21:00:13 <timburke> #startmeeting swift
21:00:13 <opendevmeet> Meeting started Wed Sep 20 21:00:13 2023 UTC and is due to finish in 60 minutes.  The chair is timburke. Information about MeetBot at http://wiki.debian.org/MeetBot.
21:00:13 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
21:00:13 <opendevmeet> The meeting name has been set to 'swift'
21:00:26 <timburke> who's here for the swift team meeting?
21:00:40 <kota> o/
21:02:58 <timburke> i failed to update the agenda, so i'm gonna kinda wing it ;-)
21:03:04 <timburke> #topic vPTG
21:03:14 <timburke> it's only about a month away now!
21:03:23 <kota> nice
21:03:24 <timburke> Oct 23-27
21:03:35 <timburke> i just created a poll for meeting times
21:03:44 <timburke> #link https://framadate.org/liKzqQw5lZQW8Pvf
21:04:00 <timburke> (thanks for the framadate suggestion, clarkb!)
21:04:40 <timburke> the etherpad for topics is still rather empty, though
21:04:43 <timburke> #link https://etherpad.opendev.org/p/swift-ptg-caracal
21:05:14 <timburke> i'll aim to add a few topics i'm interested in after this meeting
21:06:15 <timburke> if you haven't already, don't forget to register
21:06:18 <timburke> #link https://ptg2023.openinfra.dev/
21:06:48 <timburke> i understand it helps the foundation track things
21:08:24 <timburke> next up
21:08:42 <timburke> #topic labeled metrics
21:09:18 <timburke> i finally got around to rebasing my main patch for this
21:09:24 <timburke> #link https://review.opendev.org/c/openstack/swift/+/885321
21:10:13 <timburke> and got a vagrant-swift-all-in-one patch together to add something approaching a "real" metrics pipeline
21:10:37 <timburke> #link https://github.com/NVIDIA/vagrant-swift-all-in-one/pull/150
21:10:57 <kota> i see, it's sort of prometheus exporters.
21:12:09 <timburke> yup! it's similar to what we're running in prod (which explains some of the warts in the statsd mapping file)
21:12:22 <kota> nice work
21:14:35 <timburke> one handy thing about statsd_exporter, though, is that it already supports some statsd extensions add labeling; my hope is that i can come up with a patch that will let you switch to emitting one of the various extensions without any gap in your graphs (if you were already using some statsd_exporter-based metrics pipeline)
21:15:45 <timburke> at some point i want to include an upstream docs patch similar to what we did for adding a replication network to your SAIO (see https://docs.openstack.org/swift/latest/replication_network.html)
21:16:08 <kota> +1
21:16:28 <timburke> but i might wait on that until after i've got a firmer plan on that migration path
21:18:13 <timburke> i'd love for us to express more of an opinion about a recommended metrics pipeline upstream
21:19:59 <timburke> thinking about stats more has led me to write a couple patches already (and surely will lead to more down the line :-)
21:20:18 <timburke> #topic swift-recon-cron
21:20:56 <timburke> one of them was to emit a gauge for how many asyncs are on disk
21:21:12 <timburke> #link https://review.opendev.org/c/openstack/swift/+/895737
21:23:23 <timburke> i see this as having two benefits over the existing recon drop: first, it's a little more timely (since it's not waiting for whatever other process to go read recon); second, it offers these stats per-disk (the distribution across disks may have gotten lumpy if a disk was unmounted for some time)
21:25:33 <timburke> it got me thinking more about swift-recon-cron, though, and wondering why we have it as a separate cron-based task
21:28:17 <timburke> i might try out having swift-object-updater (or maybe swift-object-server? like, the worker-management process) spin out a separate thread/process to handle swift-recon-cron's job, and have the scheduling go in that config
21:29:32 <timburke> that's surely going to warrant some discussion, so it's one of the topics i should add to the etherpad :-)
21:30:09 <timburke> #topic error-limiting stats
21:30:13 <timburke> the other patch was
21:30:16 <timburke> #link https://review.opendev.org/c/openstack/swift/+/895976
21:30:45 <timburke> which would add a counter for when we actually skip a node in NodeIter because of error limiting
21:31:14 <timburke> this was in part me trying to figure out how to assess one of acoles's changes
21:31:31 <timburke> #link https://review.opendev.org/c/openstack/swift/+/895976
21:31:41 <timburke> #undo
21:31:41 <opendevmeet> Removing item from minutes: #link https://review.opendev.org/c/openstack/swift/+/895976
21:31:44 <timburke> #link https://review.opendev.org/c/openstack/swift/+/890527
21:31:51 <timburke> that's the one
21:32:42 <timburke> where he's making sure that we're incrementing errors for a node even if we can't find a replacement
21:33:21 <kota> it looks like what Pete discussed a few weeks ago.
21:35:00 <timburke> possibly -- all the more reason for us to want to get better metrics/logging around it :-)
21:35:12 <kota> :)
21:37:05 <timburke> that's at the head of a chain of to unify _get_next_response_part in the proxy for replicated and EC
21:37:26 <timburke> #link https://review.opendev.org/c/openstack/swift/+/890919
21:38:46 <timburke> next up
21:39:08 <timburke> #topic part-number queries for slo/MPU
21:40:13 <timburke> indianwhocodes has rebased his patches on top of some of clayg's refactors -- we're getting a decently-long chain now
21:40:28 <timburke> #link https://review.opendev.org/c/openstack/swift/+/894800
21:41:07 <timburke> has s3api ignore some MPU-specific sysmeta if the object isn't an slo
21:41:20 <timburke> #link https://review.opendev.org/c/openstack/swift/+/893578
21:42:33 <timburke> refactors slo a good bit, but (shouldn't) result in any behavioral change
21:42:49 <timburke> #link https://review.opendev.org/c/openstack/swift/+/894570
21:43:33 <timburke> adds the ability to request the range corresponding to part number N from an slo
21:43:43 <timburke> and finally
21:43:49 <timburke> #link https://review.opendev.org/c/openstack/swift/+/894580
21:44:40 <timburke> provides the (now actually documented!) partNumber support for s3api
21:44:49 <timburke> see also, https://bugs.launchpad.net/swift/+bug/1735284
21:45:45 <timburke> all right, i think that's most of what all's been going on this past week
21:45:50 <timburke> #topic open discussion
21:45:59 <timburke> anything else we ought to talk about today?
21:46:31 <kota> not from my side, thanks for proceeding
21:46:58 <timburke> all right, i'll let you get on with your morning then
21:47:08 <timburke> thank you for coming, and thank you for working on swift!
21:47:13 <timburke> #endmeeting