We haven’t seen these issues on our hosting, or from any other self-hosters, so this is likely a problem specific to your server. We don’t have any relevant fixes in the pipeline.
From the symptoms, it sounds like Sidekiq (the background job processor) is not running correctly.
Do you see any recent errors when visiting /logs on your forum?
How does /sidekiq look? Does it show jobs being processed?
In a healthy state, the ‘enqueued’ number should be close to zero. I suspect if things have been stuck, your number is very high, and may take some time to process down to 0.
Sidekiq was (seemingly happily) running, but it seemed to be stuck since the update to 3.4 with about 1000 jobs queued. Since updates often do rebakes or async things, we didn’t really see that as something suspicious.
A restart of the process did the trick. I suspect the who’s online plugin might be the culprit but I’m not 100% sure.
The sidekiq dashboard showed nothing active, all five job slots were filled with 21-hour-old jobs, so sidekiq wasn’t pulling any new jobs of any sort, and the enqueued count was growing. I needed to apply system updates anyway, so I just restarted the system, which fixed the the problem.
The who’s online queue growing was my first tip-off.
Visiting the sidekiq dashboard did not “give it a kick”
Additionally, I noticed this warning on the Discourse dashboard:
I’m wondering whether that meant that there’s a warning that just didn’t show in the UI somewhere. I didn’t notice that warning immediately after my last update, to 176ee0bf60. I don’t check the dashboard often enough to know when it would have shown up.