Sender | Message | Time |
---|---|---|
28 Mar 2020 | ||
tgr | in any case, the job queue should not be used for critical tasks | 17:12:15 |
tgr | when it is, it should use some sort of journal system | 17:12:30 |
tgr | RenameUser does that, for example | 17:12:39 |
bryandamon | Possibly poorly configured extensions, I'm not sure, but I've had stuck jobs that never resolved themselves and had to manually delete them from the database to allow them, and other jobs to continue. | 17:13:49 |
tgr | I'm still not sure what that means | 17:24:21 |
tgr | there's no timeout configured for the job runner and the job gets into an infinite loop? | 17:24:43 |
tgr | the job is in the queue and never gets executed? | 17:25:02 |
tgr | it does not get removed from the queue after it fails? | 17:25:16 |
darenwelsh | the job is attempted once and never re-tries, IIRC | 17:25:52 |
darenwelsh | I think i have also tried changing the "attempts" value from 1 to 0 before, but that was a long time ago and I don't remember for sure | 17:26:14 |
bryandamon | Yeah, it's there, doesn't leave, and prevents other jobs from running (my experience). | 17:26:19 |
tgr | that sounds like a problem with the job runner more than the job | 17:27:07 |
darenwelsh | re: preventing ... not sure it has prevented independent jobs, but maybe didn't run and therefore didn't create new jobs as an expected result for that one job (In how I interpreted it) | 17:27:44 |
tgr | in theory failed jobs are logged and discarded, possibly after some retries | 17:27:58 |
darenwelsh | in my case, the job runner is just a cron task to runJobs.php | 17:28:03 |
darenwelsh | which is MW core | 17:28:14 |
tgr | would be a core bug then, if it can be reproduced | 17:29:09 |
darenwelsh | if I could reproduce it, I'd know more about why it's happening, or at least based on more specific actions that lead to it. But I think it's just based on updates to templates. | 17:30:21 |
tgr | well, when the job gets stuck you can run the job runner and observe what happens | 17:36:02 |
tgr | that should be enough of a reproduction | 17:36:11 |
darenwelsh | Redacted or Malformed Event | 17:50:34 |
darenwelsh | So here's an example error from runJobs.php
I have no idea what I'm supposed to do with that. | 18:00:28 |
tgr | is it causing problems with the job queue or are you just unhappy about the job failing? | 18:10:42 |
tgr | it will cause the link tables (and thus what-links-here and similar) to be out of sync, which is not a tragedy | 18:11:14 |
tgr | the runJobs log channel should include job parameter information, at a glance this error is caused by something creating a RefreshLinksJob with none of the recursive , pages , title options set | 18:17:38 |
tgr | (which arguably should be reported as a bug - errors like that should throw in the constructor, not when the job is executed) | 18:18:06 |
darenwelsh | both - I'm unhappy because something isn't being done that should be and yes it messes up the job running because runJobs.php exits when it hits one of these. So then other jobs aren't run that could be until that job runner is started again. | 18:18:48 |
darenwelsh | so, how would I find out what created a job like this that is missing those options? | 18:21:45 |
tgr | on a closer read, if the title is not specified, it will just default to Special:BadTitle | 21:31:15 |
tgr | so something must pass an invalid title | 21:32:08 |