| *** rcernin has quit IRC | 02:19 | |
| DHE | I recall someone made a whole object server implementation - I think it was a new app rather than a plugin for the existing object servers - that used rados and could store objects in a Ceph pool... which seems backwards to me but whatever | 02:21 |
|---|---|---|
| DHE | so I guess anything's possible? | 02:21 |
| kota_ | DHE: are you meaning sort of swiftonfile? https://opendev.org/x/swiftonfile | 02:35 |
| kota_ | it's fs translation on the swift backend (i.e. object servers access to some different fs like gluster fs). | 02:37 |
| DHE | no, this was some 3rd party project that makes something that stands in as an object server and actually stores data in Ceph | 02:40 |
| DHE | this makes is feasible to have only 1 replica for that storage policy since ceph already has its own redundancy... | 02:40 |
| DHE | but I mean... it's a kinda weird use case | 02:40 |
| *** rcernin has joined #openstack-swift | 02:43 | |
| *** rcernin has quit IRC | 02:43 | |
| *** rcernin has joined #openstack-swift | 02:44 | |
| *** psachin has joined #openstack-swift | 03:38 | |
| kota_ | sounds similar because swiftonfile also assume glusterfs creates some replicas so the swift configuration will be 1 replica. | 03:46 |
| *** evrardjp has quit IRC | 04:33 | |
| *** evrardjp has joined #openstack-swift | 04:33 | |
| *** m75abrams has joined #openstack-swift | 05:32 | |
| *** viks____ has joined #openstack-swift | 06:42 | |
| *** ccamacho has joined #openstack-swift | 06:47 | |
| *** rpittau|afk is now known as rpittau | 06:51 | |
| *** ccamacho has quit IRC | 07:02 | |
| *** takamatsu is now known as mauro|hackfest-o | 07:08 | |
| *** mauro|hackfest-o is now known as mauro|hackfest | 07:09 | |
| *** mauro|hackfest is now known as takamatsu | 07:09 | |
| *** ccamacho has joined #openstack-swift | 07:29 | |
| *** rcernin has quit IRC | 07:52 | |
| *** hoonetorg has quit IRC | 10:01 | |
| *** tkajinam has quit IRC | 10:05 | |
| *** rpittau is now known as rpittau|bbl | 10:21 | |
| *** hoonetorg has joined #openstack-swift | 11:37 | |
| *** jv__ has joined #openstack-swift | 11:45 | |
| *** rpittau|bbl is now known as rpittau | 12:00 | |
| *** jv__ has quit IRC | 12:33 | |
| *** manuvakery has joined #openstack-swift | 13:23 | |
| *** jv__ has joined #openstack-swift | 14:57 | |
| *** m75abrams has quit IRC | 14:58 | |
| *** jv__ has quit IRC | 15:02 | |
| *** gyee has joined #openstack-swift | 15:03 | |
| *** psachin has quit IRC | 15:07 | |
| *** zaitcev has joined #openstack-swift | 15:12 | |
| *** ChanServ sets mode: +v zaitcev | 15:12 | |
| *** manuvakery has quit IRC | 15:33 | |
| ormandj | is it expected that the sample proxy-server.conf has differences from what's referenced in docs? for example, there is no service_token_roles_required in the example config, in fact, there's no [keystone_authtoken] section at all, as referenced in the ocata release notes. but there's plenty of deprecation warnings in the logs ;) and right now, all the config lives in the [authtoken] subsection | 15:49 |
| ormandj | https://docs.openstack.org/keystonemiddleware/latest/middlewarearchitecture.html would indicate to expect a keystone_authtoken section | 15:50 |
| ormandj | https://github.com/openstack/swift/blob/stable/ussuri/etc/proxy-server.conf-sample has no such | 15:51 |
| ormandj | we've also got another fun one with a bunch of warnings about a bunch of options being unknown: https://pastebin.com/raw/0RiUX36L (seems same as: https://ask.openstack.org/en/question/128361/proxy-server-option-not-known-to-keystonemiddleware/ ) | 15:58 |
| ormandj | still seems to function ok, in our case, aside from warnings | 15:59 |
| ormandj | service_token_roles_required doesn't even appear in the sample config, just see it in the deprecation warning on startup | 16:04 |
| *** manuvakery has joined #openstack-swift | 16:43 | |
| *** ccamacho has quit IRC | 17:26 | |
| *** ccamacho has joined #openstack-swift | 17:34 | |
| clayg | timburke: what was the motivation for moving .misplaced_objects out of the reconciler and into request helpers? | 18:52 |
| timburke | clayg, import loop -- i forget exactly what the cycle was | 18:53 |
| timburke | i agree, though -- i'd kinda prefer to have that constant in reconciler.py 😞 | 18:54 |
| clayg | change looks great, i'll do some greps and kick the tires | 18:54 |
| timburke | yeah, i'm pretty happy with where it's landed -- and feeling better about the lack of support for legacy behavior after shopping it around in the meeting a bit and talking to other operators like ormandj | 18:55 |
| timburke | clayg, the more i poke at p 711342, the more p 733911 feels like a good idea. thanks for shaving that yak! | 19:00 |
| patchbot | https://review.opendev.org/#/c/711342/ - swift - Add concurrent_gets to EC GET requests - 12 patch sets | 19:00 |
| patchbot | https://review.opendev.org/#/c/733911/ - swift - Start to decouple Object GET path - 2 patch sets | 19:00 |
| *** hoonetorg has quit IRC | 19:58 | |
| *** jv__ has joined #openstack-swift | 20:06 | |
| clayg | timburke: uhhh... https://bugs.launchpad.net/swift/+bug/1446873 | 20:11 |
| openstack | Launchpad bug 1446873 in OpenStack Object Storage (swift) "ssync doesn't work with replication_server = true" [Medium,Confirmed] | 20:11 |
| *** hoonetorg has joined #openstack-swift | 20:12 | |
| timburke | well shoot :-( | 20:14 |
| timburke | good find, though! | 20:14 |
| clayg | there's a few daemons that are still using http_connect directly - and if they're doing "public" requests to storage servers they're using the client ips | 20:15 |
| clayg | but the consensus on the bug seemed to be that we can just rework base_storage_server allowed methods | 20:16 |
| timburke | makes sense -- it sure seems to me like the @replication decorator ought to be flipped to mean "this method is only allowed on replication servers" instead of "replication servers only expose these methods" | 20:18 |
| clayg | timburke: 👍 💯 | 20:21 |
| ormandj | timburke: btw, py3 storage + py2 proxy == not great time :) we are deploying ussuri + your patch right now | 20:53 |
| ormandj | https://bugs.launchpad.net/swift/+bug/1863053 appears to be biting us pretty hard | 20:53 |
| openstack | Launchpad bug 1863053 in OpenStack Object Storage (swift) "Swift on py3 doesn't like message/rfc822 data" [High,Fix released] - Assigned to Tim Burke (1-tim-z) | 20:53 |
| ormandj | will let you know how the whole $ business works out | 20:53 |
| timburke | ormandj, sorry -- there's a fix on stable/train, but i hadn't gotten to doing a fresh release :-( | 20:56 |
| timburke | but now i'm torn about whether i should wait on also backporting https://review.opendev.org/#/c/737856/ | 20:57 |
| patchbot | patch 737856 - swift - py3: Stop munging RAW_PATH_INFO - 2 patch sets | 20:57 |
| ormandj | yeah, we see it, we're also just releasing ussuri right now + that patch | 20:57 |
| timburke | clayg, ha! i just realized that because of the way the @public and @replication decorators work, you can't issue OPTIONS requests against replication servers 😆 | 22:11 |
| clayg | poor replication_server = true option 👎 | 22:13 |
| *** rcernin has joined #openstack-swift | 22:18 | |
| *** rcernin has quit IRC | 22:19 | |
| timburke | it's a great option! it just needs to work better ;-) | 22:19 |
| *** rcernin has joined #openstack-swift | 22:19 | |
| timburke | though -- i'm starting to worry a bit about what it'd take to get probe tests able to pass with split servers like that... | 22:23 |
| clayg | Yeah. We’ll have to verify manually first. This may take a while. | 22:26 |
| clayg | Thanks for looking at waterfall! I’ll respond tomorrow before I package. | 22:26 |
| *** tkajinam has joined #openstack-swift | 22:47 | |
| zaitcev | Guys | 22:49 |
| zaitcev | How do I prevent sharder from starting? | 22:49 |
| zaitcev | It is part of swift-init's rest. | 22:49 |
| zaitcev | For things like expirer it's possible to remove the config file. But I guess not anymore? | 22:50 |
| zaitcev | Should I comment out the [container-sharder] section? | 22:50 |
| *** jv__ has quit IRC | 23:31 | |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!