[08:19:19] <atsukoito>	 Good morning! eqiad reindexing finished overnight, codfw need a lil bit more time (it hit a snag)
[08:19:50] <atsukoito>	 I guess we can switch over from CirrusSearch by the afternoon
[08:19:51] <pfischer>	 o/ atsukoito: thanks!
[08:35:07] <dcausse>	 o/
[10:02:20] <atsukoito>	 dcausse: https://gerrit.wikimedia.org/r/c/operations/mediawiki-config/+/1302197 we are finally at this point
[10:02:30] <dcausse>	 nice! looking
[10:04:38] <atsukoito>	 do you think we can do it in the afternoow window? https://schedule-deployment.toolforge.org/window/1781614800
[10:06:02] <dcausse>	 atsukoito: yes I think it should fit
[10:07:47] <atsukoito>	 thanks, I scheduled it
[10:11:12] <dcausse>	 lunch
[10:40:26] <atsukoito>	 lunch, too
[13:00:25] <inflatador>	 <o/
[13:07:54] <dcausse>	 o/
[13:10:36] <ebernhardson>	 \o
[13:11:54] <dcausse>	 o/²
[14:03:26] <ebernhardson>	 fingers just failing today...trying to go to Special:ApiSandbox and somehow typed Special:Search
[14:40:52] <atsukoito>	 Cross-post from hash-wikimedia-operations, opensearch on k8s doesn't work with Special:Translate, just doesn't return anything.
[14:41:06] <ebernhardson>	 :S do you have a request body?
[14:41:45] <atsukoito>	 ebernhardson: dcausse and me has only discovered this during the release, so no.
[14:41:49] <inflatador>	 d-causse e-bernhardson just checking in on https://docs.google.com/document/d/1XJ6_9ZHJ7yluAOA8DMhxmdDhLJZn8HaA5ADokZLpbr8/edit?tab=t.0 , are the prereqs done? I'm mainly curious about the maintenance scripts
[14:42:07] <inflatador>	 sounds like y'all are actively working on the indices part ;)
[14:42:23] <ebernhardson>	 atsukoito: ahh if david's looking i'm sure you'll find something :)
[14:42:49] <atsukoito>	 I think I either switch testwiki to k8s-only and we can test in ~prod~staging, or we need to set up a dev env with it :)
[14:42:57] <ebernhardson>	 inflatador: looking
[14:43:17] <dcausse>	 inflatador: yes I think cirrus maint scripts should now support opensearch2, striking
[14:43:26] <ebernhardson>	 inflatador: yes the patch from david is merged to allow maint scripts
[14:43:32] <ebernhardson>	 and should be deployed
[14:44:54] <dcausse>	 we still have translate I guess it's the last blocker?
[14:45:08] <ebernhardson>	 yea from this list that should be it
[14:45:13] <dcausse>	 will have to dig into the query to understand why
[14:45:40] <ebernhardson>	 i was hoping for a dump query option, but no luck.  Can maybe get one from shell.php
[14:46:04] <dcausse>	 it relies on fuzzy like & levenshtein distance something we maintain could well be a bug/regression there?
[14:46:13] <ebernhardson>	 yea it could be
[14:46:20] <dcausse>	 yes translate does not have the debug tools we're used to :/
[15:01:56] <inflatador>	 dcausse ACK thanks, it sounds like there are no more blockers to 2.x migration on the Search Platform side? LMK if I'm wrong here
[15:06:07] <dcausse>	 inflatador: we still have to move translate away, we had to rollback today because it's not yet working with opensearch2 in dse-k8s
[15:07:27] <dcausse>	 added a quick note in the doc
[15:09:37] <dcausse>	 I'll run another extraction of the index version
[15:11:16] <dcausse>	 atsukoito: is it OK to delete toolhub_* indices from the cirrussearch cluster?
[15:11:49] <atsukoito>	 dcausse: yes, toolhub_* is migrated
[15:11:58] <dcausse>	 thanks, dropping then
[16:14:37] <atsukoito>	 dcausse: do you need any help reproducing/debugging the issue?
[16:15:03] <dcausse>	 atsukoito: sorry was busy on something else, I'll dig into this now
[16:15:52] <atsukoito>	 I'm planning to check out for the day, ping me up, I'll read everything first thing in the morning
[16:19:51] <dcausse>	 sure!
[17:05:41] <dcausse>	 hm.. using top-queries I found some translate queries, replaying them just work :/
[17:06:18] <ebernhardson>	 :S Often at that point i start a shell.php session and start fake-debugging (step by pasting the next line of code :P)
[17:06:32] <dcausse>	 yes :)
[17:12:15] <dcausse>	 results are identical (minus the _type field) between prod and dse-k8s... good news is that dse-k8s seems a lot faster (1.5s vs 7.3s for prod)
[17:17:11] <ebernhardson>	 interesting! I wouldn't have expected that
[17:20:01] <dcausse>	 same, I doubt fuzzy like is more performant, could perhaps be the translate index having hard time to compete with prod search to keep its data in mem on the prod cluster?
[17:20:29] <ebernhardson>	 hmm, yea that seems highly plausible
[17:20:52] <ebernhardson>	 7s though, that's quite a bit of waiting.
[17:21:13] <dcausse>	 yes...
[17:28:41] <dcausse>	 well, not finding anything obvious playing with opensearch directly... will have to go with the manual mw-script debug session :/
[17:28:45] <dcausse>	 dinner
[19:20:26] <inflatador>	 Has anyone ever tried to regenerate indices in beta cluster? Looks like mwscript is available: https://wikitech.wikimedia.org/wiki/Nova_Resource:Deployment-prep/Overview
[19:59:52] <ebernhardson>	 inflatador: should work fine, if you mean ForceSearchIndex.php.  We use it all the time for dev
[20:01:00] <inflatador>	 ebernhardson ACK, was trying to remember the right script. thanks