| *** openstackgerrit has quit IRC | 00:49 | |
| *** openstackstatus has quit IRC | 01:20 | |
| *** openstack has joined #openstack-swift | 01:22 | |
| *** ChanServ sets mode: +o openstack | 01:22 | |
| *** rcernin has quit IRC | 02:01 | |
| *** thiago__ has quit IRC | 02:21 | |
| *** rcernin has joined #openstack-swift | 02:25 | |
| *** rcernin has quit IRC | 02:30 | |
| *** rcernin has joined #openstack-swift | 03:04 | |
| *** abelur has quit IRC | 03:26 | |
| *** abelur has joined #openstack-swift | 03:28 | |
| *** rcernin has quit IRC | 03:50 | |
| *** openstackgerrit has joined #openstack-swift | 04:03 | |
| openstackgerrit | Pete Zaitcev proposed openstack/swift master: Add a read-only role to keystoneauth https://review.opendev.org/c/openstack/swift/+/771158 | 04:03 |
|---|---|---|
| *** rcernin has joined #openstack-swift | 04:28 | |
| *** rcernin has quit IRC | 04:40 | |
| openstackgerrit | Matthew Oliver proposed openstack/swift master: Add a config file option to swift-manage-shard-ranges https://review.opendev.org/c/openstack/swift/+/774584 | 04:41 |
| *** rcernin has joined #openstack-swift | 04:41 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: Fix s3api unit tests when run at roughly midnight UTC https://review.opendev.org/c/openstack/swift/+/774589 | 05:28 |
| timburke | well *that* was stupid | 05:28 |
| zaitcev | So, what does midnight have to do with it? | 05:30 |
| zaitcev | What I see is only someone forgot that v2 existed | 05:30 |
| *** evrardjp has quit IRC | 05:33 | |
| *** evrardjp has joined #openstack-swift | 05:33 | |
| zaitcev | Or wait. It's all v4 | 05:34 |
| zaitcev | I see, one of them supplied by the client (I guess) and other is generated from the server time. | 05:43 |
| zaitcev | Still, it's unclear what midnight has with anything. | 05:44 |
| openstackgerrit | Matthew Oliver proposed openstack/swift master: Add shrink candidates to recon dump https://review.opendev.org/c/openstack/swift/+/772624 | 05:46 |
| openstackgerrit | Matthew Oliver proposed openstack/swift master: Add shrink candidates to recon dump https://review.opendev.org/c/openstack/swift/+/772624 | 05:53 |
| *** dsariel has joined #openstack-swift | 05:53 | |
| *** dsariel has quit IRC | 05:58 | |
| *** gyee has quit IRC | 06:11 | |
| openstackgerrit | Merged openstack/swift master: Add a read-only role to keystoneauth https://review.opendev.org/c/openstack/swift/+/771158 | 06:34 |
| *** rcernin has quit IRC | 07:15 | |
| *** rcernin has joined #openstack-swift | 07:32 | |
| *** rcernin has quit IRC | 07:43 | |
| *** rpittau|afk is now known as rpittau | 07:59 | |
| *** rcernin has joined #openstack-swift | 08:10 | |
| *** rcernin has quit IRC | 08:20 | |
| *** dsariel has joined #openstack-swift | 08:35 | |
| *** dsariel has quit IRC | 08:41 | |
| *** dsariel has joined #openstack-swift | 08:41 | |
| *** cschwede has joined #openstack-swift | 09:41 | |
| *** ChanServ sets mode: +v cschwede | 09:41 | |
| *** evrardjp has quit IRC | 11:10 | |
| *** coreycb has quit IRC | 11:10 | |
| *** nicolasbock has quit IRC | 11:11 | |
| *** coreycb has joined #openstack-swift | 11:11 | |
| *** nicolasbock has joined #openstack-swift | 11:11 | |
| *** evrardjp has joined #openstack-swift | 11:12 | |
| *** m75abrams has quit IRC | 11:13 | |
| *** m75abrams has joined #openstack-swift | 11:27 | |
| *** coreycb has quit IRC | 12:08 | |
| *** nicolasbock has quit IRC | 12:09 | |
| *** coreycb has joined #openstack-swift | 12:10 | |
| *** nicolasbock has joined #openstack-swift | 12:12 | |
| *** rcernin has joined #openstack-swift | 12:27 | |
| *** rcernin has quit IRC | 12:31 | |
| *** sorrison has quit IRC | 12:48 | |
| *** sorrison has joined #openstack-swift | 12:49 | |
| openstackgerrit | Merged openstack/swift master: relinker: Pull arg parsing into module https://review.opendev.org/c/openstack/swift/+/773571 | 14:26 |
| openstackgerrit | Merged openstack/swift master: relinker: Allow conf files for configuration https://review.opendev.org/c/openstack/swift/+/773572 | 14:27 |
| *** fingo has quit IRC | 14:46 | |
| *** m75abrams has quit IRC | 14:47 | |
| *** m75abrams has joined #openstack-swift | 14:48 | |
| *** stand has joined #openstack-swift | 14:51 | |
| *** fingo has joined #openstack-swift | 14:54 | |
| zigo | timburke: On my preprod cluster, I have proxy, account and containers using uwsgi. This works well. However, when I switch the object server to uwsgi as well, when I upload an object, I get this: | 15:10 |
| zigo | http://paste.openstack.org/show/802480/ | 15:10 |
| zigo | Yet another Eventlet woopsy ... Any idea what this could be ? | 15:10 |
| zigo | On the object server side, it logs a HTTP 499 error. | 15:10 |
| zigo | As if the swift-proxy was disconnecting... | 15:11 |
| zigo | I'm having a hard time to understand who's at fault (ie: the proxy or the server). | 15:11 |
| zigo | Could it be that uwsgi isn't supporting the chuncks that the swift-proxy is trying to use? | 15:11 |
| openstackgerrit | Alistair Coles proposed openstack/swift master: sharder: Ignore already shrinking sequence when compacting https://review.opendev.org/c/openstack/swift/+/774560 | 15:36 |
| openstackgerrit | Alistair Coles proposed openstack/swift master: find_compactable_shard_sequences: fix skipped range https://review.opendev.org/c/openstack/swift/+/774684 | 15:36 |
| *** m75abrams has quit IRC | 16:04 | |
| timburke | good morning | 16:16 |
| zigo | Hi ! | 16:18 |
| timburke | zaitcev, the problem was entirely in the test -- v4 signatures have the request-signing-day as part of the credential scope, and the test would always use "now" even when it was moving the signing time around a bit to now +/- some time deltas | 16:18 |
| timburke | zigo, looking... | 16:18 |
| *** m75abrams has joined #openstack-swift | 16:19 | |
| timburke | is that for a replicated or EC policy? is encryption enabled? | 16:19 |
| zaitcev | timburke: I understand, but why midnight? There's no wrap-around issue that I can see (using the hour of the day without carry into the day - it's all in seconds everywhere). | 16:20 |
| zigo | No encryption, just normal replicated setup. | 16:20 |
| zigo | If I switch back the object-server to Eventlet, everything goes back to working (though, still this LogThread issue... which is what I'm trying to solve with uwsgi). | 16:22 |
| zigo | Note that proxy, account and container servers are all already working very well with uwsgi, as much as I can tell. | 16:23 |
| timburke | zaitcev, it's that the date in the credential scope wasn't matching the date in the x-amz-date header -- so you get failures like with https://zuul.opendev.org/t/openstack/build/f407f778665a4bcf81790eb1bfd1c479/log/job-output.txt#9421-9435 where the credential had a date of 2021-02-08 but the header had a date of 2021-02-09 because the test kicked it 10 minutes into the future | 16:26 |
| *** rcernin has joined #openstack-swift | 16:27 | |
| *** rcernin has quit IRC | 16:32 | |
| timburke | zigo, so there are definitely some eventlet-isms in the object server -- look for set_hundred_continue_response_headers and send_hundred_continue_response in particular -- but i wouldn't have expected that to be needed for non-encrypted replicated PUTs | 16:36 |
| timburke | (see also, https://bugs.launchpad.net/swift/+bug/1496636) | 16:36 |
| openstack | Launchpad bug 1496636 in OpenStack Object Storage (swift) "EC: Chunked transfer/commit protocol is *not* HTTP" [Medium,In progress] - Assigned to Pete Zaitcev (zaitcev) | 16:36 |
| zigo | timburke: That's precisely what I encountered, and double-guessed (ie: that there's an issue with Transfer-Encoding: chunked). | 16:37 |
| zigo | timburke: So, this means object-server is not uwsgi ready, right? | 16:37 |
| zaitcev | zigo: That bug only deals with EC which has commit phase. | 16:38 |
| zigo | Well, in production, we might use EC and object encryption ... :/ | 16:38 |
| timburke | certainly not if you want a full-featured swift. this case (non-encrypted, replicated) should be simple enough that standard 100-continue support is all that's needed (and not the crazy hooks that we got into eventlet) | 16:39 |
| timburke | i wonder if uwsgi doesn't support 100-continue *at all*, though -- the proxy would send the Expect header and wait for the object servers to ack before sending data; meanwhile, the object servers ignore this header they don't understand and keep waiting for the proxy to send data before sending a response | 16:40 |
| timburke | :-/ this sure *looks* like support, though: https://github.com/unbit/uwsgi/blob/ce787a6e3210333a71f5074e5775885428e8b73c/plugins/http/http.c#L1259-L1262 | 16:52 |
| timburke | might need to tcpdump the connections to the backend to figure out what going on on-the-wire, see who's not holding up their end of the conversation | 16:55 |
| *** m75abrams has quit IRC | 17:00 | |
| zigo | timburke: I *do* want full feature (ie: EC + object encryption), so I will leave it using Eventlet then. | 17:04 |
| *** gyee has joined #openstack-swift | 17:20 | |
| *** dsariel has quit IRC | 17:25 | |
| *** dsariel has joined #openstack-swift | 17:26 | |
| *** rpittau is now known as rpittau|afk | 17:56 | |
| *** cschwede has quit IRC | 19:29 | |
| openstackgerrit | Clay Gerrard proposed openstack/swift master: Shink the public shrinking interface https://review.opendev.org/c/openstack/swift/+/774731 | 20:39 |
| *** timburke_ has joined #openstack-swift | 21:09 | |
| *** ChanServ sets mode: +v timburke_ | 21:09 | |
| *** timburke has quit IRC | 21:11 | |
| *** ianychoi__ has joined #openstack-swift | 21:40 | |
| *** ianychoi_ has quit IRC | 21:43 | |
| *** rcernin has joined #openstack-swift | 21:50 | |
| *** gmann is now known as gmann_afk | 22:13 | |
| *** timburke_ has quit IRC | 22:49 | |
| *** timburke__ has joined #openstack-swift | 22:50 | |
| *** ChanServ sets mode: +v timburke__ | 22:50 | |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!