[06:58:33] 10DBA, 13Patch-For-Review: Rampant differences in indexes on enwiki.revision across the DB cluster - https://phabricator.wikimedia.org/T132416#3060297 (10Marostegui) So, db2034 finished and as per our discussion yesterday evening it is currently like this: ``` PRIMARY KEY (`rev_id`,`rev_user`), KEY `rev_ti... [07:33:05] 10DBA, 06Operations, 10ops-eqiad: Degraded RAID on db1053 - https://phabricator.wikimedia.org/T151465#3060322 (10Marostegui) a:03Cmjohnson Hey Chris! Would you have time to replace this disk today? Thanks! [07:35:29] 10DBA, 06Operations, 10ops-eqiad: Degraded RAID on db1060 - https://phabricator.wikimedia.org/T158193#3029364 (10Marostegui) [07:43:39] 07Blocked-on-schema-change, 10DBA, 06Community-Tech, 06Stewards-and-global-tools (Temporary-UserRights): Schema changes for expiring user groups - https://phabricator.wikimedia.org/T155605#3060335 (10Marostegui) s4 is done. As it had the primary key, this is what was run (on the master - db1040): ``` ./osc... [08:20:35] 07Blocked-on-schema-change, 10DBA, 06Community-Tech, 06Stewards-and-global-tools (Temporary-UserRights): Schema changes for expiring user groups - https://phabricator.wikimedia.org/T155605#3060372 (10Marostegui) s5 (dewiki and wikidatawiki) are done. dewiki only needed the column and the index as the PK wa... [08:37:53] 07Blocked-on-schema-change, 10DBA, 06Community-Tech, 06Stewards-and-global-tools (Temporary-UserRights): Schema changes for expiring user groups - https://phabricator.wikimedia.org/T155605#3060395 (10Marostegui) s6 (frwiki,jawiki and ruwiki) done, only needed the column and the key, the PK was already ther... [08:39:57] Can I add T155605 to the deployments page? [08:39:59] T155605: Schema changes for expiring user groups - https://phabricator.wikimedia.org/T155605 [08:40:21] jynus: I didn't add it because it wasn't a long running change [09:06:42] 07Blocked-on-schema-change, 10DBA, 06Community-Tech, 06Stewards-and-global-tools (Temporary-UserRights): Schema changes for expiring user groups - https://phabricator.wikimedia.org/T155605#3060415 (10Marostegui) s1 (enwiki) done, only needed the column and the index: ``` ./osc_host.sh --host=db1052.eqiad.... [09:11:49] s5 problems on dbstore2001? [09:13:02] checking [09:13:13] A delete with type ALL [09:13:20] probably a missing index [09:13:22] uh?? [09:13:35] oh i see [09:15:01] no, same schema [09:17:02] but for some reason it is not using wb_changes_change_time [09:17:09] on other slaves it does [09:19:31] yes, I see no reason for it, as it is a pretty simple delete [09:19:44] simple as in it is just using one index (or it should) [09:20:21] I would run analyze when the query finishes [09:21:56] yeah, let's do that [09:22:11] actually, it is not one query [09:22:13] but many [09:22:17] I am going to stop the slave [09:22:28] even if that means contention for a few minutes [09:22:33] all deletes actually [09:22:44] "stop the slave" [09:26:58] fixed, I think [09:27:23] it will take a few minutes for s5 to catch up [09:27:48] it looks good now yes :) [09:27:49] thanks [09:27:57] analyze worked [09:28:07] for some reason it was reejecting the index [09:28:43] hey, not all bugs are that bad, that require months of debugging! [09:29:04] hahaha [09:29:26] and it is much easier than with tokudb! [09:30:00] yeah, I still remember: https://jira.mariadb.org/browse/MDEV-10796 [09:43:10] 07Blocked-on-schema-change, 10DBA, 06Community-Tech, 06Stewards-and-global-tools (Temporary-UserRights): Schema changes for expiring user groups - https://phabricator.wikimedia.org/T155605#3060532 (10Marostegui) s7 done for all wikis (centralauth doesn't have that table). Only the column and index was need... [10:16:16] 07Blocked-on-schema-change, 10DBA, 06Community-Tech, 06Stewards-and-global-tools (Temporary-UserRights): Schema changes for expiring user groups - https://phabricator.wikimedia.org/T155605#3060614 (10Marostegui) s2 is done for all wikis except l10nwiki, which doesn't have that table. The PK was there in al... [10:20:47] 07Blocked-on-schema-change, 10DBA, 06Community-Tech, 06Stewards-and-global-tools (Temporary-UserRights): Schema changes for expiring user groups - https://phabricator.wikimedia.org/T155605#3060617 (10TTO) >>! In T155605#3060614, @Marostegui wrote: > s2 is done for all wikis except l10nwiki, which doesn't h... [10:21:45] 07Blocked-on-schema-change, 10DBA, 06Community-Tech, 06Stewards-and-global-tools (Temporary-UserRights): Schema changes for expiring user groups - https://phabricator.wikimedia.org/T155605#3060620 (10Marostegui) >>! In T155605#3060617, @TTO wrote: >>>! In T155605#3060614, @Marostegui wrote: >> s2 is done f... [10:55:24] 07Blocked-on-schema-change, 10DBA, 06Community-Tech, 06Stewards-and-global-tools (Temporary-UserRights): Schema changes for expiring user groups - https://phabricator.wikimedia.org/T155605#3060718 (10Marostegui) So, to sum up: s1 done s2 done s3 pending s4 done s5 done s6 done s7 done I will start with s... [11:17:20] 07Blocked-on-schema-change, 10DBA, 06Community-Tech, 06Stewards-and-global-tools (Temporary-UserRights): Schema changes for expiring user groups - https://phabricator.wikimedia.org/T155605#3060763 (10Marostegui) For s3 I am doing batches of wikis, starting with those which don't have the PK (129 wikis in t... [12:44:12] 07Blocked-on-schema-change, 10DBA, 06Community-Tech, 06Stewards-and-global-tools (Temporary-UserRights): Schema changes for expiring user groups - https://phabricator.wikimedia.org/T155605#3060871 (10Marostegui) All the 129 wikis in s3 that didn't have the PK have been modified, so they now have the new PK... [12:50:13] hey yall, when I see old_text = DB://cluster24/6158167 in the text table, where can I get that text from? [12:52:07] 07Blocked-on-schema-change, 10DBA, 06Community-Tech, 06Stewards-and-global-tools (Temporary-UserRights): Schema changes for expiring user groups - https://phabricator.wikimedia.org/T155605#3060901 (10Marostegui) I am also filtering the wikis that do not have that table in s3: ``` zh_twwiki|zh_cnwiki|wikim... [12:52:09] marostegui ^? [12:52:28] milimetric: hi! where do you see that? [12:53:14] I got a rev_text_id from revision and did "select * from text where old_id = 16291975;" [12:55:02] marostegui: ^ [12:55:10] and you get: DB://cluster24/6158167 ? [12:55:29] yes, it's saying the revision text is stored externally [12:55:39] but I didn't know where that is [12:55:48] https://noc.wikimedia.org/conf/highlight.php?file=db-eqiad.php [12:56:09] look for 'cluster24' => [ [12:56:12] milimetric: yes, that is stored on the es servers [12:56:21] oh wow, what are es servers? [12:56:45] look what Reedy said :-) [12:56:50] also, Reedy, you're an encyclopedia man [12:56:53] haha [12:56:53] thank you [12:57:06] It's for all of the blobs [12:57:16] well, most [12:57:17] right, but where/how do I get to them? [12:57:25] Depends how you're trying [12:57:30] MW itself will do it all nicely [12:57:37] oh, right, so there's a problem :) [12:57:40] somewhat self-inflicted [12:57:42] If you're doing the sql queries yourself... You need to connect to one of the hosts yourself [12:57:55] You could poll the MW API or similar [12:57:59] I deployed a new extension that hid some pages from MW [12:58:31] they used to be on Config:Something and we configured Config:Dashiki: to be a sub-namespace of Config:, so now all those pages aren't visible anymore [12:58:38] so the API can't get to them [12:58:43] milimetric: namespaceDupes ? [12:59:17] I heard there was a script yeah that moved stuff, is that what you mean? I didn't want to run it because I didn't want to migrate all of them [12:59:48] but when I tried this on beta, it didn't have this effect of hiding the page [12:59:58] https://phabricator.wikimedia.org/P4999 [13:00:15] There's various config pages that look broken, but no Config:Dashiki [13:00:40] right, Config:Dashiki would be the destination [13:01:05] For all of those? [13:01:16] well some I'd want to just delete [13:01:18] Does MW actually support subnamespaces? Or does it just go lolno [13:01:59] I don't know the details of that, Yuri's extension JsonConfig allows you to configure "sub-namespaces" but I don't know if they're an actual first-class concept or like made up [13:02:22] I guess they're made up 'cause they're not in the db or anything [13:02:25] You're having to ask? :P [13:03:12] I'm embarassingly new to mediawiki extensions and how all this functions [13:04:12] ok, so is there a way to pull text out of cluster24? [13:04:36] I was just going to manually make the 6 configs I need to migrate and deal with the broken pages later [13:06:34] we're getting what we need out of dumps right now, just wondering if there's another way [13:06:35] milimetric: There's a load of pages in NS 0... With a Dashiki prefix [13:06:57] yes, those are going to be deleted too, after I fix all this [13:07:52] As for your actual question [13:07:52] sql metawiki -h 10.64.16.186 [13:08:00] Should only be one table, blobs_cluster24 [13:08:28] so then select blob_text from blogs_cluster24 where blob_id = 6158167 [13:08:42] from tin? [13:08:56] well, less blogs, more blobs [13:09:05] yep, tin, thx! [13:09:06] yup [13:09:17] Of course [13:09:22] That may or may not actually help you [13:09:31] Cause it's compressed [13:09:42] yep, that's ok, it's good to know how [13:09:48] I need this for a lot of upcoming work [13:10:49] so after I migrate, is the right way to fix all the old pages that namespaceDupes script? [13:11:11] Depends what you mean by "migrate" :) [13:11:18] I fixed all the Config: ones that were in NS 0 [13:11:40] So you should be able to interact with them more easily onwiki now [13:11:57] Or do you mean the Dashiki prefixed ones? [13:12:21] In that case... If you use MW to move them to Config:Dashiki, MW will automatically put them in the correct NS in the database [13:13:53] Reedy: ok, I'll try that, thanks for fixing them! [13:17:46] 07Blocked-on-schema-change, 10DBA, 06Community-Tech, 06Stewards-and-global-tools (Temporary-UserRights): Schema changes for expiring user groups - https://phabricator.wikimedia.org/T155605#2948316 (10Reedy) >>! In T155605#3060901, @Marostegui wrote: > I am also filtering the wikis that do not have that tab... [13:17:57] 07Blocked-on-schema-change, 10DBA, 06Community-Tech, 06Stewards-and-global-tools (Temporary-UserRights): Schema changes for expiring user groups - https://phabricator.wikimedia.org/T155605#3060930 (10Marostegui) We have 753 wikis (yay) to alter that already have the PK on s3. I will do that in batches. The... [13:20:20] Reedy: ok, so I tried to move just one of them and it looks like it doesn't account for the new content model of that namespace: https://meta.wikimedia.org/w/index.php?title=Config:Dashiki:LanguageReportcard&action=edit [13:21:01] so it's ok, I'll just move them manually by creating the new pages and flagging the old ones for deletion [13:21:07] https://meta.wikimedia.org/wiki/Config:Dashiki:LanguageReportcard [13:21:17] Should be fixable using https://meta.wikimedia.org/wiki/Special:ChangeContentModel [13:21:24] I guess it was never set [13:21:29] Let me login with my WMF account [13:21:40] oh ok, can I get access to that? [13:21:53] guess I'd have to be an Administrator [13:21:55] Do you have a WMF account with staff/sysop or similar rights on it? [13:22:05] not on meta, just on mediawiki [13:22:20] Reedy: but don't fix it, there will be a bunch of others if I move them [13:22:22] I was meaning global rights :) [13:22:46] Just seems silly moving them out of the way, to delete them, to recreate them when you should be able to fix it [13:23:26] I agree, but is it ok to put me in Administrators? I just don't want to cause you more work [13:24:03] Well, that's special [13:24:07] "The action you have requested is limited to users in one of the groups: Administrators, WMF Support and Safety." [13:24:10] Why can't my account do it [13:24:19] oh yeah, I get that too [13:24:27] if you can't do it... it means it's broken :) [13:24:36] Well, that's with my "staff" account [13:24:46] What's your username on meta? [13:25:11] Milimetric (WMF) [13:26:34] milimetric: done [13:26:37] Granted myself the rights to check [13:26:46] MW doesn't know it's another content... [13:26:56] thanks Reedy :) [13:27:01] So setting them to JSON you should be good to go [13:27:17] you should never touch the external storage directly [13:27:36] we use maintenance scripts for that [13:27:45] nice, works well [13:27:56] oh, I was never going to update jynus, just read the text [13:28:05] but Reedy fixed it anyway so I can do everything I need on MW [13:28:22] thanks very much all, this has been 1. enlightening 2. useful and 3. fun [13:28:23] <3 [13:28:26] prefer mediawiki wrappers, there is lot of old stuff to do things directly [13:28:45] Hence my suggestion to use the MW API to query it :) [13:28:53] yes [13:29:55] even when there is some mixup, it is always to do the least amount of stuff manually [13:30:11] *better [13:35:59] hm, Reedy the JSON content model works ok, but it's not the same as the one for newly created pages in that namespace, check out the difference between https://meta.wikimedia.org/wiki/Config:Dashiki:MultimediaHealth and https://meta.wikimedia.org/wiki/Config:Dashiki:Sample/tabs [13:36:34] https://meta.wikimedia.org/w/index.php?title=Config:Dashiki:MultimediaHealth&action=edit [13:36:38] The content format application/json+pretty is not supported by the content model JSON. [13:36:49] The content format application/json+pretty is not supported by the content model JSON. [13:37:02] right, so it's still broken I guess [13:37:13] Sounds like it [13:37:26] Presumably, Config or something isn't setting it correctly on creation? [13:37:42] Though.. [13:37:46] You've changed that one [13:37:52] right, kind of tangled at this point [13:41:07] | page_id | page_namespace | page_title | page_content_model | [13:41:07] +----------+----------------+-----------------------------+--------------------+ [13:41:08] | 9959595 | 482 | Dashiki:FlowByWiki | json | [13:41:08] | 8387927 | 482 | Dashiki:LanguageReportcard | json | [13:41:08] | 8551857 | 482 | Dashiki:MultimediaHealth | json | [13:41:11] | 10203443 | 482 | Dashiki:Sample/tabs | JsonConfig.Dashiki | [13:41:20] yes, was just looking at that [13:41:39] so Special:ChangeContentModel won't let me do a custom content_model [13:41:43] Should the extension be registering that contentmodel somewhere? [13:42:30] It configures it but I guess I didn't know about Special:ChangeContentModel or how to register a model for it [13:42:37] ContentHandlerDefaultModelFor and/or ContentHandlerForModelID [13:43:24] Oh.. Is this JsonConfigModels [13:43:34] Which we had fun and games with Zero extensions for [13:43:44] Is this related to https://phabricator.wikimedia.org/T125885 ? [13:43:46] sounds right, but again, really out of my depth, just used the JsonConfig extension and then Yuri got let go so I got lost... [13:43:56] no jynus [13:44:06] not related to flow [13:44:10] I mean the same issue [13:44:41] oh, uh doesn't sound like it [13:45:16] milimetric: I think I can make a patch that should fix it [13:45:29] Reedy: what are you thinking? [13:46:24] "JsonConfigModels": { [13:46:25] "Dashiki": { [13:46:25] "view": "Dashiki\\DashikiView" [13:46:25] } [13:46:25] }, [13:46:29] I think something like that [13:46:48] would this have to be deployed to meta somewhere? [13:46:52] Based on the fun and games we had with Zero and extension registration [13:47:00] Needs to go into gerrit, merged/cherry picked, deployed from tin [13:47:05] Easy enough to do [13:47:50] so I need to fix these dashboards, and I don't want to put you on the hook to do an ad-hoc deploy because I screwed up [13:47:59] so I can just move / create / delete my way out of this just fine [13:48:35] up to you, really [13:48:51] Might aswell get it fixed properly while it's fresh [13:48:55] https://gerrit.wikimedia.org/r/340310 [13:49:56] milimetric: Even better, it's the SWAT window in 10 :) [13:50:05] ok, merged [13:50:22] Does jenkins not run on that repo? You shouldn't V+2 generally ;) [13:51:31] oops, it runs, my bad [13:51:41] heh [13:51:53] I have a good habit of committing invalid json [13:52:08] I checked it first (man, with morning brain I do everything manually) [13:52:23] sorry I woke up to this little outage, not at my best [14:09:30] Reedy: so now that's deployed but what does it all mean? :) [14:10:05] should I move the pages and move them back again? [14:12:26] I just noticed in your README, you suggest to do that in config [14:13:46] I wonder if JsonConfig really isn't setup for helping to change the config of a page properly [14:14:09] the config has... https://github.com/wikimedia/operations-mediawiki-config/blob/9d42afdff9ff9ba9c89ba39ddc9224d89c14e9fe/wmf-config/CommonSettings.php#L3185-L3199 [14:14:30] yeah, it's possible, without yurik I'm not sure [14:16:52] milimetric: If in doubt, just fix it in the DB [14:16:52] update page set page_content_model = 'JsonConfig.Dashiki' where page_namespace = 482 AND page_title LIKE 'Dashiki%'; [14:17:06] https://meta.wikimedia.org/wiki/Special:ChangeContentModel?pagetitle=Config%3ADashiki%3AMultimediaHealth [14:17:10] Wrong link [14:17:14] https://meta.wikimedia.org/w/index.php?title=Config:Dashiki:MultimediaHealth&action=edit [14:17:16] oooh, I know someone whose name starts with a j that might be mad at us :) [14:17:33] milimetric: Have a look at how the maintenance script in core does it :) [14:17:45] $dbw->update( [14:17:45] 'page', [14:17:45] [ 'page_content_model' => CONTENT_MODEL_JSON ], [14:17:45] [ 'page_id' => $row->page_id ], [14:17:45] __METHOD__ [14:17:46] ); [14:17:48] Reedy: so I still have to move a bunch of other pages, should I just do that and then run the update again? [14:17:51] :) [14:17:56] Yeah, that's what I'd do [14:18:09] I'm gonna file a bug about not being able to change the content model to these JsonConfig "sub types" [14:18:14] But I doubt anyone is likely to fix it [14:18:15] ok, but I don't think I have rights to run that update, will you be around for a little longer? [14:18:27] yeah, I was going to say, I'd file a task but who would work on it [14:18:47] I need to go out to the pharmacy, but won't be too long so can always run it in a bit for you [14:18:52] oh! never mind, I can just create / delete instead of move for the rest [14:19:21] unless you object [14:19:40] Well, are all those in hte Config namespace yours? [14:20:09] everything in Config: I think is mine, and anything in Dashiki: is mine [14:20:34] but a lot of that stuff just has to get deleted anyway [14:20:46] yeah man, don't worry, I got this, thanks very much for the help, I know a lot more [14:20:56] I was just going to suggest we can just change all of the config ns stuff to the dashiki content model [14:21:25] nah, it's ok, more than half has to be deleted anyway [14:21:54] heh [14:22:52] and it's not like history for these things matters much, thanks again, really [14:54:55] jynus, marostegui: any objections against upgrading apache now on dbmonitor*? (sec update) [14:55:15] no, go on [14:55:19] ok [14:55:51] :) [14:56:08] can you stop it completely for a second? [14:56:15] me? [14:56:25] ^moritzm [14:56:28] ah [14:56:59] (after upgrade, I mean) [14:57:07] there is something I want to check [14:57:32] upgraded, shall I stop it now on 1001? [14:57:37] yes, please [14:57:39] ping me when done [14:57:52] stopped [14:58:04] yeah, it is not in production [14:58:07] you can start it [14:58:09] ok [14:58:27] I could check the config, but this is easier :-) [14:58:37] and more reliable [14:58:56] :-) [14:58:58] we should move ther, soon [15:25:08] we have a few "Table definition has changed, please retry transaction (10.64.16.191)" [15:25:21] about user_groups [15:26:02] where? [15:26:37] on the mw logs [15:26:40] maybe that was a race condition? [15:26:44] when the alter was going thru? [15:27:00] (it is still ongoing) [15:27:31] (only pending s3) [15:27:54] I can only see db1077 (s3) and not enwiki which was altered in the morning, so maybe it was just when the alter was being done [15:28:36] it seems only 35 in the last 24 hours [15:29:11] yeah and the last ones only s3 [15:29:27] I would note it in the ticket for the owner to have it a look, in case it was not just temporary [15:29:36] good point [15:29:39] I will do that now [15:29:58] better overreport than underreport [15:30:54] I was looking closely at the logs, because I am agraid my changes could also break stuff [15:30:59] *afraid [15:31:17] but the rate is so small that it is difficult to say [15:40:09] 07Blocked-on-schema-change, 10DBA, 06Community-Tech, 06Stewards-and-global-tools (Temporary-UserRights): Schema changes for expiring user groups - https://phabricator.wikimedia.org/T155605#3061407 (10Marostegui) From those 753 pending wikis, 308 have already been altered in different batches. We have seen... [16:15:49] 10DBA, 10MediaWiki-Categories, 03Community-Tech-Sprint, 13Patch-For-Review: Increase size of categorylinks.cl_collation column - https://phabricator.wikimedia.org/T158724#3061473 (10thiemowmde) > What are your thoughts on my comments at T158724#3048124 and T158724#3054267? As I said: just use a hash funct... [16:18:56] 07Blocked-on-schema-change, 10DBA, 06Community-Tech, 06Stewards-and-global-tools (Temporary-UserRights): Schema changes for expiring user groups - https://phabricator.wikimedia.org/T155605#3061475 (10Marostegui) No errors in the last hour since I stopped the ALTERs - looks good. [16:26:13] 10DBA, 13Patch-For-Review: Deploy gtid_domain_id flag in our mysql hosts - https://phabricator.wikimedia.org/T149418#3061502 (10JAllemandou) Hi @Marostegui We have questions around that task ass it prevents us to move forward in productionizing our data extraction pipeline (one major goal for us this quarter /... [16:44:04] 10DBA, 13Patch-For-Review: Deploy gtid_domain_id flag in our mysql hosts - https://phabricator.wikimedia.org/T149418#3061581 (10Marostegui) >>! In T149418#3061502, @JAllemandou wrote: > Hi @Marostegui > We have questions around that task ass it prevents us to move forward in productionizing our data extraction... [17:14:44] 10DBA, 13Patch-For-Review: run pt-table-checksum before decommissioning db1015, db1035,db1044,db1038 - https://phabricator.wikimedia.org/T154485#3061679 (10Marostegui) The following wikis were checksummed in s2 during the day with no issues: bgwiki - no differences bgwiktionary - no differences cswiki - no di... [17:51:52] 10DBA, 06Community-Tech, 10MediaWiki-Categories, 13Patch-For-Review: Increase size of categorylinks.cl_collation column - https://phabricator.wikimedia.org/T158724#3061779 (10DannyH) [19:10:34] 10DBA, 06Community-Tech, 10MediaWiki-Categories, 13Patch-For-Review: Increase size of categorylinks.cl_collation column - https://phabricator.wikimedia.org/T158724#3062053 (10kaldari) >If this is a relevant issue you must change the column type to integer. You can create a lookup table for all known collat... [19:46:49] ema: Question...I see this from time to time in logstash and wondering what it's indicative of... [19:47:00] Warning: Unable to record MySQL stats with: EXPLAIN /* CheckUser::doUserIPsRequest */ SELECT * FROM `cu_changes` WHERE cuc_ip_hex = '' AND (1 = 1) [19:47:10] (not always CheckUser, but it seems to be the most common offender) [19:49:14] Oh, hmm, _joe_...you filed https://github.com/facebook/hhvm/issues/3760 ~3y ago [20:03:18] 10DBA: Investigate db1047 replication lag - https://phabricator.wikimedia.org/T159266#3062225 (10jcrespo) [20:45:11] 10DBA, 06Community-Tech, 10MediaWiki-Categories, 13Patch-For-Review: Increase size of categorylinks.cl_collation column - https://phabricator.wikimedia.org/T158724#3062467 (10Bawolff) > I'm not worried about collisions. I'm worried about bloating an already huge table with more data for no practical reason... [22:35:35] 10DBA: Investigate db1047 replication lag - https://phabricator.wikimedia.org/T159266#3062951 (10Marostegui) It caught up before I was able to check it. I saw it 13 minutes delayed but when I got it to the terminal it was already up to date. Did you see which thread was lagging? S1 or S2? I will investigate furt...