21:00:03 #startmeeting swift 21:00:03 Meeting started Wed Sep 14 21:00:03 2022 UTC and is due to finish in 60 minutes. The chair is timburke. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:00:03 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:00:03 The meeting name has been set to 'swift' 21:00:12 who's here for the swift meeting? 21:01:10 hi 21:02:35 sorry for missing the last meeting -- i fell down a bit of an existential hole trying to write my ptl candidacy, and still didn't get around to doing it 21:02:56 but, i've finally caught up a bit on the mailing list! should've done it sooner 21:03:01 first up 21:03:09 #topic openstack antelope 21:03:17 no worries, thanks for the ptl nomination 21:04:12 with the zed cycle coming to a close, the official name for the next cycle is out! looping back to A, looks like the first release of 2023 will be OpenStack Antelope 21:04:41 we're back to A! 21:07:00 i vaguely wanted to propose we start taking letters from On Beyond Zebra, but never actually brought it up :-) 21:07:54 think something like https://i.pinimg.com/originals/90/cc/de/90ccde246b7d04704388248e11a754a5.jpg 21:08:50 lol 21:08:56 anyway, not much to discuss about it, but figured i ought to announce it in case anyone else missed it 21:09:04 #topic ptg 21:09:45 i also discovered that the ptg in Ohio has been changed to be all-virtual again! 21:09:57 #link https://lists.openstack.org/pipermail/openstack-discuss/2022-August/029879.html 21:10:48 unfortunately, i'd stopped paying much attention to emails with "ptg" in the title, as i knew we couldn't hope for much presence 21:11:09 but fortunately, we *are* presumed to be attending! 21:11:17 #link https://lists.openstack.org/pipermail/openstack-discuss/2022-August/030225.html 21:11:55 oh, all-virtual 21:12:58 not sure whether someone submitted an update to the team signup for me, or diablo_rojo or somebody just kind of knew that swift would want some time if it was virtual :-) 21:13:44 i still need to reserve some time slots -- i'll try to get a doodle up 21:13:48 :-) 21:13:52 and set up an etherpad 21:14:31 i'd also encourage everyone to register 21:14:34 #link https://openinfra.dev/ptg/ 21:16:07 * kota registered just now 21:16:07 the dates are October 17-21 -- i think since it went virtual, we may have picked up an extra day 21:16:39 but it's not far off now! a little over a month away 21:17:15 those were the two main things i wanted to call out from the broader ecosystem 21:17:23 #topic ring v2 21:18:24 the only other thing i wanted to bring up was that i've continued to poke at the new ring format and mattoliverau's extension to it to add a rebalance history 21:18:39 a couple new patches came out of it 21:18:48 #link https://review.opendev.org/c/openstack/swift/+/857234 21:20:03 adds a new swift-ring-info utility, which takes a ring.gz and tells you some basic info about it (what ring-builder version it corresponds to, which ring format version it uses, how many replicas, partitions, etc.) 21:20:13 #link https://review.opendev.org/c/openstack/swift/+/857758 21:20:51 is a first pass at having the proxy use the rebalance history when trying to GET/HEAD 21:22:57 we've seen some cases where that latter patch could come in handy: a lot of clients (including python-swiftclient) like to make HEADs before PUTting a bunch of data -- but if you've got issues with some nodes, you may not get back enough 404s to be confident in returning that to the client 21:24:52 i see 21:25:07 in particular, we've occasionally had a node go offline and another primary be slow to respond -- and with the default rebalance_missing_suppression_count of 1, a set of responses like (404, timeout, no route to host) from primaries and nothing but 404s from handoffs, 21:25:54 the proxy can't be sure whether the 404 was because the primary was recently given the assignment and waiting to be filled, or there's genuinely no data to be found in the cluster 21:27:09 the idea of the patch is to add a for_read kwarg when getting more nodes -- when true, we add in nodes from the rebalance history at the start of the handoff list 21:28:04 sounds nice 21:28:19 and when we do, *we include an index* so it looks kind of like a primary -- that way, we can treat its 404 as more or less "authoritative" 21:29:27 one of the last primary or the new primary could reasonably be expected to 404 during a rebalance, but presumably not *both* unless no one else has data, either 21:29:59 it needs tests, but i've managed to use it to good effect in my saio 21:30:21 anyway, that's all i've got for this week 21:30:30 #topic open discussion 21:30:43 kota, anything else you'd like to bring up? 21:31:01 not from me, thanks 21:32:13 oh! one more thing: i'll be out of town next week -- in light of the ptg, i might see if mattoliverau is interested in running the meeting, but i think it's fair to assume we'll skip it 21:33:09 i'll make sure the agenda gets updated 21:33:17 good to know, I'll be in later summer vacation in the next week. 21:33:28 so I am going to be absent too. 21:33:49 yeah, we may as well skip then :-) 21:34:04 all right, i'll let you get on with your morning then 21:34:14 thanks timburke 21:34:17 thanks for coming kota! enjoy your vacation 21:34:21 #endmeeting