[01:44:11] <wikibugs>	 10DBA, 10Wikidata, 07Performance: 5-10 times more rows are currently read from the s5 master than from all other masters together - https://phabricator.wikimedia.org/T146529#2664113 (10hoo) >>! In T146529#2664889, @aaron wrote: > Also seeing: >…  That has been fixed with 37a2c43273571af41ea38ad39c84f7cdde869...
[07:35:40] <wikibugs>	 10DBA, 06Operations, 10ops-eqiad: db1060: Degraded RAID - https://phabricator.wikimedia.org/T146449#2666237 (10Marostegui) All good now!  ```                 Device Present                 ================ Virtual Drives    : 1   Degraded        : 0   Offline         : 0 Physical Devices  : 14   Disks...
[07:36:57] <wikibugs>	 10DBA, 06Operations, 10ops-eqiad: db1060: Degraded RAID - https://phabricator.wikimedia.org/T146449#2666238 (10Marostegui) 05Open>03Resolved
[07:59:39] <addshore>	 hey jynus! just checking, on db tables do we 1) always need an auto increment id, that is a primary key OR 2) always need a primary key ?
[08:11:32] <jynus>	 addshore, 2
[08:11:41] <addshore>	 okay! many thanks! :)
[08:12:00] <jynus>	 in some cases it may be an auto_increment, but prefer a natural one if it makes sense
[08:12:06] <jynus>	 and it is small
[08:12:54] <jynus>	 3 varchar(255) == large; 2 ints == small 
[08:21:05] <addshore>	 jynus: how about 1 VARBINARY(255), 1 VARBINARY(32) and 1 int?
[08:21:44] <jynus>	 addshore, in that say I would suggest, create at ticket, and discuss it
[08:21:57] <jynus>	 it always is an "it depends"
[08:22:17] <jynus>	 there may be reasons to go for it
[08:22:41] <jynus>	 s/say/case/
[08:22:52] <addshore>	 okay, well right now this will be for a new extension so I'll initially just choose one or the other, but changing it before it would go near production should be easy enough! I can alos create a ticket, the create is at https://gerrit.wikimedia.org/r/#/c/312257/2/db/addCognateTitles.sql currently
[08:22:58] <jynus>	 it will be on a case by case scenario
[08:23:42] <jynus>	 I see, add me to the review and make sure I review it
[08:23:58] <jynus>	 I may need some background about it
[08:23:59] <addshore>	 okay! :)
[08:24:34] <jynus>	 for example, if that was an n:n table full of foreign keys
[08:24:39] <jynus>	 it would make sense
[08:24:46] <jynus>	 but again, it depends on the context
[08:25:26] <addshore>	 yeh, so if there was an auto inc id instead of the composite primary key described above it would essentally never be used for selecting, also also would never be a foreign key.
[08:25:44] <jynus>	 that doesn't necesarrily rules it out
[08:25:51] <jynus>	 because it may be used for table maintenance
[08:26:05] <jynus>	 there are some general tips
[08:26:14] <jynus>	 like trying to avoid that if it is not used
[08:26:38] <jynus>	 but then there are some edge cases for pure implementation side
[08:26:51] <jynus>	 again, I will need some context
[08:27:06] <jynus>	 I am busy these week, can this wait a bit?
[08:27:10] <jynus>	 *this
[08:28:12] <addshore>	 well, im actively working on it now and I expect to go through quite a few patches in the next week, but again, as none of the code I see myself writing will depend on an id auto inc right now I can ave the primary key eaither way and easily switch to the best before deployment! :)
[08:28:26] <jynus>	 can you send me a summary mail
[08:28:41] <jynus>	 with the general idea of the table (it can be ticket links)?
[08:28:43] <addshore>	 I'll file a ticket for review the db schema for the extension once I have the first version merged?
[08:28:51] <jynus>	 and I will review it this week
[08:28:54] <jynus>	 ok?
[08:28:55] <addshore>	 okay!
[08:29:04] <addshore>	 thanks! (as always)!
[08:29:07] <jynus>	 sorry, this week because the offsite
[08:29:15] <addshore>	 ahh yes!
[08:29:15] <jynus>	 I am a bit distracted with pure ops work
[08:46:34] <jynus>	 "<addshore> I'll file a ticket for review the db schema for the extension once I have the first version merged?"
[08:46:48] <jynus>	 Note the PK is a must, no matter the format
[08:46:55] <addshore>	 yup! :)
[08:47:11] <jynus>	 I wouls suggest in case of doubt, use the natural key
[08:47:19] <jynus>	 but add it from the start
[08:47:32] * addshore will never introduce a table with no PK, especially after knowing the pain adding wl_id has caused etc.
[08:47:33] <jynus>	 otherwise schema changes will be impossible
[08:47:37] <jynus>	 exactly
[08:47:39] <jynus>	 :-)
[08:47:47] <addshore>	 [=
[10:27:32] <wikibugs>	 10DBA, 10Continuous-Integration-Infrastructure, 10MediaWiki-Database: Enable MariaDB/MySQL strict mode on CI slaves - https://phabricator.wikimedia.org/T119371#2666528 (10hashar)
[13:50:05] <wikibugs>	 10DBA, 06Operations, 10ops-eqiad: db1060: Degraded RAID - https://phabricator.wikimedia.org/T146449#2667177 (10Marostegui) thanks
[15:30:46] <wikibugs>	 10DBA, 10CatWatch, 10MediaWiki-General-or-Unknown, 06TCB-Team, and 4 others: SELECT /* CategoryMembershipChangeJob::run 127.0.0.1 */ GET_LOCK('CategoryMembershipUpdates:XXXX', 10) AS lockstatus - https://phabricator.wikimedia.org/T133801#2667501 (10hashar) 05Open>03Resolved a:03aaron Fixed by https:/...