The Job queue grows and grows. This causes a lot of issues. For example the candidates of speedy deletion are not at the category for that anymore.
Description
Details
| Subject | Author | Repo | Branch | Lines +/- | |
|---|---|---|---|---|---|
| Job runners: Add a dedicated htmlCacheUpdate runner | Ori | operations/puppet | production | +8 -0 |
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | ori | T124194 Job queue is growing and growing | |||
| Resolved | None | T122069 jobrunner memory leaks | |||
| Resolved | aaron | T123284 Record per-job-type memory usage statistics | |||
| Declined | None | T123815 Rewrite/modify $wgJobBackoffThrottling handling code to be less pessimistic |
Event Timeline
Change 265438 had a related patch set uploaded (by Ori.livneh):
Job runners: Add a dedicated htmlCacheUpdate runner
Change 265438 merged by Ori.livneh:
Job runners: Add a dedicated htmlCacheUpdate runner
The bulk of the jobs in the queue are htmlCacheUpdate jobs, and @aaron suspects that the runners are slow to clear the backlog because the job back-off throttling configuration is not having the effect that we expect it to have (thus T123815: Rewrite/modify $wgJobBackoffThrottling handling code to be less pessimistic). As an interim measure, we added (in I180856917b) a dedicated runner for htmlCacheUpdate jobs, and we're monitoring the situation.
Seems like it is fixed now, (the queue needs still some time to make the backlog smaller) but categorysation etc. works now.