Dear instances, make sure to update your sidekiq processes. Indead, we are adopting a new strategy for daemons that process our background
jobs.
Now their is only 3 queues on sidekiq (still, we keep archive/export on delayed job for now).
The big idea : allow any worker to deal with any queue (capacity mgmt) :
- critical: must be procesed now
- default: will be processed asap
- low: do it after everything else
FYI: we had 50% of our server capacity on reserved queues. Leading to
bottle neck processing (reserved queues/worker not processing other
queues). We HAD TO fix it. Sorry for the inconvenience
TBH, this is an idea of Colin Darie <colin@darie.eu>. I'm just pushing
it forward.
Co-Authored-By: Colin Darie <colin@darie.eu>
Ces jobs sont empilés par milliers d'un coup et peuvent mettre des dizaines
de minutes à s'écouler. On ne veut pas que ça interfère avec d'autres jobs,
notamment les scans antivirus qui, si retardés, empilent des "poll" pour
connaitre le status du fichier.
