Claimed jobs don't run when queue is cleared

[Mortalife]

I have a few hundred items in a queue which has a concurrency of 4.

It seems that upon starting, 4 are immediately ran, and then 3 or so end up as "claimed".

Once those 4 are complete nothing happens, I have to wait for releaseStaleJobsIntervalMin: 1 and releaseStaleJobsMaxClaimedMs: 60_000 to remove the claimed jobs and then retry. This then results in all outstanding jobs being marked as claimed but only 4 get processed. After those 4 have be completed, none of the claimed jobs are ran, I have to wait for the cleanup again.

This doesn't seem right? I would expect that once a queue becomes available that the next job is processed.

On the first iteration I see:

[debug] [2025-12-31 18:38:03] [Sidequest] [Dispatcher] : Queue background limit reached!

However not one subsequent cleanups.

I do see a mass of spam for no available slots.

[debug] [2025-12-31 18:38:00] [Sidequest] [Executor Manager] : Submitting job 597 for execution in queue background
[debug] [2025-12-31 18:38:00] [Sidequest] [Executor Manager] : No available slots for job 597 in queue background
[debug] [2025-12-31 18:38:00] [Sidequest] [JobTransitioner] : Transition SnoozeTransition not applicable for job 597. Returning original job data.
[debug] [2025-12-31 18:38:00] [Sidequest] [Executor Manager] : Submitting job 598 for execution in queue background
[debug] [2025-12-31 18:38:00] [Sidequest] [Executor Manager] : No available slots for job 598 in queue background
[debug] [2025-12-31 18:38:00] [Sidequest] [JobTransitioner] : Transition SnoozeTransition not applicable for job 598. Returning original job data.
[debug] [2025-12-31 18:38:00] [Sidequest] [Executor Manager] : Submitting job 599 for execution in queue background
[debug] [2025-12-31 18:38:00] [Sidequest] [Executor Manager] : No available slots for job 599 in queue background
[debug] [2025-12-31 18:38:00] [Sidequest] [JobTransitioner] : Transition SnoozeTransition not applicable for job 599. Returning original job data.

Seems to me that too many jobs are being claimed, and there's no mechanism notify claimed jobs that they can execute.

I would expect that N jobs be pulled from jobs table whenever there is remaining queue concurrency, and they wouldn't be claimed until that point, since in theory, they could be processed by another runner.

20251231-1858-01.4077712.mp4

This one shows the 4 running, and then nothing.

20251231-1900-05.8843504.mp4

[Mortalife]

I've done some testing locally, and it seems that your SnoozeTransition job is set to only run with waiting and running, however this is used when the state of the job is in claimed and as such cannot be moved back into waiting.

I believe this should be changed from waiting to claimed as they're the states you'd want to progress a job into a waiting state.

https://github.com/sidequestjs/sidequest/blob/master/packages/core/src/transitions/snooze-transition.ts#L43

[merencia]

Hey @Mortalife, thanks for reporting that.

Yep, we miss that transition between claimed to waiting, and this line does that:

sidequest/packages/engine/src/execution/executor-manager.ts

Line 92 in 5528527

await JobTransitioner.apply(this.backend, job, new SnoozeTransition(0));

[merencia]

fixed by #151

[merencia]

@Mortalife we found a definitive fix for that: #152

[GiovaniGuizzo]

Released in 1.13.3: https://github.com/sidequestjs/sidequest/releases/tag/v1.13.3

[GiovaniGuizzo]

@Mortalife we have thoroughly tested it and created a regression test for the issue. Could you please try this new version and let us know? It should be more efficient than snoozing claimed jobs. Now Sidequest correctly claims the exact number of jobs.

[Mortalife]

Thanks for looking more deeply, I'll apply your changes and see how it behaves with my workload