| mattoliver | We are working on adding some extra smarts in the ring to know the last primary a partition sat so we can be smarter in the reconstructor, ie. don't bother rebuilding if the missing peice is just over there on an old primary. Should give us a build in handoffs_only after a rebalance. In local testing looks like a pretty good improvement. | 00:07 |
|---|---|---|
| mattoliver | And yes, that's what handoffs_only does, so if there are any valid missing frags that do need to be rebuilt.. it wont. so not something to have on all the time.. but great post rebalance. | 00:10 |
| reid_g | Nice. Makes sense | 00:12 |
| reid_g | Anybody here maintain the puppet-swift repo? I don't see that as parameter... but I guess that reconstruction isn't even setup there | 00:14 |
| reid_g | Maybe it won't ever support it... the comment in the readme is from 3/5 years ago | 00:15 |
| opendevreview | Matthew Oliver proposed openstack/swift master: container-updater: no incoming syncs no account update https://review.opendev.org/c/openstack/swift/+/811833 | 01:26 |
| acoles | timburke: apologies, I cannot make today's meeting | 20:09 |
| timburke | almost meeting time! | 20:54 |
| kota | morning | 20:59 |
| seongsoocho | kota: good morning | 21:00 |
| kota | seongsoocho: o/ | 21:01 |
| timburke | #startmeeting swift | 21:03 |
| opendevmeet | Meeting started Wed Oct 6 21:03:00 2021 UTC and is due to finish in 60 minutes. The chair is timburke. Information about MeetBot at http://wiki.debian.org/MeetBot. | 21:03 |
| opendevmeet | Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. | 21:03 |
| opendevmeet | The meeting name has been set to 'swift' | 21:03 |
| timburke | who's here for the swift meeting? | 21:03 |
| seongsoocho | o/ | 21:03 |
| mattoliver | o/ | 21:03 |
| kota | o/ | 21:05 |
| timburke | as usual, the agenda's at https://wiki.openstack.org/wiki/Meetings/Swift | 21:05 |
| timburke | first up | 21:05 |
| timburke | #topic gerrit outages | 21:05 |
| timburke | just an fyi, in case you missed the service-announce email | 21:05 |
| timburke | #link http://lists.opendev.org/pipermail/service-announce/2021-October/000024.html | 21:05 |
| timburke | gerrit will be down a couple times between now and the ptg, but both should be fairly short | 21:06 |
| timburke | #topic PTG | 21:06 |
| timburke | only a week and a half away, now! | 21:06 |
| timburke | still i added a few more topics to the etherpad | 21:07 |
| timburke | #link https://etherpad.opendev.org/p/swift-ptg-yoga | 21:07 |
| timburke | feel free to add more! | 21:07 |
| mattoliver | cool, look forward to chatting with everyone again (in video) | 21:08 |
| timburke | on that note... | 21:09 |
| timburke | #topic summit | 21:09 |
| timburke | we're going to have an in-person summit again! | 21:09 |
| timburke | #link https://openinfra.dev/summit/ | 21:09 |
| mattoliver | \o/ | 21:09 |
| kota | wow | 21:09 |
| seongsoocho | 🎉 yay~~~ | 21:09 |
| timburke | Jun 7-9 in Berlin | 21:09 |
| timburke | i should look at submitting a talk about pushing swift towards exabyte scale :-) | 21:10 |
| mattoliver | So long as international Borders are open from Oz and my employer will send me, I'll be there :) | 21:11 |
| seongsoocho | I'm looking for a flight ticket :) | 21:11 |
| timburke | i look forward to actually seeing everyone in person again :-D | 21:11 |
| mattoliver | oh maybe I could write one about request tracing though the cluster... might be fun to see different usage patterns. | 21:12 |
| timburke | #topic dnspython 2.0.0+ support | 21:13 |
| timburke | just another fyi -- looks like eventlet finally declared support for dnspython 2! | 21:13 |
| timburke | #link https://github.com/eventlet/eventlet/issues/619 | 21:13 |
| mattoliver | oh nice | 21:14 |
| timburke | when prometheanfire went to update global-requirements, it broke our tests, though :-( | 21:14 |
| timburke | the fix was pretty small, as best i could tell, so i just landed it | 21:14 |
| timburke | #link https://review.opendev.org/c/openstack/swift/+/812424 | 21:15 |
| mattoliver | Thanks for fixing it | 21:15 |
| timburke | but i haven't actually *functionally* tested it. so if anyone uses cname_lookup and could do some testing, i'd appreciate feedback on how it goes | 21:16 |
| timburke | #topic reconstructor and the replication network | 21:18 |
| mattoliver | this was a good find. We definitely don't want to be pulling frags over the cluster network. | 21:20 |
| timburke | reid_g recently noticed that reconstructor GETs don't use the replication network. i've got a patch to change that (in the spirit of https://review.opendev.org/c/openstack/swift/+/735751) | 21:20 |
| timburke | #link https://review.opendev.org/c/openstack/swift/+/812614 | 21:20 |
| timburke | thanks for the reviews; i'll write up a bug and some tests shortly | 21:21 |
| timburke | that's all i've got | 21:21 |
| timburke | #topic open discussion | 21:21 |
| timburke | what else should we bring up this week? | 21:21 |
| mattoliver | We ran into a fun bug where account stats would get stuck. I wrote a probe test to prove it. And here is also the first attempt at fixing it: https://review.opendev.org/c/openstack/swift/+/811833 | 21:23 |
| mattoliver | So any thoughts on that approach would be good. | 21:23 |
| mattoliver | The code was simple until acoles pointed out we'd need to support single replica container rings.. and the patch in essence only allows updates from non handoffs that have had an incoming sync. So opts for an escape hatch (config option) to override if your running with 1. (don't know why you would ever though) | 21:24 |
| timburke | it's a good corner case, and i like the config-option escape hatch. maybe better named something like "update_when_never_replicated" though? | 21:25 |
| mattoliver | ahh yeah nice, I suck at coming up with config option names :P | 21:26 |
| mattoliver | having spent time in the incoming_syncs code it does make me worry about some potentual extra work on some edgecases. But I might cofirm my understanding before I write up anything like a bug because I could be wrong :) | 21:27 |
| timburke | fwiw, i know i've played around with single-disk (so, single-replica) stuff before: https://github.com/tipabu/sds | 21:27 |
| timburke | it's worth making sure single-replica works for dev/test environments | 21:28 |
| mattoliver | cool, a usecase means it's worth the effort, thanks! | 21:28 |
| timburke | all right, let's make it a short meeting so you guys can get on with your mornings :-) | 21:30 |
| timburke | thank you all for coming, and thank you for working on swift! | 21:30 |
| timburke | #endmeeting | 21:30 |
| opendevmeet | Meeting ended Wed Oct 6 21:30:54 2021 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4) | 21:30 |
| opendevmeet | Minutes: https://meetings.opendev.org/meetings/swift/2021/swift.2021-10-06-21.03.html | 21:30 |
| opendevmeet | Minutes (text): https://meetings.opendev.org/meetings/swift/2021/swift.2021-10-06-21.03.txt | 21:30 |
| opendevmeet | Log: https://meetings.opendev.org/meetings/swift/2021/swift.2021-10-06-21.03.log.html | 21:30 |
| opendevreview | Tim Burke proposed openstack/swift master: ec: Use replication network to get frags for reconstruction https://review.opendev.org/c/openstack/swift/+/812614 | 23:38 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!