[05:36:16] jynus: db2097 migrated to 10.6 [07:27:09] ok, will run a backup to test it [07:29:45] great [09:54:04] marostegui: any suggestion on https://phabricator.wikimedia.org/T334650#9050440 ? I can keep s6 backup sources eqiad on 10.4 (and so, its backups) or upgrade them already? [10:01:20] As I don't think I am a blocker, and it seems it may still take a few more days for a 100% migration, I will leave eqiad backups on 10.4 for now, but ready to migrate- as it seems the most conservative action [10:02:24] so doing https://phabricator.wikimedia.org/T334650#9047479 but just short of the actual upgrade [10:45:12] yeah I'm going to migrate the candidate on Monday [10:45:30] and give it 48h before doing the master switch [10:50:38] thats for codfw, right? [10:50:54] or you mean eqiad? [11:04:15] yeah codfw [11:05:05] just replied in phab for the equiad comment [11:06:10] thank you, will be ready by then [12:11:06] another heads up for Amir1 that a new instance of s6 has appeared (db1225:s6) https://grafana.wikimedia.org/goto/ePgwqA34k?orgId=1 [12:12:03] new instances everywhere [12:12:20] technically, moved, but this is a 1 time thing because upgrade [12:12:55] jynus: just to check, you are updating zarcillo with all these changes or you want me to? [12:13:04] I am doing it [12:13:12] ah thanks <3 [12:13:16] otherwise prometheus wouldn't work :-D [12:13:20] as it depends on that [12:13:28] but if the host already exists it would [12:13:32] Ah, but it is a different port, yeah [12:13:44] what I may not be doing is handling orch, I am not too familiar with that [12:13:52] altough I think I was told it was not a big deal? [12:14:45] that's fine yeah [12:14:51] It would delete itself after a few days [12:14:54] If I don't do it first [12:15:00] And it will show up by itself too, so that's fine [12:15:18] do you remember what major db change happened in august/july/sep 2020? [12:15:31] I don't even remember what I did 2 weeks ago [12:15:49] https://usercontent.irccloud-cdn.com/file/YW7H3Ah6/grafik.png [12:15:52] haha [12:15:53] Amir1: unsure but I would guess the first normalization [12:16:21] MCR maybe? [12:16:49] there is a small drop before that for MCR in archive table [12:17:10] Amir1: the stats contain per-file stats [12:17:21] so you may be able to get more info [12:17:26] there is an SQL query for that [12:17:46] aah, I actually did that for s4 [12:17:53] https://phabricator.wikimedia.org/P49455 [12:17:54] because it's growing wayyy tooo fast [12:18:24] just change the jobids there [12:18:49] https://usercontent.irccloud-cdn.com/file/IaRrYeWv/table.png [12:19:04] this is for commons [12:19:48] I can make it run for enwiki [12:20:04] you can also check the differences between dump and the snapshots to avoid index or optimization differences [12:20:36] yeah I have dumps too but they also don't take account structural changes that were important (dropping index, etc.) [12:21:03] yeah, plus they are compressed, which hides some of the redundancies [12:21:45] after this graph for s4, I'll go talk to the community to cut off some of the templatelinks and categorylinks growth, I think there is a bit of an issue there. e.g. stuff like https://commons.wikimedia.org/wiki/Category:M%C3%A9rim%C3%A9e_ID_same_as_Wikidata [12:22:54] jynus: funnily enough, I think backup_files needs normalizing of wiki id and file name, it's 32GB, queries take a looong time :P [12:23:09] ha ha ha [12:23:19] in theory it should be trimmed after 3 months [12:23:26] good that I didn't [12:23:46] Yup, they are mountain of useful information [12:23:59] also the idea of backup_objects was to create some automation to aggregate per object (e.g. table, trigger, db) but was never done [12:24:44] I think adding indexes should be enough, but certainly the db hasn't been optimized for analytics (yet) [12:25:48] the other extension was to run the same scripts over a live db to compare and check backups more thoroughly [12:26:50] * Emperor watches their laptop try and examine 2 billion thumbnail records [12:27:10] xD [12:27:14] big data [12:29:50] https://usercontent.irccloud-cdn.com/file/lWpOUK4m/table.png [12:29:52] for enwiki [12:29:58] looks like contemporary art [12:31:17] that looks like MCR on revision [12:35:48] found it T238966 [12:35:49] T238966: Apply updates for MCR, actor migration, and content migration, to production wikis. - https://phabricator.wikimedia.org/T238966 [13:05:29] MariaDB 10.6 backups working well, x1 and s6 worked well [13:05:44] 🎉 [13:21:51] Amir1: +1 on those graphs being art :) [15:43:46] who thought databases could become a source of art [15:43:51] you learn a new thing every day [16:02:02] :)