21:00:09 #startmeeting swift 21:00:09 Meeting started Wed Sep 15 21:00:09 2021 UTC and is due to finish in 60 minutes. The chair is timburke. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:00:09 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:00:09 The meeting name has been set to 'swift' 21:00:19 who's here for the swift meeting? 21:00:36 o/ hi! 21:01:02 o/ 21:02:49 o/ 21:02:50 as usual, the agenda's at https://wiki.openstack.org/wiki/Meetings/Swift 21:03:11 i haven't updated it since last week (sorry), but i mostly wanted to check in on the same things 21:03:25 #topic PTG 21:03:46 just a reminder about the call for discussion topics at https://etherpad.opendev.org/p/swift-ptg-yoga 21:04:08 little more than a month to the PTG 21:04:57 #topic v2 ring format 21:05:42 minor update here -- i've got a new chain up, starting at https://review.opendev.org/c/openstack/swift/+/808530 21:06:29 Thanks for working on it timburke 21:06:35 clayg and mattoliver have started taking a look and offered some good feedback 21:07:11 we'll continue to see how it evolves 21:07:54 #topic misplaced objects and sharded containers 21:08:49 now that acoles is back, i wanted to check in on https://review.opendev.org/c/openstack/swift/+/800748, see how much we're still concerned about the issue and what we need to keep making progress 21:09:37 I still need to loop back on it. Will take a new look today. As I said last week I made a suggestion in the review. 21:10:22 👍 21:10:29 But good news is the solution involved 2 things, 1. Allowing shards to return objects from any given policy 21:10:44 I need to go back to it too, sorry i have not revisited it since my vacation 21:10:52 Which has landed which bandaids the main problem. 21:11:25 It's the second that listed.. actually making the change so objects can get reconciled. 21:11:42 acoles nps you've been busy :) 21:12:22 I'm going on vacation next week, but nobody will ever notice. 21:12:46 zaitcev: we will! have a good one 21:12:50 Well bandaids one of the issues we saw. 21:13:06 zaitcev, enjoy! 21:13:14 all right, that's all i've got 21:13:17 #open discussion 21:13:22 #topic open discussion 21:13:32 anything else we ought to bring up? 21:13:55 I'm thinking but I don't have anything super urgent. 21:16:01 We've been having fun with go aws clients no always handling 100-continues properly or rather timing out waiting for 100-continue I think. But never fear timburke has an upstream eventlet patch to handle it. 21:16:34 Don't have it handy as I'm on my phone 21:16:49 oh yeah :-) https://github.com/eventlet/eventlet/pull/578 21:18:23 which also reminds me that eventlet still needs py310 support: https://github.com/eventlet/eventlet/pull/715 (just a start; still some issues remain, not least of which is CI) 21:19:14 py310 is expected next month: https://www.python.org/dev/peps/pep-0619/ 21:19:50 That Expect: 100 is 2 years old 21:20:39 yup :-( 21:21:59 fwiw, we were running with the original patch (discard the unread request) when we ran into further issues. seems we can't win; some clients will go either way 21:22:05 only thing to do is close out the connection 21:23:18 We've seen a rare case where a root containers own_shard_range's epoch is reset. Been playing with probetests to try and recreate the issue, attempting weird hand off or rebalance situations. Still can't reproduce unless I force it to happen. But do have this which might solve it in any case https://review.opendev.org/c/openstack/swift/+/807824 21:24:00 There's are probe test in there that forces the problem. 21:24:38 Anyway, let me know if it makes enough sense to attempt to fix it that way or not. 21:25:01 thanks mattoliver 21:29:45 timburke: we're done, right? 21:30:04 yeah, seems like we're winding down 21:30:15 (sorry, had to take a call) 21:30:26 thank you all for coming, and thank you for working on swift! 21:30:30 #endmeeting