Wednesday, 2026-02-11

mattoliverHaven't had a meeting in a while so might have a quick one that basically will mostly just cover what Tim covered last week... sorry for not updating the agenda until close to the meeting07:02
mattoliver#startmeeting swift07:03
opendevmeetMeeting started Wed Feb 11 07:03:10 2026 UTC and is due to finish in 60 minutes.  The chair is mattoliver. Information about MeetBot at http://wiki.debian.org/MeetBot.07:03
opendevmeetUseful Commands: #action #agreed #help #info #idea #link #topic #startvote.07:03
opendevmeetThe meeting name has been set to 'swift'07:03
mattoliverWhose here for the swift meeting.. assuming just me :P 07:03
cschwedeo/07:03
mattoliveras always the agenda is at:07:03
mattoliver#link https://wiki.openstack.org/wiki/Meetings/Swift07:03
mattoliveroh hey cschwede !07:04
mattoliverNot too much to report today, so will probably be a quick one07:04
cschwedeI have some eventlet news07:04
mattoliver#topic eventlet removal07:04
mattoliverGreat, go for it!07:04
cschwedeDid some initial benchmarking with threading, and it looks like it might be ~12% slower than evenlet. So...07:05
mattoliveroh well that's interesting07:05
cschwedeI tried another approach using asyncio, and - without publishing numbers yet - this looks way more performant.07:05
mattoliverOK, well that's good to know. 07:06
cschwedeWhat I'm doing now is collecting all the data, pros and cons for both approaches including a list of required work and possible blockers. I would like to discuss this in a mid-cycle meeting with some of you07:06
cschwedeWell, with all of you, but not sure if all of you will be able to join ;)07:06
mattoliverThat sounds like a great idea! and I appreciate all the work your putting into it!07:07
cschwedeSo maybe we can schedule a call next week or the week after?07:07
mattoliveryeah, sounds good. I'll reach out to the others.07:07
mattoliverAnd ping you with the others availability.  07:08
cschwedeI'm fine with meeting late at my local time, trying to talk to as many people as possible07:08
mattoliverCan you do a late night to make it easier for the US for should I try and get them to stay up late07:08
mattoliveroh well there you go, already answered :P 07:09
cschwede:)07:09
mattolivercool, well leave it with me and I'll poke people :) I'm looking forward to it!07:09
cschwedeYes, me too! Thx for organizing this!07:09
mattoliverno worries!07:09
mattoliverLet's move on then07:10
mattoliver#topic Timestamp collisions07:10
mattoliverThis is still going on07:10
mattoliverThe "Band aid" fix first version still isn't out. Which it annoying. But turns out timestamps are everywhere and getting confidence is hard07:10
mattoliverBut we still have to plan to only start with obj PUTs and roll it out to other verbs and types later07:11
mattoliverAl, Tim, Jian and I has a meeting about it. We've merged more then 40 patches to get clean up testing so we could get better confidence in it07:12
cschwedeWow, that's a significant number of patches!07:12
mattoliverAnd just deciding about the things that need a more precise timestamp/jitter and things that don't. 07:12
mattoliverAnd we seem to have come to the conclusion that things that need to have a reference in the system, like a request then we should be using the timestamp with offset and jitter (like a uuid or a refence to an event) 07:14
mattoliverand other things, like sharding epochs and x-delete-at's are more like a simple wrapping of time.time but swift-ified 07:14
mattoliverSo there is a new precursor patch to add Timestamp and SimpleTimestamp... although there is still arguements on the name.. SimpleTimestamp07:15
mattoliver#link https://review.opendev.org/c/openstack/swift/+/96874007:15
patch-botpatch 968740 - swift - timestamps: add SimpleTimestamp for ShardRanges - 22 patch sets07:15
mattoliverLooks like Al has a new patch up to help with timestamp normailisation. acoles sorry only just noticed. Will look at it in the morning (got really distracted with down stream things today).07:16
mattoliverAnyway, getting closer, we hope to have something to carry in a downstream release soon if not land something. 07:17
mattoliverLet's move on to the next topic then..07:18
cschwedeDownstream release for you means running in (pre-)production?07:18
mattoliverlol, you don't want to know :P But yeah both pre-prod first. 07:18
mattoliverBut to see if it stops our problems and we can trust it, prod too07:19
cschwedeI did not read all review comments, is this safe to run with some servers using this already and others not yet?07:20
mattoliver*once we can trust it in pre-prod07:20
mattoliverwell I hope so, at least for upgrade impact on our clusters. 07:20
cschwedeNice!07:21
mattolivergood thing is, we're reusing the existing offset field, so old swift will just think it's a timestamp with a bunch of offset07:21
mattoliverand like we said it's for extra presision on time, so only looked at mostly on collisions. 07:22
mattoliverWill make filenames on disks longer.. which is why I keep playing with nanosecond resolution as a side quest.. but don't thinks its anywhere near as close and the offset jitter.07:23
mattoliver*approach07:23
mattoliverMight be good the chat about at a face2face :) 07:23
mattoliver#topic vPTG07:24
mattoliverThis is more just informational. We have another vPTG coming up07:24
mattoliver#link  https://ptg.openinfra.org/07:24
mattoliver^ register there07:24
mattoliverApr 20-2407:25
mattoliverNo doubt we'll need cschwede to remind us to set things up again :P 07:25
cschwederofl07:25
mattoliverSo register, come, and let's make swift better! 07:26
mattoliverFinal topic I have is...07:26
mattoliver#topic LRC (locally repairable/recoverable code) scheme for liberasurecode07:26
mattoliver#link https://review.opendev.org/c/openstack/liberasurecode/+/95928007:27
patch-botpatch 959280 - liberasurecode - feature: LRC (locally repairable code backend) - 21 patch sets07:27
mattoliver#link https://review.opendev.org/c/openstack/pyeclib/+/96428307:27
patch-botpatch 964283 - pyeclib - prepare LRC backend integration - 6 patch sets07:27
mattoliverI feel these are getting close. I did a review, tho I should probably test them out. My linear algebra isn't the strongest, but timburke  seems to be all on it.07:28
mattoliverI know tim wants to do some more releases of liberasurecode,  so if anyone feel like getting some mathematics on or want to test it out, or want to review some c, go to it!07:29
mattoliverThats all I got for the moment.. so07:30
mattoliver#topic open discussion 07:30
mattoliverAnything else anyone wants to bring up?07:30
cschwedeNothing from my side, thx for the updates mattoliver!07:30
mattoliverI've been reworking some of my older patches, rebacing and testing etc. I pushed new version of tracing, SO_TIMESTAMPING and db vacuum. All of which I think are pretty interesting and I want to get into swift..  07:31
mattoliverBut sounds like my dinner is ready.. so that can wait for a later meeting or better yet the PTG where I can figure out if there are any blockers in anyones mind 07:32
mattoliverWith that I'll call it early then! 07:33
mattoliverThanks for coming cschwede ! 07:33
mattoliverThanks everyone for working on swift!07:33
mattoliver#endmeeting07:33
opendevmeetMeeting ended Wed Feb 11 07:33:30 2026 UTC.  Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)07:33
opendevmeetMinutes:        https://meetings.opendev.org/meetings/swift/2026/swift.2026-02-11-07.03.html07:33
opendevmeetMinutes (text): https://meetings.opendev.org/meetings/swift/2026/swift.2026-02-11-07.03.txt07:33
opendevmeetLog:            https://meetings.opendev.org/meetings/swift/2026/swift.2026-02-11-07.03.log.html07:33
cschwedeThanks mattoliver for running the mtg, enjoy dinner!07:33
opendevreviewAlistair Coles proposed openstack/swift master: trivial: remove print statement in relinker unit test  https://review.opendev.org/c/openstack/swift/+/97637512:15
opendevreviewAlistair Coles proposed openstack/swift master: WIP proxy_logging: use consistent labels for all stasd metrics  https://review.opendev.org/c/openstack/swift/+/97594916:17
opendevreviewMerged openstack/swift master: trivial: remove print statement in relinker unit test  https://review.opendev.org/c/openstack/swift/+/97637517:44
opendevreviewWael Halbawi proposed openstack/swift master: relinker: Prefix log messages with invoked action  https://review.opendev.org/c/openstack/swift/+/97555217:52
opendevreviewAlistair Coles proposed openstack/swift master: timestamps: add SimpleTimestamp for ShardRanges  https://review.opendev.org/c/openstack/swift/+/96874019:07
opendevreviewAlistair Coles proposed openstack/swift master: WIP Timestamp: add random jitter to offset field  https://review.opendev.org/c/openstack/swift/+/96773819:07
opendevreviewAlistair Coles proposed openstack/swift master: tests: Stop using normalize_timestamp  https://review.opendev.org/c/openstack/swift/+/97628819:07
opendevreviewAlistair Coles proposed openstack/swift master: swob: fix error message from ensure_x_timestamp  https://review.opendev.org/c/openstack/swift/+/97653219:07
opendevreviewAlistair Coles proposed openstack/swift master: object-server: use internal form for default GET x-timestamp  https://review.opendev.org/c/openstack/swift/+/97653319:07
opendevreviewClay Gerrard proposed openstack/swift master: DNM: Add debug logging to ssync sender and receiver  https://review.opendev.org/c/openstack/swift/+/97653920:24

Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!