| *** lyrrad has quit IRC | 01:54 | |
| *** openstack has joined #swift3 | 13:23 | |
| *** acoles is now known as acoles_ | 13:47 | |
| *** acoles_ is now known as acoles | 14:23 | |
| *** chsc has joined #swift3 | 15:17 | |
| *** chsc has quit IRC | 16:40 | |
| *** lyrrad has joined #swift3 | 17:04 | |
| *** chsc has joined #swift3 | 17:10 | |
| *** chsc has quit IRC | 19:41 | |
| *** acoles is now known as acoles_ | 19:52 | |
| *** chsc has joined #swift3 | 20:05 | |
| *** acoles_ is now known as acoles | 21:00 | |
| *** acoles is now known as acoles_ | 22:46 | |
| *** chsc has quit IRC | 22:52 | |
| *** chsc has joined #swift3 | 22:57 | |
| kota_ | hello | 23:00 |
|---|---|---|
| chsc | hello | 23:01 |
| kota_ | bi-weekly meeting time. | 23:02 |
| kota_ | chsc: o/ | 23:02 |
| kota_ | anyone else? | 23:02 |
| kota_ | chsc: looks like just you and me :) | 23:03 |
| chsc | Yes. | 23:03 |
| chsc | I only have a small item: https://review.openstack.org/#/c/279234/ | 23:03 |
| patchbot | chsc: patch 279234 - swift3 - Accept HTTP_ACCEPTED 202 from proxy server | 23:03 |
| kota_ | alright, looking | 23:04 |
| chsc | We have seen that swift core can return 202 for a PUT, and swift3 currently turns that into a 500. | 23:04 |
| timburke | hello! | 23:04 |
| kota_ | timburke: hello | 23:04 |
| kota_ | chsc: oh, looks too good catch :) | 23:05 |
| chsc | cool :) | 23:06 |
| chsc | Is there anything else for me to do to get the patched integrated? | 23:06 |
| kota_ | chsc: i thought we could make a test... | 23:06 |
| kota_ | to prevent regressions | 23:07 |
| chsc | How would you setup the test? Have the 202 returned somewhere and verify that swift3 lets it go through? | 23:08 |
| kota_ | it might be hard because we have no way to make 409 confict via swift3, right? | 23:08 |
| kota_ | seems to trigger by difference between proxy server time. | 23:09 |
| kota_ | ah, | 23:09 |
| kota_ | alright, just making unit test seems enough | 23:09 |
| chsc | I don't see an easy way. Yes, time difference between machines was the problem trigger for us. | 23:09 |
| kota_ | wait a bit. | 23:10 |
| kota_ | an example is here, https://github.com/openstack/swift3/blob/master/swift3/test/unit/test_obj.py#L398-L420 | 23:11 |
| kota_ | it's a basic unit test for PUT objects. In the test, we use FakeSwift app instead of actual vanilla swift. | 23:11 |
| kota_ | so I wonder if you could try a new test method with the case swiftt returning back 202 accepted while PUT object. | 23:12 |
| kota_ | easy way to hook to make the FakeSwift to return 202 is like as https://github.com/openstack/swift3/blob/master/swift3/test/unit/test_obj.py#L425-L427 | 23:13 |
| kota_ | register the response you intend. | 23:13 |
| kota_ | chsc: thought? | 23:14 |
| chsc | That looks useful. I will try to put a test together. | 23:15 |
| kota_ | chsc: great, thanks. | 23:15 |
| chsc | That is the only item i had. | 23:15 |
| kota_ | ok, thanks. | 23:16 |
| kota_ | timburk: do you have something? | 23:16 |
| timburke | just to mention that i've been hearing more rumblings about adding v4 signature support, so i'll probably be reviewing patch 211933 soon | 23:16 |
| patchbot | timburke: https://review.openstack.org/#/c/211933/ - swift3 - support AWS signature v4 | 23:16 |
| kota_ | oh yeah, that is in agenda added by andrew | 23:17 |
| kota_ | looks like he is absent, though. | 23:17 |
| kota_ | timburke: how do you think of that? | 23:18 |
| timburke | it seems promising. i still have to get my saio set up for it, though, i think. looked like i'd need to use keystone auth? | 23:19 |
| kota_ | timburke: maybe, I don't have a deeper look too. | 23:20 |
| timburke | oh, also, i'll try to get the ranged-copy patch mergable. although, with the current movement on COPY requests upstream, it may require that we handle the GET/PUT ourselves instead of relying on COPY | 23:20 |
| kota_ | that's why, I asked for Andrew to make functional to make us eas to review. | 23:21 |
| kota_ | but | 23:21 |
| kota_ | https://review.openstack.org/#/c/272516 | 23:21 |
| kota_ | everything broken :/ | 23:21 |
| timburke | yeah, i remembered seeing that patch a while ago but hadn't really looked at it yet. checking it out now... | 23:22 |
| kota_ | timburke: alright, will see also COPY stuff. | 23:23 |
| kota_ | in my memory, he told us we need later version of keystone to enable v4 but no dependecy forced in the patch so no way we can know the constraint in upstream. | 23:24 |
| kota_ | s/keystone/keystone and keystone_middleware/ ??? not sure. | 23:25 |
| kota_ | hmm... it seems we need more conversation with Andrey to progress v4 auth | 23:27 |
| kota_ | I think we don't want to break any exisiting stuff for just adding a new authentication mechanism. | 23:27 |
| timburke | on the plus side, only one test seemed to break across those gates... | 23:30 |
| kota_ | hmmm looks related to the v4 thing. | 23:32 |
| kota_ | Location seems to be tied up to the region and v4 using the region info in the auth. | 23:33 |
| kota_ | not sure, need more deeper eye though. | 23:33 |
| timburke | hmm...and yeah, those signatures on the failed test are still v2... definitely needs a more thorough review | 23:34 |
| kota_ | to be honest, I don't like to have many tox env though :/ he added v4xkeystone env into tox. | 23:34 |
| timburke | while i'm thinking of it, we may want to put keystonemiddleware in setup.cfg as an "extra"...then we could at least make a version recommendation (if not actually require it) | 23:34 |
| kota_ | timburke: sounds good idea | 23:35 |
| timburke | it's pretty easy to do; see patch 254346 | 23:36 |
| patchbot | timburke: https://review.openstack.org/#/c/254346/ - python-swiftclient - Move python-keystoneclient to "extras" | 23:36 |
| kota_ | looks suitable. to catch up quickly, what means for "extra" option? | 23:37 |
| kota_ | extras, i mean | 23:38 |
| timburke | i agree we don't want too many tox envs, though. maybe there's a way we could have it switch between v2 and v4 signatures based on an environment variable? we'd still need to set up gate jobs, though | 23:38 |
| timburke | so, you're used to doing `pip install swift3`, right? it will install swift3 and everything in requirements.txt (and their requirements, etc) | 23:39 |
| timburke | if you do `pip install swift3[extra-name]` it will still do that, plus look up extra-name for a lit of additional packages to install | 23:40 |
| kota_ | timburke: might be. Or if we can run both v2, v4 at the same time, running up whole tests via both auth likely making v4 test case inherits v2 test case (the difference is just auth in self.connection) | 23:40 |
| timburke | i think in deb packages it might be comparable to a Recommends? something like that. | 23:41 |
| kota_ | timburke: gotcha, sounds reasonable. | 23:41 |
| timburke | see https://pythonhosted.org/setuptools/setuptools.html#declaring-extras-optional-features-with-their-own-dependencies | 23:42 |
| kota_ | one thing i am still wondering what's happen if the extras is insuffisient at the environment, though. | 23:42 |
| kota_ | will take a look. | 23:42 |
| kota_ | ok, let's move onto last part of today's meeting. | 23:46 |
| kota_ | I didn't add anything but I would like to here opinions for | 23:46 |
| kota_ | swift3 may back to swift core repo. | 23:46 |
| kota_ | that item addressed in the hackathon next week. | 23:47 |
| kota_ | not yet determined anything, just I wan to hear. | 23:47 |
| timburke | my only concern is maintaining support for old clusters. but honestly, i don't mind having an extra reason for customers to upgrade, either | 23:49 |
| timburke | well, that, and managing the functests, since the style is so different between swift and swift3 | 23:50 |
| timburke | (not saying one style is better or worse. just...different) | 23:50 |
| kota_ | sure thing, we need continuous conversation, "how we can land into core repo" | 23:50 |
| kota_ | abosolutely we need a lot of changes (e.g. namesapaces, tests, whatever) if moving to core. | 23:51 |
| kota_ | alright, not so bad feeling but we have concerns, that fits with my thought. | 23:52 |
| kota_ | thanks | 23:52 |
| timburke | and i expect that the larger swift community will have some concerns as well. but we'll certainly start a discussion in Bristol | 23:53 |
| *** chsc has quit IRC | 23:53 | |
| kota_ | timburke: :) | 23:54 |
| kota_ | anything else for today/ | 23:54 |
| kota_ | ? | 23:54 |
| kota_ | looks nothing. | 23:55 |
| kota_ | finish up, thanks for working swift3. | 23:56 |
| timburke | thanks kota_! | 23:56 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!