| *** dasp_ has joined #openstack-swift | 00:28 | |
| *** dasp has quit IRC | 00:29 | |
| *** m75abrams has joined #openstack-swift | 04:19 | |
| *** evrardjp has quit IRC | 04:33 | |
| *** evrardjp has joined #openstack-swift | 04:33 | |
| *** irclogbot_1 has quit IRC | 06:05 | |
| *** irclogbot_0 has joined #openstack-swift | 06:07 | |
| *** psachin has joined #openstack-swift | 07:26 | |
| *** dosaboy has joined #openstack-swift | 07:43 | |
| -openstackstatus- NOTICE: uWSGI made a new release that breaks devstack, please refrain from rechecking until a devstack fix is merged. | 07:44 | |
| *** rpittau|afk is now known as rpittau | 08:00 | |
| *** ccamacho has joined #openstack-swift | 08:36 | |
| *** mvkr has joined #openstack-swift | 09:48 | |
| *** rpittau is now known as rpittau|bbl | 10:03 | |
| *** ccamacho has quit IRC | 10:20 | |
| *** rcernin has joined #openstack-swift | 10:44 | |
| *** rcernin has quit IRC | 10:54 | |
| *** ccamacho has joined #openstack-swift | 10:56 | |
| *** rpittau|bbl is now known as rpittau | 12:04 | |
| *** psachin has quit IRC | 13:15 | |
| *** jv__ has joined #openstack-swift | 13:52 | |
| *** jv__ has quit IRC | 13:57 | |
| *** jv__ has joined #openstack-swift | 13:57 | |
| *** jv has joined #openstack-swift | 13:58 | |
| *** gyee has joined #openstack-swift | 16:03 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: py3: Fix expirer container generation https://review.opendev.org/735262 | 16:11 |
|---|---|---|
| timburke | so i got a nudge last week about doing a stable release for train -- and i realized that there are a couple py3 issues in flight that we probably ought to backport | 16:12 |
| timburke | that's one ^^^ -- the other is https://review.opendev.org/#/c/734721/ (though it maybe needs tests -- there's some exercising of it with the tests added by https://review.opendev.org/#/c/685455/ at least) | 16:13 |
| patchbot | patch 734721 - swift - py3: (Better) fix percentages in configs - 4 patch sets | 16:13 |
| patchbot | patch 685455 - swift - py3: Allow percentages in configs (MERGED) - 1 patch set | 16:13 |
| *** rpittau is now known as rpittau|afk | 16:17 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: Get rid of memcache attr on proxy-server app https://review.opendev.org/735484 | 16:50 |
| timburke | i can't help but feel like https://review.opendev.org/#/c/30481/ went a bit too far in trying to have clear_info_cache call into set_info_cache :-/ | 17:08 |
| patchbot | patch 30481 - swift - get_info - removes duplicate code (Take 3) (MERGED) - 17 patch sets | 17:08 |
| openstackgerrit | Tim Burke proposed openstack/swift master: proxy: Stop killing memcache entries on 5xx responses https://review.opendev.org/735359 | 18:19 |
| clayg | do any of our s3api experts know anything about this decorator: https://github.com/openstack/swift/blob/master/swift/common/middleware/s3api/controllers/multi_upload.py#L134 | 18:26 |
| clayg | it seems to fire off a for-realzy-to-disk HEAD request to the container DB for every part upload 🤷♂️ | 18:27 |
| timburke | clayg, yeah, it tries to do it as a get_container_info call, but since we haven't done auth yet to translate to a for-realzy swift path, that basically doesn't work -- which is part of what prompted me to do https://review.opendev.org/#/c/668064/ | 18:38 |
| patchbot | patch 668064 - swift - Give internal clients the option of getting contai... - 5 patch sets | 18:38 |
| timburke | needs tests tho | 18:38 |
| clayg | it's not obvious to me why "doesn't have auth yet" => "go to disk NOW then" | 19:00 |
| *** m75abrams has quit IRC | 19:09 | |
| timburke | basically, no part of the system has the context to know both (1) the real swift path and (2) that s3api would be perfectly content to get this info from memcache | 19:09 |
| timburke | in s3api, we know we're happy with memcache results, and will even try to do the normal get_container_info thing if we expect that'll work (ie, if we've already made an auth'ed request) | 19:10 |
| timburke | but if we haven't authed, we expect get_container_info to fail -- without a real swift path, get_container_info's gonna fail: https://github.com/openstack/swift/blob/2.25.0/swift/proxy/controllers/base.py#L337-L340 | 19:11 |
| timburke | so our only real option is to do a HEAD instead | 19:11 |
| clayg | what about these TEST requests that the ACL stuff will do? | 19:12 |
| clayg | is there any request besides a HEAD on the container we could throw at swift that will make it through auth and return the account name? | 19:12 |
| timburke | might be a viable alternative, get us to the auth'ed state without any backend requests 🤔 | 19:12 |
| zaitcev | I do not understand. Do you want to take AUTH_xxxxx from the container path and reverse-lookup the account name in the auth system? | 19:16 |
| clayg | this doens't obviously break; but I'm not sure what it's doing exactly -> https://gist.github.com/clayg/72494e96b616c92fb0ea95e062f8793d | 19:18 |
| zaitcev | e.g. KSAU_7f80fce9f306430f87aaaebf29ec15c7 -> ksadmten | 19:18 |
| clayg | zaitcev: I think when a s3 client wants to HEAD their bucket "foo" - we don't know (yet) what the swift path will be | 19:18 |
| clayg | at some point somehow the bucket controller seems to figure it out - and there was "cheat" where calling get_container_info would sometimes just use a bucket HEAD | 19:19 |
| zaitcev | I thought that we cheated by examining the auth token. The token says the account that obtained it. Then, we assume that account as the account under which we search for a container with the same name as S3 bucket. That is what makes our buckets non-global. | 19:24 |
| timburke | zaitcev, we do get back an auth token from keystone and use that to map to the account, but that's down in s3token, away from the caller in s3api | 19:42 |
| timburke | (and of course, other auth systems do something similar but a little different) | 19:42 |
| timburke | clayg, that gist seems reasonable | 19:43 |
| clayg | it seems totally opaque to me - i do NOT understand what these TEST requests are | 19:43 |
| clayg | but it does avoid going to disk on part upload | 19:44 |
| timburke | TEST is relying on the proxy server's handling around https://github.com/openstack/swift/blob/master/swift/proxy/server.py#L544-L546 | 19:45 |
| clayg | oh, so we expect a 405 response - seems fine to me - but it would have gone ahead and AUTH'd!? that's *awesome*!? right? | 20:03 |
| *** evrardjp has quit IRC | 20:11 | |
| *** evrardjp has joined #openstack-swift | 20:13 | |
| timburke | gotta make sure that if some middleware actually implemented a TEST method that they'd be allowed, right? ;-) | 20:22 |
| *** ccamacho has quit IRC | 20:51 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: Get rid of memcache attr on proxy-server app https://review.opendev.org/735484 | 20:57 |
| openstackgerrit | Clay Gerrard proposed openstack/swift master: s3api: Don't do naive HEAD request for auth https://review.opendev.org/735738 | 21:19 |
| *** threestrands has joined #openstack-swift | 21:36 | |
| *** gmann is now known as gmann_afk | 21:39 | |
| openstackgerrit | Clay Gerrard proposed openstack/swift master: Clean up some proxy tests https://review.opendev.org/735741 | 22:18 |
| *** rcernin has joined #openstack-swift | 23:16 | |
| timburke | thinking about https://bugs.launchpad.net/swift/+bug/1883302 -- reaper and reconciler also use direct_client; is that desirable? | 23:31 |
| openstack | Launchpad bug 1883302 in OpenStack Object Storage (swift) "container-sharder should send stat updates using replication network" [Undecided,Confirmed] | 23:31 |
| timburke | expirer, too, since it uses some reconciler machinery | 23:32 |
| timburke | then there's all the uses of internal_client, too... | 23:34 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!