[10:58:48] <ozge_>	 Hello folks, This is Ozge from the ML Team, and I’m working on the add-a-link model.
[11:01:05] <ozge_>	 Can we query the current add-a-link recommendations here in Elastic by using a UI e.g. Kibana, Logstash and what is the name of the index? Page I refer to: https://en.wikipedia.org/w/index.php?search=hasrecommendation%3Alink+articletopic%3Abiography&title=Special%3ASearch&ns0=1&searchToken=7jnqtn2my5wsn8irmvphrjfdb and log stash I’m looking into: https://wikitech.wikimedia.org/wiki/Logstash and more information about add 
[11:01:05] <ozge_>	 a link model: https://wikitech.wikimedia.org/wiki/Add_Link
[11:29:52] <pfischer>	 ozge_: Hi! I am not the expert here, dcausse should be back in a bit. Until then: I am not aware of a UI for our Elastic/OpenSearch cluster serving search results. There’s an HTTP API though, you can use: https://wikitech.wikimedia.org/wiki/Help:CirrusSearch_OpenSearch_replicas
[11:31:22] <pfischer>	 The information backing hasrecommendation is a weighted tag, AFAIK, see https://www.mediawiki.org/wiki/Extension:CirrusSearch/Schema#weighted_tags
[12:04:45] <dcausse>	 ozge_: no we don't have a kibana in front of the search cluster, is there anything specific you're looking into? 
[12:06:37] <dcausse>	 you can access the "cloudreplica" of the search indices (https://wikitech.wikimedia.org/wiki/Help:CirrusSearch_OpenSearch_replicas)
[12:07:04] <dcausse>	 in theory I suppose you could setup your own opensearch-dashboard/kibana to connect to
[12:07:10] <dcausse>	 it
[12:10:00] <dcausse>	 but note that I'm not sure that our index setup is kibana "friendly"
[12:32:59] <ozge_>	 Thank you very much Peter and David for the quick reply. Add-a-link project recommends some links to the articles and enhancing it with topics then saves the enhanced articles to Elastic.These articles are the ones I shared in the link above. My objective is to check what type of recommendations we have available (number of recs, their topics, etc.) in scope of this issue https://phabricator.wikimedia.org/T393474 . I can 
[12:32:59] <ozge_>	 also look into an older dump or an analytics database if we have a dump somewhere. But probably I should find the index name first.
[12:38:09] <dcausse>	 ozge_: do you have access to hadoop?
[12:38:57] <ozge_>	 yes
[12:39:35] <dcausse>	 ok, prepping a quick snippet to show how to access this data from the search index dumps we have there
[12:40:06] <ozge_>	 Awesome, thank you very much David.
[13:14:22] <inflatador>	 <o/
[13:16:09] <dcausse>	 o/
[13:20:14] <btullis>	 Hello. Quick question: how bad would it be if I were to kill this task and allow it to restart? https://airflow-search.wikimedia.org/dags/mjolnir_weekly/grid?task_id=feature_vectors-norm_query-20180215-query_explorer&tab=details&dag_run_id=scheduled__2025-05-15T00%3A00%3A00%2B00%3A00
[13:21:14] <btullis>	 I'm in the middle of a rolling reboot of the dse-k8s-eqiad Kubernetes cluster (for T394897) and I am having trouble draining the node where this task is running.
[13:21:15] <stashbot>	 T394897: [Cephfs] Clients occasionally fail to release caps, resulting in blocked requests and Airflow service disruption - https://phabricator.wikimedia.org/T394897
[13:22:34] <btullis>	 I can see that it's usually a 12 hour task , so I don't know whether it would cause a user-facing issue if I were to kill it.
[13:22:42] <dcausse>	 btullis: should be fine
[13:23:27] <btullis>	 OK, many thanks. I will add a `--force` to my `kubectl drain` command and try again.
[13:38:56] <dcausse>	 ozge_: please see: https://phabricator.wikimedia.org/P76406
[14:11:33] <ozge_>	 This is awesome David! Thank you! I’ll try it now.
[14:59:38] <gehel>	 Andreas will be facilitating our retro today!
[15:01:07] <gehel>	 dcausse: ^
[15:01:17] <gehel>	 retrospective: https://meet.google.com/eki-rafx-cxi
[15:01:21] <dcausse>	 gehel: sorry I have a conflict
[15:01:23] <gehel>	 pfischer: ^
[15:03:11] <pfischer>	 gehel: i'll be 10' late, picking up Luise
[15:03:31] <gehel>	 pfischer: we'll start without you! Join when you can!
[15:50:00] <ebernhardson>	 Trey314159: i added the ac/ft targetpage split, and put back in some of the lines i was eliding, i think it gets far too busy :P https://phabricator.wikimedia.org/F60380122
[15:57:29] <Trey314159>	 ebernhardson: I kind of like it. The overall structure of the majority of the flow is still very clear. All the little spiderweb lines can be ignored to get a good overview, but I like to look at them and think about the weird flows those people following.
[15:58:12] <Trey314159>	 The FT/AC Target split is very nice. No surprises, but it's good to see concretely how their outflows differ.
[16:14:00] <inflatador>	 workout, back in ~40
[17:10:08] <dcausse>	 dinner
[18:26:40] <inflatador>	 sorry, came back and then went to lunch. back!
[21:27:36] <ebernhardson>	 final-ish report: https://people.wikimedia.org/~ebernhardson/T392525-Sankey-Exploration-2025-04.html
[21:37:24] <Trey314159>	 ebernhardson: I will check out your report tomorrow! Now I'm off to dinner...