[00:32:54] [1/3] what [00:32:54] [2/3] https://cdn.discordapp.com/attachments/1006789349498699827/1570959676517130250/image.png?ex=6af8fdb5&is=6af7ac35&hm=a850726487227381ee742e75a448703b94ae6fc86ae3a490ca9a9dc2ac5c804c& [00:32:55] [3/3] https://cdn.discordapp.com/attachments/1006789349498699827/1570959676957397023/image.png?ex=6af8fdb6&is=6af7ac36&hm=7ef8ae6e5213cfc7a56485e66a19ae9656425fc668bc27d1daea22bdd60a84b0& [00:38:58] @honoka55 anyway done [00:56:36] oh yeah this has happened before [00:57:08] https://issue-tracker.miraheze.org/T15799 [01:07:08] Ah, I see, it's happening because main in mw-config has beta at 1.47, but the branch I switched to has it at 1.46 [01:07:33] Still, it's an issue that ultimately needs to be resolved, I suppose [01:22:17] it happens regardless of mw-config iirc [01:28:53] It has to do with mw-config. All that script does is run `MirahezeFunctions::getMediaWikiVersion( 'metawikibeta' );` [01:31:30] then why doesn't that happen when running that exact same function in shell.php? [01:31:53] Because then it uses the version from the command line [01:32:16] [1/10] ``` [01:32:17] [2/10] if ( PHP_SAPI === 'cli' ) { [01:32:17] [3/10] $parts = explode( '/', $_SERVER['SCRIPT_NAME'] ); [01:32:17] [4/10] $version = $parts[3] ?? null; [01:32:17] [5/10] if ( $version && in_array( $version, self::MEDIAWIKI_VERSIONS, true ) ) { [01:32:18] [6/10] return $version; [01:32:18] [7/10] } [01:32:18] [8/10] } [01:32:19] [9/10] ``` [01:32:19] [10/10] I assume anyway [01:33:34] oh I guess the original thing was about a 1.45 wiki that incorrectly said 1.46 whereas this is a 1.47 wiki incorrectly saying 1.46 [01:34:40] Yeah, it could use some improvement regardless but if isn't a real issue on-wiki so I never really wanted to spend more time in it. I built the whole multiversion really poorly tbh, I need to rework it at some point. [05:20:44] i've mended the pr. does it need to be deployed again? or i can just go on? [10:48:50] i've finished the two prs. could someone plz deploy them to beta? [10:52:55] Will do [10:56:41] [1/2] Done. [10:56:42] [2/2] @suzuneu I stashed your changes in `mw-config` with comment `Waki's CentralAuth fixes`. [11:00:23] [1/3] yeah it's working https://exttest.mirabeta.org/wiki/User:Honoka55 [11:00:23] [2/3] https://cdn.discordapp.com/attachments/1006789349498699827/1571117587901849691/image.png?ex=6af990c6&is=6af83f46&hm=e8342c80f3a01c353737879a43266c1b7ac059a2192692e0ce45448229b153b6& [11:00:24] [3/3] https://cdn.discordapp.com/attachments/1006789349498699827/1571117588476330058/image.png?ex=6af990c7&is=6af83f47&hm=f41188cc77672c2e97a4a56e8c63d746584b8509b6989764bcb288bc3898d965& [11:02:07] just need approval to merge them. open to reviews [11:47:20] Thank you. I wanted to test something in beta that didn't quite warrant a PR to mw-config, and I wasn't really sure how I should handle it after deploying it. [11:55:40] [1/2] I feel like most things in the stash can be dropped because they are from months ago, but the list is manageable so it should be fine to stash more temporary changes. [11:55:40] [2/2] https://cdn.discordapp.com/attachments/1006789349498699827/1571131498948595752/image.png?ex=6af99dbb&is=6af84c3b&hm=39646f60c8eaf1b89cac3179222215e9c93fb31e11d3411408f1d0307d62e6b0& [12:20:28] the bucket ones at least can be dropped [12:47:29] [1/2] https://issue-tracker.miraheze.org/T16515 [12:47:29] [2/2] I tried doing normal index rebuilding but that doesn't work on crashed tables apparently. Based on what I could find a `REPAIR TABLE searchindex;` would resolve that? [12:49:48] I'm thinking there shouldn't be any issues with doing any SQL hocus pocus on the table, since it's all generated by the maintenance script anyway. But would like a +/-1 on that take [20:44:10] when this happened for the WMF they recreated it (by dropping it and then creating it again). That also sounds sensible to me, as it doesn't hold any data that couldn't be rebuilt [22:52:10] I'll take a look [22:53:00] it's db161 so that's a slight concern [22:53:09] Hopefully nothing else is corrupt [22:59:01] That wiki is buggered [23:00:00] I mean tbf [23:00:16] does it surprise anybody that a database that OOMs every other day has some corruption [23:00:35] The script is up to templatelinks at the moment [23:00:51] I guess I'm spending tomorrow morning check it's not just this wiki [23:00:56] No [23:01:20] It isn't just searchindex though [23:02:16] I mean as long as it's not data that can't be reconstructed [23:03:16] It might be [23:04:37] It seems to be fixable corruption though [23:07:43] @cosmicalpha are you around? [23:07:57] Yes [23:08:30] @jenny_on_wiki @abaddriverlol I'm half asleep so can one of you run rebuild search index on that wiki once I've finished repairing the db [23:08:50] if you mean the searchindex issue that started right after the last OOM on db161 so was likely caused by that [23:08:51] @cosmicalpha I think we need to run mysqlcheck on db161, happy for RW or read only? [23:09:03] yes, ping me when you want me to do that [23:09:08] Yes it's 100% caused by OOM but I'm more concerned it's more than 3 tables [23:09:16] MariaDB isn't supposed to keep crashing [23:09:32] We can go read only if necessary. [23:10:01] @abaddriverlol wait a few moments please [23:10:06] I'm asking if it is [23:11:28] @abaddriverlol go for it please [23:11:39] We do not have to if the server can handle it. Also isn't it called mariadb-check now? [23:12:03] Yes MariaDB would like me to retrain my brain on the name [23:12:15] I'm leaning towards it probably won't [23:12:32] started, but might take a while because there are 140k pages [23:12:34] And to put it RO MW side tomorrow morning while we perform checks [23:12:56] Because it's quite possible it's badly buggered [23:13:04] And will lock things up unexpectedly [23:14:15] Just to note that wiki is the one that has caused at least one of the OOMs recently, and I believe it did the last one as well. [23:14:42] Though there was no signs of corruption last I looked. [23:14:51] So that changed after the last time [23:14:57] Do we know what extension caused it (maybe post this on MM if you know)? Surely MW didn't cause it on its own? [23:15:35] There were 3 tables corrupt on that wiki [23:15:57] I could run check and see how bad the rest of the db is in the morning [23:16:12] But if it's widespread, go RO then to rebuild everything [23:16:23] If not, probably just do individual wikis [23:16:46] Does that sound like a plan? [23:20:11] @cosmicalpha [23:20:28] yes [23:20:32] sounds good [23:20:46] Ok [23:21:01] Am I okay to announce possible disruption tomorrow morning then @cosmicalpha [23:21:40] yes, make sure at least one other is around when you do it in case it does cause other issues though please. [23:23:39] @paladox will you be awake by 9am tomorrow and around for a few hours just in case? [23:24:32] i won't be around by 9 [23:24:43] what's you needing? [23:25:00] Someone to be able to help if MySQL does anything really weird [23:25:03] @abaddriverlol ? [23:25:09] 9 am in what timezone [23:25:11] utc? [23:25:12] oh, are you deploying something [23:25:14] yeh [23:25:15] UTC [23:25:18] uk is utc+0 now [23:25:33] Running mysqlcheck and fixing any issues I find [23:25:57] i could be online at that time if I force myself to go to bed at a reasonable time today [23:26:30] I'm about to go to bed so I wake up too [23:26:44] if you want you can run it now if you gives you some sort of comfort (just putting it out there that mysqlcheck shouldn't take anything down). [23:27:10] mysqlcheck is the easy bit. More if we find hard to repair corruption and I need advice. [23:27:17] oh i see [23:27:32] Hopefully most of it either self heals or is like searchindex and can be thrown away [23:28:10] I will start MySQL check at 9am and ping you two to see if you're awake when I know how fun it's going to be