openstackgerrit | Tim Burke proposed openstack/swift master: s3api: Get rid of slo_enabled flag https://review.opendev.org/c/openstack/swift/+/770685 | 00:51 |
---|---|---|
zaitcev | Does anyone else have a weird interaction of s3api and ratelimit? | 01:02 |
zaitcev | Oh, actually etc/proxy-server.conf-sample talks about it. | 01:04 |
openstackgerrit | Tim Burke proposed openstack/swift master: Replace S3Request.location with S3Request.s3api_region https://review.opendev.org/c/openstack/swift/+/770690 | 01:20 |
timburke_ | i wonder if i should just take the region-for-the-sake-of-auth from the Authorization header, and stop looking at the configured location in S3Request entirely... | 01:22 |
openstackgerrit | Tim Burke proposed openstack/swift master: Replace S3Request.location with S3Request.s3api_region https://review.opendev.org/c/openstack/swift/+/770690 | 01:23 |
timburke_ | (forgot about fixing up tests) | 01:24 |
*** hoonetorg has joined #openstack-swift | 01:42 | |
timburke_ | zaitcev, note that prior to https://github.com/openstack/swift/commit/1db11df you kinda had to have ratelimit right of auth (or put up with swift requests double-counting against your limits, if you did a double-placement anyway) | 01:42 |
*** timburke_ has quit IRC | 01:43 | |
*** baojg has quit IRC | 02:06 | |
*** baojg has joined #openstack-swift | 02:06 | |
openstackgerrit | Tim Burke proposed openstack/swift master: prefactor: Pass s3api Config to S3Requests https://review.opendev.org/c/openstack/swift/+/729091 | 02:08 |
openstackgerrit | Tim Burke proposed openstack/swift master: s3api: Get rid of slo_enabled flag https://review.opendev.org/c/openstack/swift/+/770685 | 02:08 |
openstackgerrit | Tim Burke proposed openstack/swift master: Replace S3Request.location with S3Request.s3api_region https://review.opendev.org/c/openstack/swift/+/770690 | 02:08 |
_david_sohonet | If I'm not using sharding, is it correct to ignore the 'states' URL request param ? I'm seeing in proxy-server log lines ? I've not noticed it before. I assume it's added after the request hits the proxy-server ? | 02:53 |
*** rcernin has quit IRC | 03:04 | |
*** rcernin has joined #openstack-swift | 03:25 | |
*** rcernin has quit IRC | 03:27 | |
*** rcernin has joined #openstack-swift | 03:31 | |
*** rcernin has quit IRC | 03:36 | |
*** rcernin has joined #openstack-swift | 03:37 | |
*** rcernin has quit IRC | 03:38 | |
*** rcernin has joined #openstack-swift | 03:38 | |
*** rcernin has quit IRC | 03:39 | |
*** rcernin has joined #openstack-swift | 03:39 | |
*** psachin has joined #openstack-swift | 03:58 | |
*** evrardjp has quit IRC | 05:33 | |
*** evrardjp has joined #openstack-swift | 05:35 | |
*** m75abrams has joined #openstack-swift | 06:06 | |
*** gyee has quit IRC | 06:31 | |
openstackgerrit | Matthew Oliver proposed openstack/swift master: Reclaim shards when root container is reclaimed https://review.opendev.org/c/openstack/swift/+/770529 | 06:41 |
*** dsariel has joined #openstack-swift | 06:54 | |
*** rcernin has quit IRC | 07:40 | |
*** rpittau|afk is now known as rpittau | 07:47 | |
*** rcernin has joined #openstack-swift | 07:56 | |
*** rcernin has quit IRC | 08:01 | |
*** rcernin has joined #openstack-swift | 08:26 | |
*** m75abrams has quit IRC | 09:45 | |
*** m75abrams has joined #openstack-swift | 09:49 | |
*** m75abrams has quit IRC | 09:51 | |
*** m75abrams has joined #openstack-swift | 09:53 | |
*** rcernin has quit IRC | 10:14 | |
*** rcernin has joined #openstack-swift | 10:23 | |
*** rcernin has quit IRC | 10:40 | |
*** baojg has quit IRC | 11:09 | |
*** baojg has joined #openstack-swift | 11:11 | |
*** baojg has quit IRC | 11:11 | |
*** baojg has joined #openstack-swift | 11:12 | |
*** baojg has quit IRC | 11:12 | |
*** baojg has joined #openstack-swift | 11:12 | |
*** baojg has quit IRC | 11:13 | |
*** baojg has joined #openstack-swift | 11:13 | |
*** baojg has quit IRC | 11:14 | |
*** baojg has joined #openstack-swift | 11:14 | |
*** baojg has quit IRC | 11:14 | |
*** baojg has joined #openstack-swift | 11:15 | |
*** baojg has quit IRC | 11:15 | |
openstackgerrit | Alistair Coles proposed openstack/swift master: s3api: Make allowable clock skew configurable https://review.opendev.org/c/openstack/swift/+/770005 | 11:53 |
*** rcernin has joined #openstack-swift | 11:55 | |
openstackgerrit | Alistair Coles proposed openstack/swift master: Make test_verify_server independent of test env https://review.opendev.org/c/openstack/swift/+/770796 | 11:58 |
*** rcernin has quit IRC | 11:58 | |
*** psachin has quit IRC | 12:39 | |
*** takamatsu has joined #openstack-swift | 14:35 | |
*** m75abrams has quit IRC | 14:48 | |
acoles | _david_sohonet: | 14:52 |
acoles | _david_sohonet: yes, the 'states' param has no effect with unsharded containers, and yes it is added by the proxy to GET requests to backend container servers | 14:53 |
*** m75abrams has joined #openstack-swift | 15:07 | |
*** biyiklioglu has joined #openstack-swift | 15:07 | |
*** biyiklioglu has quit IRC | 15:16 | |
*** biyiklioglu has joined #openstack-swift | 15:25 | |
*** hoonetorg has quit IRC | 15:35 | |
*** hoonetorg has joined #openstack-swift | 15:40 | |
*** hoonetorg has quit IRC | 15:43 | |
*** gyee has joined #openstack-swift | 16:22 | |
*** dosaboy has quit IRC | 16:24 | |
*** dosaboy has joined #openstack-swift | 16:25 | |
*** hoonetorg has joined #openstack-swift | 16:45 | |
*** m75abrams has quit IRC | 16:57 | |
*** diablo_rojo has joined #openstack-swift | 17:04 | |
*** rpittau is now known as rpittau|afk | 17:42 | |
openstackgerrit | Alistair Coles proposed openstack/swift master: ssync: sync non-durable fragments from handoffs https://review.opendev.org/c/openstack/swift/+/770047 | 18:08 |
acoles | wow, that ^^^ has conflicts with more than 50 other patches, could be time for some abandoning | 18:12 |
*** dosaboy has quit IRC | 18:21 | |
*** diablo_rojo has quit IRC | 18:34 | |
*** diablo_rojo has joined #openstack-swift | 18:43 | |
*** timburke has joined #openstack-swift | 18:48 | |
*** ChanServ sets mode: +v timburke | 18:48 | |
timburke | hmm. do we not have any tests that exercise somethig like `swift-init container.1 start`? i think that may have stopped working following https://review.opendev.org/c/openstack/swift/+/764137 ... | 18:51 |
*** dosaboy has joined #openstack-swift | 19:11 | |
*** timburke_ has joined #openstack-swift | 19:12 | |
*** ChanServ sets mode: +v timburke_ | 19:12 | |
clayg | i don't think i've ever used that interface - but i feel like I used to be able to say something like `swift-init status proxy.noauth` | 19:12 |
clayg | `swift-init container status -c 1` works fine - maybe we could expand -c to accept strings (i.e. `swift-init status proxy -c noauth`) | 19:13 |
timburke_ | i think i only noticed the .N thing after you pointed out that `swift-init status proxy.noauth` is possible ;-) | 19:14 |
clayg | yeah, love that feature | 19:14 |
*** timburke has quit IRC | 19:15 | |
clayg | but `-c noauth` would be entirely acceptable - i always thought `container.1` was silly | 19:15 |
*** fingo has quit IRC | 20:15 | |
timburke_ | huh. so proxy.noauth gets parsed in Server.__init__ (https://github.com/openstack/swift/blob/2.26.0/swift/common/manager.py#L445-L448) and gets used to limit our search in _find_conf_files (https://github.com/openstack/swift/blob/2.26.0/swift/common/manager.py#L505-L511) | 20:16 |
timburke_ | while -c/--config-num comes gets used as number around https://github.com/openstack/swift/blob/2.26.0/swift/common/manager.py#L541-L548 and really *does* have to be a int... | 20:16 |
*** fingo has joined #openstack-swift | 20:18 | |
clayg | nah, just change the type of `-c` to be a string, set it to .conf and let _find_conf_files do it's thing (the index trick is lame) | 20:19 |
timburke_ | clayg, fwiw, you could just do `swift-init status proxy -c 1` *today* to have it only look at your noauth proxy | 20:20 |
clayg | yeah, but that's lame | 20:22 |
*** dosaboy has quit IRC | 21:00 | |
*** dosaboy has joined #openstack-swift | 21:02 | |
timburke_ | ugh. yeah, i need to get overall retries in the gate for func tests -- it sucks to have a patch wait in the check queue for 25hrs just to fail because of container-listing consistency issues in func tests :-( | 21:16 |
*** jrosser has quit IRC | 21:29 | |
*** jrosser has joined #openstack-swift | 21:32 | |
*** biyiklioglu has quit IRC | 21:39 | |
mattoliverau | morning | 21:44 |
*** diablo_rojo has quit IRC | 21:48 | |
*** rcernin has joined #openstack-swift | 22:05 | |
*** dsariel has quit IRC | 22:29 | |
openstackgerrit | Tim Burke proposed openstack/swift master: swift-init: Re-enable targeted configs https://review.opendev.org/c/openstack/swift/+/770871 | 22:45 |
openstackgerrit | Tim Burke proposed openstack/swift master: swift-init: Include message when no executables found https://review.opendev.org/c/openstack/swift/+/770872 | 22:45 |
openstackgerrit | Tim Burke proposed openstack/swift master: Monkey-patch in tests using mock.patch https://review.opendev.org/c/openstack/swift/+/770873 | 22:45 |
_david_sohonet | https://github.com/eventlet/eventlet/commit/9f49f0bbb9c18d7478d7c5e2e677ae6452845ab5, but I see you author | 23:01 |
_david_sohonet | er a later patch on the same area of code: https://github.com/eventlet/eventlet/commit/af253ad97d6e33e86aca56ea5ce5ede1ed0a51cb . Should I take the first but not the second ? | 23:01 |
_david_sohonet | Sorry about the noise ^. I suck at IRC clearly.. | 23:01 |
_david_sohonet | clayg: you mentioned your proxy-servers were more stable with https://github.com/eventlet/eventlet/commit/9f49f0bbb9c18d7478d7c5e2e677ae6452845ab5, but I see you authorer a later patch on the same area of code: https://github.com/eventlet/eventlet/commit/af253ad97d6e33e86aca56ea5ce5ede1ed0a51cb . Should I take the first but not the second ? | 23:01 |
timburke_ | _david_sohonet, you want both; they both came in as part of https://github.com/eventlet/eventlet/pull/645 | 23:02 |
clayg | 👍 whatever is the latest eventlet is fine. | 23:03 |
_david_sohonet | Ok, thanks timburke_ . As an update on the deadlock issue, I pointed the bad app (ffmpeg - encoding video with a TempURL as the video source - lots and lots of 499s) at a canary proxy server and it seems to confirm the app is the issue. | 23:05 |
DHE | there's an official fix for the logging deadlock? | 23:05 |
timburke_ | no, not yet, sorry :-( | 23:05 |
DHE | aww... | 23:05 |
timburke_ | i should really make another attempt at it, even if it's just to have the proxy-server turn off logging.logThreads or whatever during bring-up | 23:06 |
timburke_ | _david_sohonet, do you have a good repro script? that'd be *amazing*! part of my trouble thus far has been how sporadically i'd see the issue in my (very low-traffic) home cluster | 23:07 |
DHE | I don't know if it's worth trying to apply on other servers. while the proxy server is probably better behaved than the average client it's probably better to be safe | 23:07 |
timburke_ | does it usually just need to try re-encoding one (large-ish) object, or would i want to try it in a loop for a while? | 23:08 |
DHE | my workload is customer hardware downloading smallish (4 MB or so) objects. I believe there is an above average/high rate of users aborting transfers though | 23:11 |
openstackgerrit | Tim Burke proposed openstack/swift master: Turn off logging.logThreads when monkey-patched https://review.opendev.org/c/openstack/swift/+/752593 | 23:25 |
timburke_ | DHE, _david_sohonet you guys might try ^^^ (and no stdlib changes) -- i'd be real interested in feedback on it! | 23:26 |
openstackgerrit | Merged openstack/swift master: Fix __exit__ calls https://review.opendev.org/c/openstack/swift/+/770666 | 23:51 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!