21:00:01 #startmeeting swift 21:00:02 Meeting started Wed Sep 9 21:00:01 2020 UTC and is due to finish in 60 minutes. The chair is timburke. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:00:03 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:00:05 The meeting name has been set to 'swift' 21:00:12 who's here for the swift meeting? 21:00:28 hi 21:00:31 o/ 21:01:28 o/ 21:02:04 not too much on the agenda; mostly end-of-cycle wrap-up 21:02:05 https://wiki.openstack.org/wiki/Meetings/Swift 21:02:11 #topic ptg 21:02:31 the new release notes looks amazing! nice work everyone 21:02:48 reminder to sign up at https://www.eventbrite.com/e/project-teams-gathering-october-2020-tickets-116136313841 21:03:11 clayg's been doing a great job of adding topics to the etherpad 21:03:14 #link https://etherpad.opendev.org/p/swift-ptg-wallaby 21:03:55 and if you haven't yet, please respond to the doodle; i'll try to have meeting times scheduled this week 21:03:59 #link https://doodle.com/poll/ukx6r9mxugfn7sed 21:04:22 i did the doodle! and I have a ticket for the VPTG - but eventbrite still seems to let me want to register (again?) 21:04:42 clayg, i think you're good then ;-) 21:04:57 oh i'm GREAT 😉 21:05:04 i still need to do the registration iirc... i'm slacking off 21:05:14 any questions or comments on the summit/ptg? 21:07:20 all right then 21:07:25 #topic releases 21:07:39 it's that time in the cycle again! 21:08:23 i've got some draft release notes for swift at p 750537 -- please take a look and let me know if there's anything that should be corrected or clarified 21:08:24 https://review.opendev.org/#/c/750537/ - swift - Authors/ChangeLog for 2.26.0 - 1 patch set 21:09:15 i ought to get some client release notes going, too -- but i was hoping we'd be able to add in a fix for https://bugs.launchpad.net/python-swiftclient/+bug/1873435 21:09:17 Launchpad bug 1873435 in python-swiftclient "Established connection is never closed on Python 2.7" [High,In progress] - Assigned to Tim Burke (1-tim-z) 21:09:53 i believe it should be resolved by p 736401 21:09:54 https://review.opendev.org/#/c/736401/ - python-swiftclient - (Mostly) revert "Cleanup session on delete" - 3 patch sets 21:10:28 and then p 721051 more-explicitly manages some connections 21:10:29 https://review.opendev.org/#/c/721051/ - python-swiftclient - Close connections created when calling module-leve... - 5 patch sets 21:10:52 anybody think they'd have time to take a look at thos in the next day or so? 21:12:06 clayg, you'd at least taken a look at them once upon a time ;-) 21:12:32 yeah but I hate that entire series of "fixes" 😡 21:13:49 if someone else could load it into their head and either confirm or deny some of the insanity that would probably be helful 21:14:23 i never particularly liked it going back to p 674320 so.... 🤮 21:14:24 https://review.opendev.org/#/c/674320/ - python-swiftclient - Cleanup session on delete (MERGED) - 2 patch sets 21:15:50 isn't anytime you have to add a __del__ like "doing it wrong"? 21:17:24 that's my gut reaction, yeah. part of why i was ready to just rip it back out 21:18:44 ok, well p 736401 doesn't look so bad - i'm not sure what's going on in p 721051 21:18:45 https://review.opendev.org/#/c/736401/ - python-swiftclient - (Mostly) revert "Cleanup session on delete" - 3 patch sets 21:18:47 https://review.opendev.org/#/c/721051/ - python-swiftclient - Close connections created when calling module-leve... - 5 patch sets 21:19:27 I'm guessing they're separate because they're independently mergeable - and we'd be happier getting one of them than neither? 21:19:53 i broke them up because you didn't seem to like them being together! 21:21:33 well, I'll commit to looking at the first one - seems like we want to make sure py3 doesn't show the resource leak error and py2 goes back to what it was doing before the __del__ was added 21:23:55 the original trouble is that if you're using the module-level functions, we "helpfully" create a connection (and therefor requests.Session) for you, but then don't clean it up. the __del__ was a more-or-less straightforward way to appease a resource warning under py3, but (apparently?) it caused an fd leak on py2 21:24:57 i'm fairly certain that if we've got an explicit close plumbed in there (as in the second patch), the __del__ can be removed without reintroducing the warning on py3 21:27:03 all right, well -- that's all i've got 21:27:11 #topic open discussion 21:27:19 anything else we should discuss today? 21:28:02 I was gunna give an update on shrinking with https://review.opendev.org/#/c/741721/ - it's going well, but I've discovered some more stuff "in the wild" that I want to create test cases for 21:28:03 patch 741721 - swift - add swift-manage-shard-ranges shrink command - 5 patch sets 21:29:18 what were some of the new edge cases? 21:30:16 so i'm using matt's graphviz helper - or some version of it that takes input from some dumps of shard ranges I'm pulling out of the db nodes 21:31:33 I think what's happening is when there's "parallel tracks" (like a fork that comes together) the "chain building" code that makes the accepter list breaks early 21:32:26 but the mechanics of marking a sr as shrinking in the root - having the replicas pick up all the overlapping acceptors - and then draining rows out ... WORKS GREAT! 21:32:53 so it's all about picking the right path and marking bad guys as shrinking in the root now - basically all done by the typing 21:33:01 and virtual beers 21:34:01 \o/ 21:34:32 have you started developing heuristics for which side ought to win out? 21:34:51 no, just eyeballin' it 😬 21:35:30 well, the more it happens, the better we'll get at dealing with it :-) 21:35:38 when there's orphan track that start in the middle of nowhere or end early it's really easy to see "which side ought to win" 21:36:31 yeah, right now I have a LOT of shard ranges to shrink - it'll go quick when the code just "always does the right thing" - but pretty soon I'll be wanting to have it pick the ones that need to shrink for me 21:36:59 anything else we ought to bring up? 21:37:10 ... and don't really have a plan to express that in code ... I was looking at p 722570 but I don't grok yet 21:37:10 https://review.opendev.org/#/c/722570/ - swift - WIP: PLaying around with a basic shard-cleanup tool - 3 patch sets 21:37:41 i'll try to get some stable releases out once i've got 2.26.0 done 21:39:31 all right, let's call it early then 21:39:42 thank you all for coming, and thank you for working on swift! 21:40:23 please do take a look at the release notes; i feel like we've done an impressive amount this cycle 21:40:55 (or maybe i just need to make sure i get two releases out again next cycle ;-) 21:41:02 #endmeeting