[08:20:48] <gmodena>	 o/
[08:45:52] <gmodena>	 popping out for a dentist appointment
[09:10:14] <dcausse>	 o/
[09:55:11] <gmodena>	 dcausse do you usually test CirrusSearch (config) changes in Beta, or go straight to prod traffic?
[09:55:50] <dcausse>	 gmodena: we generally test while deploying on mw test servers
[09:56:46] <gmodena>	 dcausse ack - thanks
[10:56:19] <dcausse>	 lunch
[11:49:56] <gmodena>	 dcausse no rush - I tagged you in chain of three patches that attempt to deploys 2024 models and enable a mlr-2025 A/B test. One bit that's not clear is how to handle kowiki/zhwiki in https://gerrit.wikimedia.org/r/c/operations/mediawiki-config/+/1118782/1
[11:51:18] <gmodena>	 btw - I claimed T385972. Shall I assign some proxy estimate (to be refined on monday) before dragging it to "In progress"?
[11:51:19] <stashbot>	 T385972: Deploy and test new MLR models - https://phabricator.wikimedia.org/T385972
[12:45:50] <dcausse>	 gmodena: looking
[12:46:40] <dcausse>	 re: estimate, no need to estimate imo, we can do that next monday if you want
[12:51:20] <gmodena>	 thanks!
[12:51:27] <gmodena>	 sounds good re estimation
[13:10:48] <gmodena>	 lunch+meetings block
[13:55:36] <gehel>	 gmodena: estimations are there to ensure we have a common understanding of the work to be done. I think that on T385972, what we need to do and how to do it is reasonably clear, so estimation is less important anyway.
[13:55:36] <stashbot>	 T385972: Deploy and test new MLR models - https://phabricator.wikimedia.org/T385972
[14:18:52] <gmodena>	 gehel sounds good!
[14:19:46] <inflatador>	 <o/
[14:20:28] <gmodena>	 o/
[14:24:29] <gehel>	 Hey team! I'm being disruptive! I'm introducing regular milestones to our work, similar to what we do on the SRE team. This means that our current workboard is now https://phabricator.wikimedia.org/project/view/7690/. It looks very similar to the previous one :)
[14:25:11] <gehel>	 I'll create a new milestone every ~3 weeks. The current one can be found as a link in the menu of our main backlog board (https://phabricator.wikimedia.org/tag/discovery-search).
[14:26:27] <gehel>	 This makes it easier for me to get visibility on what was completed when.
[14:27:26] <gehel>	 On that note, you can now close the tasks when moving them to the "done" column, and get that endorphin rush! The default filter on the board is to show all tasks.
[14:28:37] <gehel>	 I should probably have talked to you all about the change before rushing forward. But forgiveness, permissions and all that. Hopefully this isn't too disruptive, and you'll find a way to forgive me if it is!
[15:41:44] <inflatador>	 g-ehel is closing so many tickets, I can't read my Phab comments ;)
[15:48:27] <gehel>	 just moving closed tickets around. I did not think that phab would send notifications for closed tickets!
[15:48:51] <gehel>	 I might make more noise!
[16:36:38] <gmodena>	 dcausse thanks for the reviewes! Do you usually deploy cirrus  yourself, or got through a MW backport window? Could we pair on this deployment in case?
[16:37:24] <dcausse>	 gmodena: I always use MW backport window, happy to pair on this one, tomorrow morning is OK for you?
[16:38:08] <gmodena>	 dcausse yep, that'd work for me.
[16:49:09] <dcausse>	 gmodena: just scheduled the patch (https://schedule-deployment.toolforge.org/window/1739347200)
[16:50:29] <gmodena>	 dcausse thanks! I'll set a reminder :)
[16:51:01] <gmodena>	 dcausse do we also roll out the a/b test, or would you rather do that post mlr deployment?
[16:53:27] <dcausse>	 gmodena: as you want, I thought you were waiting for models trained today
[16:59:05] <dcausse>	 gmodena: sorry, realized that I had pending comments on one of the patches
[17:00:02] <gmodena>	 dcausse np!
[17:00:27] <dcausse>	 what we can do is ship https://gerrit.wikimedia.org/r/c/operations/mediawiki-config/+/1118783 tomorrow as well, do some tests and enable the A/B as a second step, perhaps on the afternoon window tomorrow
[17:00:50] <gmodena>	 dcausse +1 
[17:00:53] <gmodena>	 let's do that.
[17:01:10] <dcausse>	 ok scheduling the second one too
[17:01:35] <gmodena>	 i'll update the a/b test name and add today's models to the first patch in the chain.
[17:01:44] <dcausse>	 +1
[17:03:08] <gmodena>	 oof. I spoke too early. Looks like mjolnir has been stuck for a while: https://airflow-search.wikimedia.org/dags/mjolnir_weekly/grid?search=mjolnir_weekly&dag_run_id=scheduled__2025-01-17T18%3A42%3A00.449096%2B00%3A00&task_id=feature_selection-20180215-query_explorer-dbn-20180215-query_explorer-pruned_mrmr
[17:04:21] <gmodena>	 20250110 are the latest models we have uploaded. Let's go with those and try to fix the training pipeline in parallel?
[17:04:33] <dcausse>	 gmodena: sounds good
[17:05:10] <dcausse>	 sent a quick invite for tomorrow
[17:05:44] <gmodena>	 thanks!
[17:08:30] <dcausse>	 errand+dinner, back later tonight
[17:14:42] <gmodena>	 dinner
[17:32:33] <inflatador>	 workout, back in ~40
[18:33:20] <inflatador>	 back
[18:50:00] <gmodena>	 o/
[19:03:46] <inflatador>	 lunch, back in ~40
[19:30:13] <inflatador>	 back
[19:39:34] <inflatador>	 oops, forgot I have physical therapy today. Back in ~1h
[21:25:20] <inflatador>	 back