Add Sentry Redis integration
This is an attempt at troubleshooting the numerous SystemExit
errors we've been getting on Sentry recently. They are all coming from /api/v1/jobs/
, just like when we had a flood of reindex
tasks for ElasticSearch. Either Redis or RQ are being incredibly slow, or there might be some connection issue, or RQ is just running a billion queries (I already had issues with RQ absolutely flooding Redis when there were too many jobs). But these errors seem to also occur even when there are no jobs to return at all!
I noticed that Sentry was discretely telling us to also enable the Redis integration. This adds a bunch of breadcrumbs that show the Redis queries directly just like SQL queries, which should make RQ's behavior much more explicit.
Edited by Erwan Rouchet