how do developers convey that they've triaged the logs?
Lars: Commenting on the train task
Brennen: Don't have to come to the meeting, but still need to check train logs
Dan: as a train deployer, I think the most valuable thing is knowing that the dashboard is "curated"
Dashboard here being mediawiki-new-errors
Lars: when there's a logspam or some other problem, the people involved with the code are awake and available -- mostly speaking about someone from the team being available.
Antoine: generic mw-error dashboard is pretty usable currently. One-off errors still show up. Adding a filter can be tricky for some messages.
Ahmon: best filter is a nag to fix it
Antoine: didn't know how to reach out to CPT -- wasn't even aware that their name changed to PET -- couldn't find those folks
Lars: update to developers maintainers page
Jeena: PHP vs JavaScript errors -- I don't think those should be seperate. There are a lot of different tabs to keep open when running train so JS errrors are extra
Ahmon: consolidated dashboards, hardest thing to figure out who does what. Sometimes people are quick, sometimes you can't find anyone
Added 'minimum hits' mod to logspam-watch to collapse errors than occur less than the minimum number of times.
https://github.com/dancysoft/mw-k8s-dev ready for Joe (and others) to check out. It builds a multiversion mediawiki image with operations/mediawiki-config (which hacks).
Need to work with somebody in the MW devel team to discuss offline l10n file generation (currently not possible without terrible hacks)
Limited bandwidth due to train dread and daughter being sick though the latter looks ok now.
Jeena and Ahmon, I came up with a syntax and possible implementation for list comprehensions in pipelinelib config that might support the multiversion build in a generalized way. Would love input if we all have time this week. Otherwise async or next Monday