2022-04-19T01:34:14 LCP0968[m], cboltz I'm getting random database errors when trying to read threads, new posts etc 2022-04-19T07:37:43 progress.o.o seems a bit slow this morning? 2022-04-19T08:24:12 pjessen, do you still see duplicates in https://mirrorcache.opensuse.org/update/leap/15.3/sle/repodata/970feef70cbd786e4312437d7b2b59286370e6cb58ea88f3d56639c8727a7cc5-primary.xml.gz.mirrorlist ? 2022-04-19T08:25:22 I did see duplicates in past, but iirc they gone relatively quickly. I have an idea how to fix it, but since it is mostly cosmetic problem - I give it low priority 2022-04-19T08:31:13 *** Papabears[m] is now known as Papabears9969[m] 2022-04-19T08:41:33 anikitin: let me check 2022-04-19T08:42:17 anikitin: yes, it looks like every entry is duplicated :-) 2022-04-19T08:43:46 not for me :) , what if you append ?COUNTRY=de in url ? In any case - I have the bug, I will have a look thx 2022-04-19T08:44:34 I'll try with country - what is also "funny" is that I don't see any Swiss mirrors, none. My own mirror is being scanned though. 2022-04-19T08:45:29 with COUNTRY=de, I see no duplicates 2022-04-19T08:46:37 with COUNTRY=it or da, I have duplicates 2022-04-19T08:52:41 COUNTRY=fr - no dupes, but otherwise be/nl/no/se/fi etc 2022-04-19T09:38:01 the wiki has problems connecting to the DB 2022-04-19T09:38:02 https://en.opensuse.org/ 2022-04-19T09:38:36 I also got reports of "slowness" for events.o.o 2022-04-19T09:38:46 anything up with the cluster? 2022-04-19T09:40:21 henne: progress.,o.o has also been slow all morning 2022-04-19T09:42:52 * henne uploaded an image: (14KiB) < https://libera.ems.host/_matrix/media/r0/download/matrix.org/gwotAtFMxqqlVoBZRPFpvfwv/Screenshot%20from%202022-04-19%2011-41-18.png > 2022-04-19T09:42:57 this is events.o.o 2022-04-19T09:45:04 cboltz was having some issues yesterday, but he is not here yet 2022-04-19T09:45:32 he was working on the wiki, some update 2022-04-19T10:36:50 so is anyone looking into this? 2022-04-19T10:39:01 * henne twists some arms 2022-04-19T11:40:02 likewise, forum (well for me) still having database issues 2022-04-19T11:46:09 cboltz: darix found out it's the wiki eating all connection slots 2022-04-19T11:46:29 cboltz: sounds like fallout from the upgrade? 2022-04-19T11:48:15 any heroes available? :D 2022-04-19T11:48:25 *stabs cboltz a bit* 2022-04-19T11:49:11 in case you wonder why barely anything works that uses mysql: the wiki eats up all the connections 2022-04-19T11:49:54 so if someone could rcapache2 stop on the VM(s) that would be awesome 2022-04-19T11:50:16 Hello team, progress-o-o is down https://progress.opensuse.org/projects/opensuse-leap-15-4/issues/gantt 2022-04-19T11:50:23 I suppose it's related to ^ 2022-04-19T11:50:35 it is 2022-04-19T11:53:17 progress has been slow all morning, but it's working 2022-04-19T11:54:44 pjessen: then you were lucky and wiki gave you some connections :D 2022-04-19T11:55:24 basically all galera nodes have 100% cpu load and the wiki seems to trigger it 2022-04-19T11:56:01 mysqladmin processlist | grep -c wiki_ 2022-04-19T11:56:01 583 2022-04-19T11:57:06 holy shit 2022-04-19T11:57:13 why is there sql traffic on more than one node 2022-04-19T11:58:12 583 - I'm sure we don't have that many apache workers 2022-04-19T11:58:37 so the wiki has connections stuck at least on node 1 and 2 ... that also explains why i saw cluster resume events from earlier in the day 2022-04-19T11:58:51 pjessen: can we turn off the wikis for a moment? 2022-04-19T11:59:10 yeah, just logged in 2022-04-19T11:59:34 because i can tell you it clogs up the whole galera cluster right now 2022-04-19T11:59:51 ]have stopped apache on riesling 2022-04-19T12:00:24 hmm, interesting[tm] 2022-04-19T12:00:35 the most likely reason is the updated mediawiki on en.o.o 2022-04-19T12:00:50 all those connections are stuck in "sending data" 2022-04-19T12:01:23 I guess they might time out, eventually 2022-04-19T12:02:45 darix-: the wikis connect only to one node (via haproxy), therefore "more than one node" might mean it _writes_ to the database (which then gets synced to all nodes) 2022-04-19T12:03:36 purged all wiki connections 2022-04-19T12:03:50 all the queries were select 2022-04-19T12:04:21 tons of '(SELECT /* SpecialRecentChangesLinked::doMainQuery...)' 2022-04-19T12:04:46 only one "SELECT /* MediaWiki\Revision\RevisionStore::getAuthorsBetween" on galera2 2022-04-19T12:05:14 galera 1 only had that SpecialRecentChangesLinked::doMainQuery 2022-04-19T12:05:27 for galera3 i didnt grab the list before purging connections 2022-04-19T12:06:31 can you paste the full SpecialRecentChangesLinked query? 2022-04-19T12:06:40 its a bit weird for SELECTs to be getting hung up - any I/O issues? 2022-04-19T12:06:49 "(SELECT /* SpecialRecentChangesLinked::doMainQuery */ rc_id,rc_timestamp,rc_namespace,rc_title,rc_" 2022-04-19T12:07:00 that is all i see in mysqladmin processlist 2022-04-19T12:07:10 looks cut off :-( 2022-04-19T12:07:32 pjessen: well one reason could be ... that table runs over a large table and e.g. indexes are missing our outdated 2022-04-19T12:07:40 so each query does a full table scan 2022-04-19T12:07:58 s/that table/that query/ 2022-04-19T12:07:59 right, would be an explanation 2022-04-19T12:08:08 e.g. any migrations which might not have been run? 2022-04-19T12:08:20 which would have created the index 2022-04-19T12:08:23 the output of the update script looked good 2022-04-19T12:08:37 let me start apache again, and then "show full processlist" so that we can find out what causes the load 2022-04-19T12:09:08 mysqladmin kill $(mysqladmin processlist | awk '/wiki_/ {print $2}' | tr '\n' ',') 2022-04-19T12:09:11 just in case :) 2022-04-19T12:09:52 hehe ;-) 2022-04-19T12:10:59 no hanging queries for wiki_en so far... 2022-04-19T12:11:26 ah, managed to catch it, and it's quite long 2022-04-19T12:12:04 https://paste.opensuse.org/80160298 2022-04-19T12:12:37 with tons of sub queries :D 2022-04-19T12:12:40 good luck! 2022-04-19T12:12:49 wow, that is long .... 2022-04-19T12:19:24 essentially 3 xc the sdame query, with different where 2022-04-19T12:21:28 main tables involved, 'change_tags' and 'recent_changes' 2022-04-19T12:25:43 explain (separate for each of the 3 UNIONed queries): https://paste.opensuse.org/view/raw/22966656 2022-04-19T12:26:08 looks like the first part misses an index 2022-04-19T12:28:32 how big are those tables? 2022-04-19T12:29:17 doug notified me tsp is broken, I should probably have a look (it's unrelated to galera issues, it's running on pg) 2022-04-19T12:30:54 recentchanges has about 1000 rows, comment (renamed to comment_rc_comment in the query) about 57000 2022-04-19T12:32:50 interestingly, comment.comment_id is PRIMARY KEY (`comment_id`), so that key should get used in the JOIN - but explain says that despite PRIMARY would be a possible key, no key gets used 2022-04-19T12:34:12 its not a lot of rows 2022-04-19T12:35:06 right, but I still wonder why the primary key doesn't get used - especially because that results in "Using temporary; Using filesort" 2022-04-19T12:35:51 any issue with space on galera? 2022-04-19T12:36:55 583 hanging queries mush have been waiting for something. 2022-04-19T12:37:29 /var/lib/mysql-files on galera2 only has 1.6G free 2022-04-19T12:38:09 hmm, not a lot, but sounds okay 2022-04-19T12:40:35 no one will ever need more than 640k disk space! 2022-04-19T12:41:14 no idea what caused the queue, and "of course" now that I look at it, there is _no_ hanging/waiting apache process and _no_ hanging query 2022-04-19T12:46:50 heisenbug 2022-04-19T13:00:28 at least my P95s are down to 2s 2022-04-19T13:03:04 progress.o.o is working too :-) 2022-04-19T13:05:36 I have to leave 2022-04-19T13:05:47 if the problem appears again, please save the output of ps Zaux before killing apache so that we can see if it's really the english wiki or another one 2022-04-19T13:06:52 forums are better too, thanks all 2022-04-19T13:11:25 you can tell him all connections were to wiki_en 2022-04-19T13:11:36 (in the mysqladmin processlist output) 2022-04-19T13:16:35 Hi.. Who can add some announcements to status.opensuse.org? 2022-04-19T13:16:49 lethliel: pjessen: might be able to 2022-04-19T13:29:36 Thanks 2022-04-19T13:42:45 Conan Kudo (ニール・ゴンパ) you tried fixing tsp, what did you actually find out? 2022-04-19T13:43:02 I did? 2022-04-19T13:43:08 no I tried to use it and it broke all over me 2022-04-19T13:43:11 that's what doug told me 2022-04-19T13:43:14 oh 2022-04-19T13:43:18 that's fun 2022-04-19T13:43:34 let's debug this then, i have no idea how to trigger the error 2022-04-19T13:47:12 I would love to actually debug this but my vpn connection seems to freeze for long periods of time 2022-04-19T14:01:46 that's your mandatory burnout protection 😉 2022-04-19T14:02:30 I feel like it will cause me burnout faster than fix anything 2022-04-19T14:11:53 I can't access reimbursements, Doug can't set the right settings so that I can actually add my bank info to the reimbursement filing, etc. 2022-04-19T14:11:53 it throws 500 errors a lot 2022-04-19T14:12:00 hhhh 2022-04-19T14:12:15 I really have no sympathy for leap staying on ruby 2.5 2022-04-19T14:12:26 updating dependencies is straight up impossible 2022-04-19T14:12:35 LCP0968[m]: you can happily run every ruby version in parallel 2022-04-19T14:12:36 just fine 2022-04-19T14:12:50 and we can easily build you all ruby extensions for it as well 2022-04-19T14:12:51 well, yes 2022-04-19T14:12:57 I do have to install them 2022-04-19T14:13:08 guess how henne runs OBS on ruby 3.1 2022-04-19T14:13:09 which means I need to get the repo 2022-04-19T14:13:45 I like that the first result on build.o.o for ruby is crystal project 2022-04-19T14:13:46 i would accept complains about ruby if it werent so easy to run multiple versions in parallel 2022-04-19T14:14:04 even packaging rails apps is super easy 2022-04-19T14:14:09 (if anyone wants to learn it) 2022-04-19T14:14:12 well, I'm complaining about leap 2022-04-19T14:14:13 I love ruby, it's great 2022-04-19T14:14:18 leap kinda blows though 2022-04-19T14:14:28 well you can run your services on TW if you want 2022-04-19T14:14:33 i do that for all my servers 2022-04-19T14:14:47 including the stuff we host under https://pixls.us 2022-04-19T14:14:58 we don't really have tw in our infra 2022-04-19T14:15:09 who is stopping you from doing it? 2022-04-19T14:15:11 we should probably run all of our rails servers with tw 2022-04-19T14:15:18 LCP#0968: I'll add TSP to my list... 2022-04-19T14:15:47 henne did the software-o-o upgrade in super fast time 2022-04-19T14:15:50 if i recall correctly 2022-04-19T14:16:05 there is a starting point for that, though idk how valuable that is anymore https://github.com/openSUSE/travel-support-program/pull/121 2022-04-19T14:16:16 LCP0968[m]: and the problem for running different ruby versions which is the non default ruby is the same on leap as it is on TW 2022-04-19T14:16:18 since we are on rails 7 2022-04-19T14:16:39 well, for tumbleweed I can just use the default version 2022-04-19T14:16:43 sure 2022-04-19T14:17:20 in any case ... https://build.opensuse.org/package/view_file/home:darix:apps/discourse/discourse.spec?expand=1 2022-04-19T14:17:30 if you are curious about rails app packaging 2022-04-19T14:17:41 (the gitlab package is pretty much the same) 2022-04-19T14:18:01 yeah 2022-04-19T14:18:08 runtime deps are autogenerated based on Gemfile.lock 2022-04-19T14:18:11 we don't package tsp in obs yet 2022-04-19T14:18:26 with hard equal requires 2022-04-19T14:18:28 mostly because it was a quick migration from connect server 2022-04-19T14:19:04 in any case ... getting ruby stuff fixed seems easier than having leap ship newer python modules :) 2022-04-19T14:19:13 your matrix server also wants TW 2022-04-19T14:19:23 oh it really does 2022-04-19T14:19:34 I did try backporting, that didn't go well at all 2022-04-19T14:19:36 no more py 3.6 support 2022-04-19T14:19:47 yeah 2022-04-19T14:37:55 henne: couldnt tsp just be merged into osem? 2022-04-19T14:39:04 OSEM already suffers from tons of unmaintained features... 2022-04-19T14:39:16 I had some TSP replacement 2022-04-19T14:40:38 I think I showed Gerald... 2022-04-19T14:43:59 https://cobudget.com/ 2022-04-19T14:45:02 they seem to be pretty far 2022-04-19T14:45:34 I guess any other tool for participatory processes could be used too 2022-04-19T14:46:41 like https://meta.decidim.org/processes or https://www.loomio.com/about