Fix & Prevent Recurrence

4.1 Post‑reset sanity check

n8n execution:get

You should see status = "failed" and a populated stoppedAt.

Check Expected result
UI shows the execution as Failed ✅
No Redis lock key exists ✅ (redis-cli EXISTS n8n:queue:lock:<id> → 0)
Queue length returns to normal ✅ (redis-cli LLEN n8n:queue:jobs)

4.2 Preventive checklist

  • Enable executionTimeout in config.yml (e.g., executionTimeout: 300).
  • Monitor queue depth via Prometheus (n8n_queue_length).
  • Alert when pending executions exceed 5 minutes.
  • Upgrade n8n and the queue‑mode plugin together (matching major versions).
  • Back up the execution_entity table nightly.

5. Real‑World Example: Kubernetes Deployment

Scenario: An n8n pod crashed during a large CSV import, leaving 12 jobs pending.

5.1 Identify the stuck executions

kubectl logs -l app=n8n -c n8n --since=2h | grep "pending"
# Example output: Execution IDs 9b1c2d3e-...

5.2 Reset each execution from inside the pod

# Replace <POD_NAME> and <EXECUTION_ID> accordingly
kubectl exec -it $(kubectl get pod -l app=n8n -o jsonpath="{.items[0].metadata.name}") -- \
  n8n execution:reset 9b1c2d3e-...

5.3 Roll out a clean queue worker

kubectl rollout restart deployment/n8n

EEFA: The rollout forces a fresh pod that starts with a clean Redis state (the PVC is ephemeral), eliminating any lingering lock keys.

Conclusion

A “job stuck” condition is almost always a state‑sync mismatch between the execution table and the queue backend. The reliable remedy is to reset the execution (CLI preferred), clear any orphaned lock, and then verify that the queue depth returns to normal. Reinforce the fix with timeout settings, queue‑depth monitoring, and version‑aligned upgrades to keep n8n running smoothly in production.

Must Read

Leave a Comment

Your email address will not be published. Required fields are marked *