Sat, Oct 13
Fri, Oct 12
Thu, Oct 11
@Marostegui These are the proposed indexes, if you want to discuss something concrete:
Wed, Oct 10
Can we shelf this in the backlog?
Nice hustle, this is deployed!
This task should be deduped or linked to T171619
Following up from IRC, I set loggers.celery.level to DEBUG in logging_config.yaml, and was able to demonstrate locally that celery debug logs would appear in /srv/log/ores/app.log.
I found this config on production, so in theory our workers should be restarting themselves:
Both issues mention a poll syscall returning POLLNVAL that we should be able to see in the strace if it's the same problem. Is it the worker manager or the worker threads which go 100% CPU? Do the straces match this profile?
Additionally, the itemquality -> articlequality is much easier, only affects Wikidata clients, and doesn't introduce a new key.
I'm going to push this forward, but would first like to reopen discussion of the name. "contentquality" is something we invented among our team, and has no inherent meaning to Wikimedians. Perhaps we should merge "itemquality" into "articlequality" instead, which would keep its approximate original meaning, and has easily discoverable documentation.
Looks like the trouble started on Sept 22nd, 16:00 UTC.
Tue, Oct 9
@Halfak I got some confirmation on IRC that nobody expects our extension to filter or rename users included in the JSON content. The analogous user signatures in wikitext aren't filtered in any special way, at least. This seems to clear our path to allowing usernames…
The fully implemented secondary schema is ready for TechCom review: