Friday, 2019-06-28

*** early` has joined #openstack-swift01:08
*** early has quit IRC01:08
*** baojg has joined #openstack-swift01:43
*** threestrands has joined #openstack-swift02:23
*** tonyb has quit IRC03:05
kota_clayg: what are you trying to is for p 667933? Basically, the algorithms is what timburke described above. the acl handler will translate the PUT with acl to PUT/POST to save the existing acls.03:35
patchbot - swift - clarify s3api ACL test HEADs - 1 patch set03:35
*** tonyb has joined #openstack-swift03:36
kota_and is to avoid unnecessary request which will be into same result. so if you find other status (i.e. the case that we cannot skip the extra request), probably we should fix the if condition.03:37
kota_hmm... it looks to be related p 580333 that looks I was looking so i should circle back to the patch what we exactly are changing...03:39
patchbot - swift - Use get_container_info to check existence before c... - 3 patch sets03:39
*** psachin has joined #openstack-swift03:40
timburkekota_, for a bit of context, we've got a customer who was doing some benchmarking/stress-testing using the S3 api and started seeing some strange behavior (like BucketNotFound errors during object PUT) once the container servers got overloaded03:54
timburkethis kicked out and as correctness issues -- better to 503 than say that a container that clearly *does exist* doesn't03:55
patchbotpatch 667411 - swift - Return 503 when primary containers can't respond (MERGED) - 2 patch sets03:55
patchbotpatch 667235 - swift - Don't handle object without container (MERGED) - 1 patch set03:55
timburkeand it renewed interest in whose symptom (large numbers of LockTimeouts) we also saw03:57
patchbotpatch 580333 - swift - Use get_container_info to check existence before c... - 3 patch sets03:57
timburke...and even when it *did* succeed, it'd cause needless memcache churn03:58
timburkeeven with all that, we're worried that the container's still overloaded because of all the HEADs we have to do as part of s3api ... but there isn't really a good way to avoid some of those -- we *have* to get something down the pipeline to let auth translate access_key_id to account04:00
timburke...which led to me playing around with
patchbotpatch 668064 - swift - WIP: Give operators the option of serving containe... - 2 patch sets04:01
kota_timburke: i see, I'll take a look the history but exactly we could have something wrong with overloaded container, especially non-atomic operation (e.g. acls)04:06
*** pcaruana has joined #openstack-swift04:36
openstackgerritTim Burke proposed openstack/python-swiftclient master: Clean up warnings from newer flake8
openstackgerritTim Burke proposed openstack/python-swiftclient master: Fix SLO re-upload
openstackgerritMerged openstack/python-swiftclient master: Isolate docs requirements
*** gkadam has joined #openstack-swift05:50
*** gkadam has quit IRC06:17
*** threestrands has quit IRC06:34
*** gkadam has joined #openstack-swift06:44
*** csmart has quit IRC06:52
*** ccamacho has joined #openstack-swift06:53
*** csmart has joined #openstack-swift07:03
*** rcernin has quit IRC07:16
*** tesseract has joined #openstack-swift07:17
openstackgerritMatthew Oliver proposed openstack/swift master: Auto-sharding: Initial steps
openstackgerritMatthew Oliver proposed openstack/swift master: sharding: first attempt at _elect_leader
*** tesseract has quit IRC07:30
*** rdejoux has joined #openstack-swift07:32
*** tesseract has joined #openstack-swift07:32
openstackgerritMerged openstack/python-swiftclient master: Fix SLO re-upload
*** mikecmpbll has joined #openstack-swift08:05
*** psachin has quit IRC08:08
openstackgerritMerged openstack/python-swiftclient master: Clean up warnings from newer flake8
*** psachin has joined #openstack-swift08:25
openstackgerritzhufl proposed openstack/swift master: Fix invalid assert states
*** psachin has quit IRC09:16
kota_timburke, clayg: After I looked at the changes timburke suggested, I'm realizing the status code map would not follow the recent swift status code changes (perhaps, it has not followed for corner cases ever though)09:21
kota_ok, let me remember the logic. probably the controllers takes a S3Request (or S3ACLRequest) object instead of swob.Request so the acl handler gets something as clayg said,
kota_and exactly as the FIXME said, potentially the secondary POST request happens anyway even the first PUT failed with 5xx or 4xx09:38
kota_oh... but the error resp should be raised instead of just a return09:41
kota_Though, InternalError would be returned but the handler will *raise* error status if swift returns unknown status code.09:44
kota_so I assume when overload happens in the swift cluster, then the PUT container (first request) failed with 503, the second request won't happen...09:45
kota_OIC, maybe I reached out what timburke said.09:47
kota_I might be missing what's going on clayg's but p 668064 looks to mitigate the number of HEAD requests for GET account operation to check the last modified?09:49
patchbot - swift - WIP: Give operators the option of serving containe... - 2 patch sets09:49
kota_if my understanding would be correct, I think the challenge was what I tried a few years ago,
kota_anyway, thanks for sharing that context, I may catch (or follow) up more on that so I'll continue to look at the patches and ircs in the next week.09:55
*** ccamacho has quit IRC10:30
*** ccamacho has joined #openstack-swift10:31
*** ccamacho has quit IRC11:11
*** mvkr has quit IRC12:29
*** tdasilva has joined #openstack-swift12:35
*** ChanServ sets mode: +v tdasilva12:35
*** pcaruana has quit IRC12:40
*** pcaruana has joined #openstack-swift12:41
*** baojg has quit IRC12:51
*** gkadam_ has joined #openstack-swift13:21
*** gkadam has quit IRC13:22
*** baojg has joined #openstack-swift13:36
*** BjoernT has joined #openstack-swift13:42
*** gkadam_ has quit IRC14:21
*** baojg has quit IRC15:19
*** baojg has joined #openstack-swift15:19
*** zaitcev_ has joined #openstack-swift15:20
*** ChanServ sets mode: +v zaitcev_15:20
*** jistr is now known as jistr|off15:20
*** zaitcev has quit IRC15:23
*** baojg has quit IRC15:29
*** jistr|off is now known as jistr15:42
*** zaitcev__ has joined #openstack-swift16:19
*** ChanServ sets mode: +v zaitcev__16:19
*** zaitcev_ has quit IRC16:23
*** BjoernT has quit IRC16:46
*** zaitcev__ has quit IRC16:53
*** zaitcev__ has joined #openstack-swift17:04
*** ChanServ sets mode: +v zaitcev__17:04
*** e0ne has joined #openstack-swift17:26
*** zaitcev__ is now known as zaitcev17:26
*** mikecmpbll has quit IRC17:34
*** rdejoux has quit IRC17:47
*** mvkr has joined #openstack-swift17:52
*** e0ne has quit IRC18:12
*** e0ne has joined #openstack-swift18:13
*** tesseract has quit IRC18:13
*** e0ne has quit IRC18:16
*** e0ne has joined #openstack-swift18:26
*** e0ne has quit IRC18:28
*** zaitcev has quit IRC18:28
*** zaitcev has joined #openstack-swift18:39
*** ChanServ sets mode: +v zaitcev18:39
*** tdasilva has quit IRC19:06
*** BjoernT has joined #openstack-swift19:59
*** BjoernT has quit IRC20:04
zaitcevWhat does Jinja2 do in our lower-constraints.txt?20:06
*** BjoernT has joined #openstack-swift20:08
*** zaitcev has quit IRC20:11
timburkethat's an excellent question :-/ i'm guessing it's a transitive dep of some sort, but i don't remember who brought it in20:21
*** zaitcev has joined #openstack-swift20:24
*** ChanServ sets mode: +v zaitcev20:24
*** mvkr has quit IRC20:40
*** mvkr has joined #openstack-swift20:53
*** mvkr has quit IRC21:07
*** pcaruana has quit IRC21:16
*** mvkr has joined #openstack-swift21:21
timburkezaitcev, looks like sphinx pulls it in. so it probably doesn't strictly need to be in lower-constraints.txt; we don't have a lower-constraints docs build or anything22:04
zaitcevI tried to run my test cluster on py3 and immediatelly object server threw up22:05
zaitcevI thought you were running func tests.22:05
zaitcevNot that I'm disappointed or anything, at last something that I can fix.22:05
timburkeoh good! that means you can review :P22:06
patchbotpatch 665494 - swift - Move call to global_conf_callback after loadapp() - 4 patch sets22:06
zaitcevso that's what it was huh22:06
timburkestill weirds me out a bit that we're getting different paste behavior based on python version... makes me wonder a bit if there's some new interpolation thing that's happening that's gonna break some existing configs...22:07
timburkeiirc, pick up that patch and and you can get func tests passing with tempauth22:09
patchbotpatch 662546 - swift - py3: Be able to read and write non-ASCII headers - 4 patch sets22:09
timburkeadd and they'll pass for keystone too22:10
patchbotpatch 653548 - swift - py3: add swift-dsvm-functional-py3 job - 14 patch sets22:10
timburkewhich we verify in
patchbotpatch 666942 - swift - Run all func tests under py2 against services runn... - 1 patch set22:10
*** zaitcev has quit IRC23:39
*** zaitcev has joined #openstack-swift23:40
*** ChanServ sets mode: +v zaitcev23:40

Generated by 2.15.3 by Marius Gedminas - find it at!