[13:06:00] \o [13:06:15] I should've thought of this Friday, but I'm gonna to base the OpenSearch image off the releng java image instead of stock bookworm [13:15:22] seems reasonable [13:55:12] * ebernhardson should better document how to release all the things...so many steps [13:55:17] s/document/automate/ [14:01:33] i wonder what we would want...the annoyance is i merge to wmf-jvm-utils, then do a release, then release cirrus-highlighter and search-extra, then update the plugins .deb repo and release the .deb, then update the opensearch container repo with the .deb and release the new container, then update the integration environment repo with the container reference [14:01:36] too many sequential steps :P [14:02:59] I guess automation counts as a form of documentation... [14:03:20] imo automation is better than documentation in most cases :P It's like documentation you can run [14:05:15] yeah, plus it's constantly being tested, documentation usually isn't ;( [14:06:22] Automation is often written in a more opaque language than documentation, though [14:06:36] Now if it were in COBOL... that's practically English! [14:06:55] * inflatador likes where this is going [14:07:44] lol [14:12:55] updated a/b test code is out, in theory we should actually be running the fuzziness AB test now [14:31:38] Taking my son to the doctor...will miss triage [15:01:40] hmm, my google meet is just spinning... [15:02:49] yea just spins even after restarting...lemme try firefox [15:49:28] ebernhardson: not quite what you were talking about wrt to measuring several things in A/B tests but it made me think of this recent Numberphile video: https://www.youtube.com/watch?v=FUQwijSDzg8 [15:49:47] maybe i can find the page...sec [15:51:21] Trey314159: i think this one was interesting: https://briefer.cloud/blog/posts/p-hacking/ [15:51:27] "So instead of working with a 5% false positive rate, you’re actually closer to 18.5%: nearly a 1 in 5 risk that you're shipping something based on a fluke." [15:52:01] and related hn discussion: https://news.ycombinator.com/item?id=44308328 [15:52:28] looks interesting.. I will give both a read... [15:56:25] i guess the part i was trying to do better on was "Pre-register hypotheses and metrics", decide the metrics up front instead of while i'm looking at the data and writing the analysis [16:06:45] back [17:00:50] * ebernhardson is midly annoyed we have cirrus-integration-test-runner, cirrus-reindex-orchestrator, and then cirrussearch-opensearch-image. It make my tab completion wonky :P [17:03:19] ebernhardson: the discussion on hn is more informative and more entertaining than the article. The article has some egregious math errors (1 - (1-0.05)^9) is both 64 and 37. It's neither.. it's 0.37 or 37%. (I think the 64 formula is a cut-n-paste error, but still..)... Anyway, there are lots of strong opinions along many different dimensions in the comments... it's easy to find ones I agree with! [17:04:13] lol, indeed. Certainly i have to side with the people saying your software is not a medical trial, and you can be a bit less rigorous [17:05:12] err. sigh....just realized i released search-extra and cirrus-highlighter to gitlab instead of maven central...now to remember how to do that [17:21:29] hmm, 401 unauthorized :S and the username/password in my password manager for oss.sonatype.org don't work [17:22:26] oh, i guess i haven't been paying attention: "As of June 30, 2025 OSSRH has reached end of life and has been shut down. All OSSRH namespaces have been migrated to Central Publisher Portal." [17:51:57] apparently central publishing portal is stricter, to release we need to specify a property in all the poms: https://gerrit.wikimedia.org/r/c/search/extra/+/1169216 [18:42:57] lunch, back in ~40 [18:53:02] oh...sigh [18:53:15] the name field ends up in the plugin path....meaning i shouldn't have used spaces :S [18:54:41] or i need to override that somewhere, it seems cirrus-highlighter already does because it has a name with spaces but the updated plugin is still cirrus-highlighter [19:30:00] ryankemper we may want to look at https://gerrit.wikimedia.org/r/c/operations/puppet/+/1169208 during pairing today [19:32:12] Agreed! [20:00:44] hmm, never seen that before. Deployment 02d4a183-7a44-460d-8777-27d5763784c8 has been validated. To finish publishing visit https://central.sonatype.com/publishing/deployments [20:00:54] going to be annoying :P [20:01:09] or maybe thats just status ...unclear