Hi All, I'm using the experimental scheduler for parallel processing of tasks that come in short bursts (up to 300 tasks every 5 to 60 minutes). Scheduler_task queue is populated by a homemade background worker (-M -N -R script) and I'm running 10 scheduler workers for task processing.
Everything works like a charm 99% of time but occasionally tasks stop being consumed (they get stuck in a QUEUED state). As far as I can tell the workers are still fine - the show up as ACTIVE and I can see their heartbeats updated. I've been chasing this for some time now and I'm considering implementing a homemade queue instead. Any help would be appreciated. Any thoughts? Thanks! --