| timburke | bah, my repro didn't pan out -- ratelimit's get_container_info call still has a PATH_INFO like /bucket/obj so the proxy-server 400s | 00:12 |
|---|---|---|
| timburke | (plus, it thinks that it wants to do account-level ratelimiting...) | 00:13 |
| timburke | so while i still think my broader concern is warranted (do we have middlewares that *do* drop the sysmeta when translating between sysmeta and client headers?), that bug seems very particular to S3Request.get_container_info rather than proxy.controllers.base.get_container_info | 00:22 |
| timburke | there's https://github.com/openstack/swift/blob/2.19.0/swift/common/middleware/symlink.py#L249-L264 but that's objects... | 00:30 |
| * timburke shrugs | 00:31 | |
| DHE | roles in keystone are a little confusing. can I just make a new role, add it to proxy-server.conf as "operator_roles = <newrole>, <existing-values>" and that user can use swift but not any other openstack stuff? (ie make virtual machines, etc) | 00:40 |
| *** gyee has quit IRC | 00:51 | |
| *** two_tired2 has joined #openstack-swift | 00:55 | |
| DHE | aside: liberasurecode_rs_vand in 10+3 configuration on an older 3.6 GHz CPU (i7-4790) in theory can hit 500 megabits of PUT performance per core... actual throughput was ~100 megabits at ~19% CPU usage | 01:12 |
| mattoliverau | DHE yup (re:keystone). You can have a list of roles in keystone that match to operator roles. Then any user in that role can access swift, and so long as you make sure members of that role don't have access to other openstack things it shold work as expected. | 01:25 |
| DHE | "other openstack things" is the tricky bit. my nova system doesn't even have a policy.json file. or rather it's just "{ }".... :/ | 01:26 |
| DHE | and reading it is weird... | 01:26 |
| DHE | I'll roll with this and see how it goes... | 01:29 |
| *** itlinux has joined #openstack-swift | 01:44 | |
| *** itlinux has quit IRC | 01:44 | |
| *** psachin has joined #openstack-swift | 03:30 | |
| *** mikecmpbll has quit IRC | 03:41 | |
| *** two_tired2 has quit IRC | 04:11 | |
| *** itlinux has joined #openstack-swift | 05:12 | |
| *** e0ne has joined #openstack-swift | 05:52 | |
| *** pcaruana has joined #openstack-swift | 06:11 | |
| kota_ | hello world | 06:16 |
| *** e0ne has quit IRC | 06:24 | |
| mattoliverau | kota_: o/ | 06:27 |
| *** e0ne has joined #openstack-swift | 06:52 | |
| *** rcernin has quit IRC | 07:12 | |
| *** gkadam has joined #openstack-swift | 07:23 | |
| *** mikecmpbll has joined #openstack-swift | 08:04 | |
| openstackgerrit | Merged openstack/swift master: Use latest eventlet in probe tests https://review.openstack.org/602526 | 08:24 |
| *** guimaluf has quit IRC | 09:44 | |
| *** e0ne has quit IRC | 10:16 | |
| *** e0ne has joined #openstack-swift | 10:25 | |
| *** psachin has quit IRC | 13:13 | |
| *** two_tired2 has joined #openstack-swift | 13:57 | |
| openstackgerrit | Thiago da Silva proposed openstack/liberasurecode master: Install ISA-L lib to be tested w/ liberasurecode https://review.openstack.org/604391 | 14:24 |
| *** two_tired2 has quit IRC | 14:35 | |
| *** e0ne has quit IRC | 14:58 | |
| *** e0ne has joined #openstack-swift | 15:02 | |
| notmyname | good morning | 16:08 |
| notmyname | timburke: do you know how to check what verion of python was running in a set of tests? | 16:18 |
| notmyname | eg http://logs.openstack.org/62/605862/2/check/swiftclient-functional-py2/b42c7e9/job-output.txt.gz is supposed to be py2 and http://logs.openstack.org/62/605862/2/check/swiftclient-functional/837299b/job-output.txt.gz is supposed to be py3. but I can't really find anything in there that actually says that | 16:18 |
| *** gyee has joined #openstack-swift | 16:24 | |
| notmyname | maybe we need to add a `python -V` to the start of all our tests | 16:35 |
| notmyname | both of those log files above are from https://review.openstack.org/#/c/605862/ | 16:36 |
| patchbot | patch 605862 - python-swiftclient - py2 functional testing - 2 patch sets | 16:36 |
| *** ianychoi has quit IRC | 16:39 | |
| *** mikecmpbll has quit IRC | 16:41 | |
| *** gkadam has quit IRC | 16:42 | |
| *** ianychoi has joined #openstack-swift | 16:44 | |
| tdasilva | notmyname: it's weird that all tests are being skipped, not sure why... | 16:53 |
| notmyname | tdasilva: turns out, these jobs run the tests twice! scroll up and you'll see another with all tests running and passing | 16:53 |
| notmyname | timburke was saying this is to test keystone vs not-keystone | 16:53 |
| tdasilva | heh, ok: i just the skipped in the logs then saw this: http://logs.openstack.org/62/605862/2/check/swiftclient-functional/837299b/testr_results.html.gz | 16:54 |
| notmyname | yep. I did exactly the same thing yesterday--sgo to the bottom, scroll up until you see results, and "wat?!" | 16:54 |
| notmyname | I was intentionally leaving that weirdness alone for now. I first want to get whatever we have today running in py2 and py3. then we can go back in and make sure the right things are being tested | 16:55 |
| notmyname | the state of func tests in swiftclient are pretty sad. I don't want to try to fix all of that right now. that will take a long time | 16:55 |
| timburke | hmm... more functional test flakiness when run in resource-starved VMs? http://logs.openstack.org/22/580122/2/check/swift-multinode-rolling-upgrade/9334639/job-output.txt.gz#_2018-09-28_04_50_28_836537 | 17:05 |
| *** psachin has joined #openstack-swift | 17:09 | |
| timburke | notmyname: tdasilva: you can indirectly determine python version from https://github.com/openstack/python-swiftclient/blob/master/requirements.txt#L1 -- look for whether futures shows up in the tox env | 17:14 |
| timburke | compare http://logs.openstack.org/62/605862/2/check/swiftclient-functional-py2/b42c7e9/job-output.txt.gz#_2018-09-28_01_23_29_134661 and http://logs.openstack.org/62/605862/2/check/swiftclient-functional/837299b/job-output.txt.gz#_2018-09-28_01_24_44_576769 | 17:14 |
| *** ejat has quit IRC | 17:29 | |
| notmyname | timburke: thanks. *sigh* that's a terrible heuristic | 17:42 |
| timburke | i wholeheartedly agree | 17:42 |
| *** e0ne has quit IRC | 17:44 | |
| notmyname | timburke: however, the good news is that the py2 functional tests are actually using py2! yay! | 17:45 |
| *** mikecmpbll has joined #openstack-swift | 17:55 | |
| *** mahatic has quit IRC | 17:56 | |
| *** e0ne has joined #openstack-swift | 18:02 | |
| *** mahatic has joined #openstack-swift | 18:04 | |
| *** ChanServ sets mode: +v mahatic | 18:04 | |
| *** e0ne has quit IRC | 18:05 | |
| *** e0ne has joined #openstack-swift | 18:25 | |
| notmyname | did anyone see the policy naming mailing list thread? | 18:48 |
| *** e0ne has quit IRC | 18:50 | |
| timburke | huh. http://logs.openstack.org/91/604391/3/check/liberasurecode-unittests/13ef1c8/job-output.txt.gz#_2018-09-28_16_51_31_916452 | 18:53 |
| timburke | "the `impossible' happened" -- sounds ominous | 18:53 |
| notmyname | lol | 18:53 |
| notmyname | the policy stuff makes me want to remember all hisashi's work on the RBAC stuff | 18:53 |
| timburke | apparently we hit https://github.com/lu-zero/vex/blob/3c7650b/priv/ir_opt.c#L1223 ? | 18:55 |
| notmyname | I wonder what the Ico_ prefix on those types means | 18:56 |
| timburke | we merged some of that RBAC stuff... looks like https://review.openstack.org/#/c/212825/ and https://review.openstack.org/#/c/253371/ are still open | 18:59 |
| patchbot | patch 212825 - swift - Add functional test for access control (COPY) with... - 10 patch sets | 18:59 |
| patchbot | patch 253371 - swift - Improves RBAC related functional test - 8 patch sets | 18:59 |
| notmyname | https://review.openstack.org/#/q/project:openstack/swift+branch:master+topic:add-functional-test | 18:59 |
| notmyname | but I'm trying to remember how it works with keyston/oslo RBAC at all. like where are these defined? | 19:00 |
| timburke | yeah, i'm trying to remember, too. i feel like there was something ... oh! here! https://review.openstack.org/#/c/149930/ | 19:01 |
| patchbot | patch 149930 - swift - WIP: Enable Role-based access control using oslo.p... (ABANDONED) - 15 patch sets | 19:01 |
| notmyname | yeah. I'd love to have the policy.jsob stuff enabled for any auth system and avoid the need to bring in oslo config. it would give operators quite a bit of control, especially when they back auth with ldap or ad | 19:07 |
| *** e0ne has joined #openstack-swift | 19:29 | |
| *** e0ne has quit IRC | 19:34 | |
| openstackgerrit | Fatema Khalid Sherif proposed openstack/swift master: Enabling direct_client users to overwrite X-Timestamp https://review.openstack.org/603261 | 19:38 |
| *** openstackgerrit has quit IRC | 20:07 | |
| *** openstackgerrit has joined #openstack-swift | 20:08 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: Ignore ENOENT and ENOTEMPTY errors in delete_partition https://review.openstack.org/606192 | 20:08 |
| *** psachin has quit IRC | 20:10 | |
| *** geaaru has joined #openstack-swift | 21:41 | |
| *** paladox is now known as paladox[EU_UK] | 22:07 | |
| *** paladox[EU_UK] is now known as paladox | 22:08 | |
| *** paladox is now known as paladox_UK_EU | 22:09 | |
| *** paladox_UK_EU is now known as paladox[UK_EU] | 22:09 | |
| *** paladox[UK_EU] is now known as paladox | 22:10 | |
| openstackgerrit | Merged openstack/python-swiftclient master: py2 functional testing https://review.openstack.org/605862 | 22:41 |
| *** fatema_ has joined #openstack-swift | 22:47 | |
| fatema_ | Good morning | 22:47 |
| timburke | a very early morning for you fatema_, no? | 22:54 |
| openstackgerrit | Merged openstack/swift master: Unquote URL before using splited parts. https://review.openstack.org/580122 | 22:55 |
| fatema_ | yes, 1 A.M. :D | 22:58 |
| fatema_ | but no lectures tomorrow so it's fine | 22:58 |
| fatema_ | timburke, I just had some questions but left them in comments. | 22:58 |
| fatema_ | I wanted to know another thing about the bug https://bugs.launchpad.net/swift/+bug/1757250 | 22:59 |
| openstack | Launchpad bug 1757250 in OpenStack Object Storage (swift) "direct_client gen_headers is inconsistent" [Wishlist,New] | 22:59 |
| fatema_ | It mentions the need to *make sure all direct_client methods accept the headers kwarg* | 23:00 |
| openstackgerrit | Tim Burke proposed openstack/swift master: Tolerate missing containers when trying to clean up https://review.openstack.org/606217 | 23:03 |
| timburke | hurray for weekends! | 23:03 |
| timburke | i'll take a look at comments next, but yeah -- it'd be nice if callers could pass headers for all api calls | 23:03 |
| timburke | i think it's just the two functions that need it | 23:04 |
| fatema_ | timburke, are callers intended to pass headers through kwargs ?? | 23:08 |
| fatema_ | I see that most of the functions in direct_client don't have kwargs | 23:09 |
| timburke | you usually see signatures like https://github.com/openstack/swift/blob/2.19.0/swift/common/direct_client.py#L214-L215 or https://github.com/openstack/swift/blob/2.19.0/swift/common/direct_client.py#L248-L251 -- the bug is just that https://github.com/openstack/swift/blob/2.19.0/swift/common/direct_client.py#L224-L225 (for example) doesn't accept a headers arg | 23:11 |
| timburke | the caller might provide it as a keyword arg like direct_head_container(..., headers={'x-blah-blah': 'blah'}) or positionally like direct_head_container(..., {'x-blah-blah': 'blah'}) -- i'm not so concerned about that part | 23:13 |
| fatema_ | ok I see now how it could be passed. why isn't this part a concern ? | 23:18 |
| timburke | because i'm fairly certain that the majority of callers are inside the swift repo ;-) | 23:22 |
| *** fatema__ has joined #openstack-swift | 23:22 | |
| *** fatema_ has quit IRC | 23:23 | |
| fatema__ | Hmm.... | 23:24 |
| fatema__ | You know better of course. But I think I would do it to close the bug | 23:25 |
| timburke | for the few that aren't, as long as we append to the end of the argument list, we won't break existing users. though i wouldn't be surprised if we've rearranged args before -- again, this doesn't seem highly-used and where we *really* care about api compat is in things like proxy-server responses | 23:25 |
| timburke | OH! i think i get the question now :-) i'm pretty sure clayg just meant that there should be a headers arg, and it should be named "headers" to be consistent with the other functions, not that we should have everything start accepting **kwargs and ignoring anything we don't recognize | 23:27 |
| *** gyee has quit IRC | 23:33 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!