Provide environment information
- Trigger.dev Cloud, project
proj_puwcdoxtxcywuaznmabj, prod env, region us-east-1
@trigger.dev/sdk 4.5.14, @trigger.dev/build 4.5.14- Node 22, deployed version
20260921.61
Describe the bug
A task that calls batchTriggerAndWait never resumed after every child run in the batch reached COMPLETED. The parent stayed frozen (no attempt error, durationMs: 0, no output) for ~95 minutes until we cancelled it by hand.
Three parent runs of the same task (scheduled-metrics-orchestration) hit this in one wave on 2026-09-21. Their siblings (17 other runs of the same task, same code, same batch sizes) resumed normally within a minute of their children completing.
Two shapes were observed:
1. Batch stuck in PROCESSING although all its runs finished (2 of 3 cases)
GET /api/v1/batches/{id} still returns status: "PROCESSING" and runs: [] hours after every child run completed:
| Parent run |
Batch |
runCount |
Batch created |
Last child finishedAt |
Batch status now |
run_06gccj5uiktpn7mrncb3mki501 |
batch_06gcclsm9uuel78rsk0jsgoa01 |
562 |
2026-09-21T23:40:24Z |
2026-09-21T23:41:13Z |
PROCESSING, runs: [], updatedAt 23:40:24Z |
run_06gccj5ugpgo8odb5kr8incm01 |
batch_06gcclu8fqilo4lfhncod8b401 |
76 |
2026-09-21T23:40:36Z |
2026-09-21T23:41:01Z |
PROCESSING, runs: [], updatedAt 23:40:37Z |
All 562 (resp. 76) child runs show status: COMPLETED under the parent's relatedRuns.children, and each child carries the batch id above. The batch record itself never transitioned out of PROCESSING, so it looks like the batch's run list was never sealed and the parent's waitpoint was never completed.
2. Batch COMPLETED, parent still never resumed (1 of 3 cases)
| Parent run |
Batch |
runCount |
Batch updatedAt |
Parent resumed? |
run_06gccj5uhhi8a3998loq7v9e01 |
batch_06gcclodi3fqtidru2fmj8bt01 |
529 |
2026-09-21T23:40:42Z (status: COMPLETED, 529 run ids listed) |
No — frozen until cancelled 2026-09-22T01:14:01Z |
All three parents were created at 2026-09-21T23:28:33Z from one batchTriggerAndWait of 20 items (batch_06gccj5sinpguml4d39lrfi101) issued by the cron task social-metrics-collection run run_06gccic2vbemcv2j2okvr9t801. Cancelling the three frozen parents (Canceled by user, 01:14–01:19Z) is what finally let the grandparent's batch wait resolve.
Reproduction
We cannot reproduce on demand. It happened to 3 of 20 concurrent parents doing the same thing: await childTask.batchTriggerAndWait(items) with 470–580 items per batch, each item carrying an idempotency key and a small JSON payload. Parent task has maxDuration: 300, child maxDuration: 600, each on its own queue with a concurrency limit.
Expected behavior
Once every run in the batch is in a final state the batch should be marked complete and the waiting parent should be resumed (or, if the batch record cannot be sealed, the parent should fail with an error rather than wait forever).
Additional information
Happy to share more run ids or grant dashboard access to the project if that helps.
Provide environment information
proj_puwcdoxtxcywuaznmabj, prod env, region us-east-1@trigger.dev/sdk4.5.14,@trigger.dev/build4.5.1420260921.61Describe the bug
A task that calls
batchTriggerAndWaitnever resumed after every child run in the batch reachedCOMPLETED. The parent stayed frozen (no attempt error,durationMs: 0, no output) for ~95 minutes until we cancelled it by hand.Three parent runs of the same task (
scheduled-metrics-orchestration) hit this in one wave on 2026-09-21. Their siblings (17 other runs of the same task, same code, same batch sizes) resumed normally within a minute of their children completing.Two shapes were observed:
1. Batch stuck in
PROCESSINGalthough all its runs finished (2 of 3 cases)GET /api/v1/batches/{id}still returnsstatus: "PROCESSING"andruns: []hours after every child run completed:finishedAtrun_06gccj5uiktpn7mrncb3mki501batch_06gcclsm9uuel78rsk0jsgoa01PROCESSING,runs: [],updatedAt23:40:24Zrun_06gccj5ugpgo8odb5kr8incm01batch_06gcclu8fqilo4lfhncod8b401PROCESSING,runs: [],updatedAt23:40:37ZAll 562 (resp. 76) child runs show
status: COMPLETEDunder the parent'srelatedRuns.children, and each child carries the batch id above. The batch record itself never transitioned out ofPROCESSING, so it looks like the batch's run list was never sealed and the parent's waitpoint was never completed.2. Batch
COMPLETED, parent still never resumed (1 of 3 cases)updatedAtrun_06gccj5uhhi8a3998loq7v9e01batch_06gcclodi3fqtidru2fmj8bt01status: COMPLETED, 529 run ids listed)All three parents were created at 2026-09-21T23:28:33Z from one
batchTriggerAndWaitof 20 items (batch_06gccj5sinpguml4d39lrfi101) issued by the cron tasksocial-metrics-collectionrunrun_06gccic2vbemcv2j2okvr9t801. Cancelling the three frozen parents (Canceled by user, 01:14–01:19Z) is what finally let the grandparent's batch wait resolve.Reproduction
We cannot reproduce on demand. It happened to 3 of 20 concurrent parents doing the same thing:
await childTask.batchTriggerAndWait(items)with 470–580 items per batch, each item carrying an idempotency key and a small JSON payload. Parent task hasmaxDuration: 300, childmaxDuration: 600, each on its own queue with a concurrency limit.Expected behavior
Once every run in the batch is in a final state the batch should be marked complete and the waiting parent should be resumed (or, if the batch record cannot be sealed, the parent should fail with an error rather than wait forever).
Additional information
Happy to share more run ids or grant dashboard access to the project if that helps.