| opendevreview | Matthew Oliver proposed openstack/swift master: Sharding: SR without an epoch can't merge over one with an epoch https://review.opendev.org/c/openstack/swift/+/807824 | 06:39 |
|---|---|---|
| mattoliver | ^ haven't decided if that's the best way forward for that. But we've seen an interesting "reset" root own_shard_range happen in prod only a few times, but enough to think about the cause. | 06:41 |
| *** elodilles_pto is now known as elodilles | 08:44 | |
| *** timburke__ is now known as timburke | 20:04 | |
| timburke | it's been quiet lately -- my usual 2-week dashboard is rather short ;-) https://review.opendev.org/dashboard/?title=The%20Fortnight%20That%20Was&foreach=-age:2week&Landed%20(Server)=is:merged%20project:openstack/swift&Landed%20(Client)=is:merged%20project:openstack/python-swiftclient&Active%20(Server)=is:open%20project:openstack/swift&Active%20(Client)=is:open%20project:openstack/python-swiftclient&Abandoned%20(Server)=is:abandoned%20pr | 20:04 |
| timburke | oject:openstack/swift&Abandoned%20(Client)=is:abandoned%20project:openstack/python-swiftclient | 20:04 |
| timburke | :-( stupid line breaks | 20:05 |
| timburke | the interesting bit is https://review.opendev.org/dashboard/?title=The%20Fortnight%20That%20Was&foreach=-age:2week&Landed%20(Server)=is:merged%20project:openstack/swift&Landed%20(Client)=is:merged%20project:openstack/python-swiftclient&Active%20(Server)=is:open%20project:openstack/swift&Active%20(Client)=is:open%20project:openstack/python-swiftclient right now, anyways | 20:05 |
| *** elodilles is now known as elodilles_pto | 20:12 | |
| timburke | almost meeting time! | 20:51 |
| timburke | our agenda's prety light, as is the recent patch activity, so it might be a fairly short one | 20:52 |
| mattoliver | Morning, oh I like a good dashboard :) | 20:57 |
| kota | morning | 20:59 |
| mattoliver | kota: o/ | 21:02 |
| kota | mattoliver: o/ | 21:02 |
| * kota is now checking https://wiki.openstack.org/wiki/Meetings/Swift for meeting... | 21:02 | |
| mattoliver | we do have one. timburke said so a few minutes ago. he's obviously distracted :) | 21:03 |
| kota | all right, thx | 21:03 |
| timburke | #startmeeting swift | 21:04 |
| opendevmeet | Meeting started Wed Sep 8 21:04:01 2021 UTC and is due to finish in 60 minutes. The chair is timburke. Information about MeetBot at http://wiki.debian.org/MeetBot. | 21:04 |
| opendevmeet | Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. | 21:04 |
| opendevmeet | The meeting name has been set to 'swift' | 21:04 |
| timburke | sorry, i *was* distracted ;-) | 21:04 |
| timburke | who's here for the swift meeting? | 21:04 |
| kota | hello | 21:04 |
| mattoliver | o/ | 21:04 |
| timburke | i think acoles is out, and we'll just see as we go whether clayg and zaitcev (and anyone else) are around | 21:05 |
| timburke | as usual, the agenda's at https://wiki.openstack.org/wiki/Meetings/Swift | 21:05 |
| timburke | #topic PTG | 21:06 |
| -opendevstatus- NOTICE: The Gerrit service on review.opendev.org is going offline momentarily for a host migration and zuul upgrade, downtime should be only a few minutes. | 21:06 | |
| timburke | just a reminder to add any topics you want to discuss to the etherpad | 21:06 |
| timburke | #link https://etherpad.opendev.org/p/swift-ptg-yoga | 21:06 |
| mattoliver | ahh, that'll explain why I couldn't get to gerrit just now | 21:06 |
| kota | lol | 21:07 |
| mattoliver | I think I added a few. But no doubt there are more and we still need to add the normal ones like opps feedback :) | 21:07 |
| timburke | we're a little more than a month away; it'll be good to see everyone again (even if just via video chat) | 21:07 |
| timburke | #topic ring v2 | 21:08 |
| zaitcev | I added that little e-mail from Arkady | 21:08 |
| zaitcev | to the Etherpad. | 21:09 |
| timburke | thanks zaitcev! | 21:09 |
| timburke | i don't know that we've proposed any new patches for ring v2 yet, but i've got a chain that i've started on; hopefully i'll have it to a state that i'm happy to push it up in another week or two | 21:10 |
| timburke | while i was doing that, i wrote up some history of ring formats (as best as i could make out): https://gist.github.com/tipabu/435fc63f1aa76d6b346956571a393665 | 21:10 |
| mattoliver | oh nice a gist, I'll take a look. we did have https://review.opendev.org/c/openstack/swift/+/803665 which implemented a part of it as POC to see how it looked. But looking forward to seeing what you've done esp post etherpad. | 21:11 |
| mattoliver | I think that was right can't get into gerrit to double check :P | 21:12 |
| timburke | huh. worked for me -- "ring: store actual ring replica count to better deserialize" | 21:12 |
| mattoliver | oh its back! | 21:13 |
| mattoliver | looking forward to working out this v2 serialization format, because it'll open the doors to last_primary improvements we can add to things | 21:14 |
| timburke | yeah, that's my thinking, too: we can make a few different things a good bit smarter if they've got some ring history | 21:15 |
| timburke | #topic sharder and reconciler | 21:15 |
| timburke | #link https://review.opendev.org/c/openstack/swift/+/800748 | 21:15 |
| mattoliver | Not as much progress here as I'd like in the last week or so | 21:16 |
| timburke | no worries -- i just wanted to check in on it and see what we can do to help get it moving again | 21:16 |
| mattoliver | reconciler does things a little different. and we were spitballing ways to make the root -> shard SPI migrations act similarly | 21:16 |
| mattoliver | I left a suggestion in the patch discussion.. and now waiting on others to comment. | 21:17 |
| mattoliver | but Al's on vacation.. so kinda got stalled. | 21:17 |
| mattoliver | hopefully have an update next week :) | 21:17 |
| timburke | sounds good. and i'll try to load up the problem in my head and get a useful review out :-) | 21:18 |
| mattoliver | thanks :) | 21:18 |
| timburke | one last-minute topic | 21:19 |
| timburke | #topic reviewed but not merged patches | 21:20 |
| timburke | i've got a few patches w/ a +2 but no +A: https://review.opendev.org/q/project:openstack/swift+label:Code-Review%252B2+is:open+-age:52w | 21:20 |
| timburke | anybody want to call for more eyes on some of those, or should we just merge them? | 21:21 |
| timburke | the query params patch still has me nervous, but i think the others are probably good to go? but then, i'm biased | 21:21 |
| zaitcev | Why nervous? Security? | 21:22 |
| mattoliver | The fact that they have a +2 means they're in good shape. Only worry is, was the +A not added because the reviewer wasn't 100% sure? I'm leaning towards just landing them. But I'll give them a parse over today and land or comment. | 21:23 |
| zaitcev | Yes. I wrote "The scenarios Tim is considering are concerning, but fortunately the allowed_headers list is empty by default." Now I cannot remember a blessed thing about that patch. What did I mean exactly .... | 21:24 |
| mattoliver | lol | 21:24 |
| zaitcev | Always write it down! | 21:24 |
| zaitcev | The EC watcher is no big deal, I can just land it now. | 21:25 |
| zaitcev | Oh, righ | 21:27 |
| zaitcev | I see why. It has no tests. | 21:27 |
| timburke | switch it to a -1 then ;-) i can work on adding some | 21:28 |
| zaitcev | I meant to signal "fine by me, but you guys are always so strict about tests, so I'll leave it with just +2 for now". | 21:28 |
| timburke | i do have some security concerns on the tempurl patch, yeah. the default is secure, but the motivation for the patch makes me think that it's likely people will allow multipart-manifest, which seems to have a security impact | 21:29 |
| timburke | all right, that's all i've got | 21:30 |
| timburke | #topic open discussion | 21:30 |
| timburke | anything else we should talk about this week? | 21:30 |
| mattoliver | maybe hold off on the tempurl and let's discuss it at PTG maybe? | 21:31 |
| mattoliver | we can brainstorm security impact maybe | 21:31 |
| zaitcev | Going back to right history gist, we have a "History" section in https://docs.openstack.org/swift/latest/overview_ring.html | 21:31 |
| zaitcev | Is this something worth rolling into one document or whatevs | 21:31 |
| zaitcev | s/right /ring / | 21:32 |
| timburke | yeah, probably a good idea. i can work on getting a docs patch up | 21:32 |
| mattoliver | made some interesting progress on the opentracing integration into swift. https://review.opendev.org/c/openstack/swift/+/803879 our SREs have been playing with it. And it's really nice visualising swift requests across a cluster. Start seeing paterns and hopefully ares to fix. | 21:34 |
| mattoliver | I'll show some traces at PTG maybe :) | 21:34 |
| zaitcev | Yeah, it's our world... Whenever our something small and useful pops, it turns into this whole thing. EC watcher needs tests, gists can be more useful in docs, etc. | 21:34 |
| timburke | always more work to be done :-) | 21:35 |
| timburke | all right, seems like we can let kota and mattoliver get on with their mornings | 21:37 |
| timburke | thank you all for coming, and thank you for working on swift! | 21:37 |
| timburke | #endmeeting | 21:37 |
| opendevmeet | Meeting ended Wed Sep 8 21:37:51 2021 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4) | 21:37 |
| opendevmeet | Minutes: https://meetings.opendev.org/meetings/swift/2021/swift.2021-09-08-21.04.html | 21:37 |
| opendevmeet | Minutes (text): https://meetings.opendev.org/meetings/swift/2021/swift.2021-09-08-21.04.txt | 21:37 |
| opendevmeet | Log: https://meetings.opendev.org/meetings/swift/2021/swift.2021-09-08-21.04.log.html | 21:37 |
| mattoliver | thanks timburke | 21:38 |
| mattoliver | oh.. meeting is over but just thought about something. but nothing meeting worthy, so just leave it here for people to think about. | 21:40 |
| mattoliver | we've had a sharded root container's epoch reset again in prod. It's only happened 3 times, and seems to be the same container. Only edgecase I can come up with is a put to a handoff that gets a defaulted own_shard_range that then merges over the other primaries.. but haven't been able to prove it. I wonder if something like: https://review.opendev.org/c/openstack/swift/+/807824 would help. ie, once a container gets an epoch (shards) | 21:41 |
| mattoliver | it never looses it, so make an epoch check in merge_shard_ranges | 21:41 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!