Earlier  
Later
New Search
Posted Nick Remark
#openstack-sdks - 2019-08-08
11:33:57 frickler mordred: o.k., 0.31.0 returned None there, too, so that's o.k., then
11:37:53 mordred cool
11:39:21 frickler mordred: but I'd actually tend to agree with Thierry on that the release might better be 0.33.0 than 0.32.1
11:44:23 stingrayza frickler, mordred: fantastic! thanks gents :D
11:44:40 stingrayza at least I know I wasn't going mad ;D
11:45:05 stingrayza ^ my default position when I find a bug.
11:49:26 openstackgerrit Brin Zhang proposed openstack/python-openstackclient master: WIP: Add server migration list CLI https://review.opendev.org/675304
11:50:29 openstackgerrit Maxim Babushkin proposed openstack/openstacksdk master: Add 'tag' support to compute with supported microversion https://review.opendev.org/675306
11:52:41 openstackgerrit Maxim Babushkin proposed openstack/openstacksdk master: Add 'tag' support to compute with supported microversion https://review.opendev.org/675306
11:53:25 openstackgerrit Loan Harrouin proposed openstack/openstacksdk master: Add get_actions for server resource https://review.opendev.org/675113
11:58:52 mordred frickler: yeah - agree re: 0.33
12:29:30 mordred kmalloc: whence you are awakened, I have a question for you on keystone v2 admin interface things
12:30:38 mordred kmalloc: using the token fixture from ksa I've got a catalog in the sdk tests now that uses differnt hostnames for admin and public endpoints - this is flushing out some places where things were off but we weren't noticing because a requests-mock uri registered for either would work (same hostname)
12:30:48 mordred in a few places I'm getting: No mock address: GET https://identity.example.com/v2.0/OS-KSADM/roles
12:31:03 mordred I would have thought that OS-KSADM would really only ever be on the admin interface
12:31:24 mordred but that's telling me the code is making a request to OS-KSADM on the public interface
12:31:31 mordred is the code wrong? or the test wrong?
12:31:38 mordred cmurphy: ^^ in case you know
12:34:18 mordred OR - should I put a pin in it for later since at this point keystonev2 is so old it makes us all want to die inside
13:00:01 openstackgerrit Andrey Kurilin proposed openstack/osc-lib master: Fix usage of NotFound exception https://review.opendev.org/675328
13:05:37 mordred kmalloc:, cmurphy: I think the right answer for now is to just skip it for now - I don't think it's worth the effort to cleanup keystonev2 unless someone actually complains
13:25:33 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: WIP Replace catalog-v3.json with keystoneauth fixture https://review.opendev.org/675187
13:25:40 dtantsur mordred: approved
13:25:44 mordred dtantsur: woot! thanks
14:20:30 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Replace catalog-v3.json with keystoneauth fixture https://review.opendev.org/675187
14:35:19 mordred Shrews, dtantsur, kmalloc: ^^ that one is ready for review - although it needs a ksa release so it can't be landed and the tests won't go totally green until that one exists
14:42:26 openstackgerrit Andrey Kurilin proposed openstack/osc-lib master: Fix usage of NotFound exception https://review.opendev.org/675328
15:01:17 openstackgerrit Merged openstack/openstacksdk master: Fix dns return values https://review.opendev.org/675298
15:10:20 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Remove container-infra.example.com hack https://review.opendev.org/675378
15:26:54 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: WIP Remove the block-storage catalog hack https://review.opendev.org/675382
15:30:40 kmalloc mordred: o/
15:32:30 kmalloc mordred: so, in v3 admin/public are the same. really the only thing anything knows is it should talk to keystone. the split between admin and public was really a "do I get a 404 or not", and people could run both on the same endpoint in reality if thye wanted (some did even in v2)
15:32:54 kmalloc mordred: soooo possibly the test or fixture are subtly wrong.
15:33:25 kmalloc and OS-XXXX extensions were weird.
15:33:29 mordred kmalloc: well - for v3 we're fine - it's all v2 cases where we're messed up
15:33:42 kmalloc mordred: if someone complains about v2 we can clean it up.
15:33:45 mordred (for v3 nothing is coded to request admin endpoints)
15:33:47 mordred cool
15:33:51 mordred yeah - that's where I got to
15:34:16 mordred kmalloc: in fact- what if I don't even have admin endpoints for anything in the v3 catalog?
15:34:17 kmalloc otherwise i would let v2 just kindof sit. my guess is we will rapidly get to a point where no one will be using v2 even with SDK.
15:34:40 mordred (so that we catch places where people are doing something that they probably shouldnt be)
15:34:46 kmalloc someone was working on fixing that in the general case (no admin endpoints in gate in v3) recently
15:34:50 kmalloc i am not sure what landed.
15:35:08 kmalloc but some things were still a little broken. i support going to no admin endpoints in v3
15:35:14 kmalloc esp. in SDK testing.
15:35:24 kmalloc no explicit admin endpoints*
15:35:44 mordred ++
16:00:00 edleafe What's that you say? The API-SIG Office Hour is upon us?
16:00:21 elmiko \o/
16:57:51 edleafe Guess all that lurking wore dtantsur|afk out!
16:58:19 elmiko haha, yeah!
16:58:34 elmiko we can blame it on the /other/ meeting XD
17:00:47 elmiko i won't be around next week, i'm on conference duty ;)
17:00:54 elmiko have a good weekend edleafe o/
17:01:03 edleafe you too, elmiko!
18:49:23 openstackgerrit David Shrewsbury proposed openstack/openstacksdk master: WIP: Retry SLO manifest upload https://review.opendev.org/675134
18:50:24 frickler mordred: kmalloc: for what happens with no admin endpoint see https://review.opendev.org/651492 , heat blows up when using the OS-TRUST interface, so that might be related to your v2 issue
18:51:06 frickler anyway help with fixing that remaining issue would be greatly appreciated, I dug through the tracebacks and didn't find anything helpful
19:00:21 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: WIP: Retry SLO manifest upload https://review.opendev.org/675134
19:21:16 Shrews mordred: omg, i got the test working
19:23:07 Shrews mordred: i'm assuming we want the same retry logic on dlo object manifests, yeah? Would it also need to delete the segments?
19:44:35 openstackgerrit David Shrewsbury proposed openstack/openstacksdk master: Retry large object manifest upload https://review.opendev.org/675134
19:45:16 Shrews mordred: I just realized that we have the same problem if any of the segment uploads fail. :/ So no limited to manifest failure.
19:45:32 Shrews So not*
20:27:32 openstackgerrit Adrian Turjak proposed openstack/keystoneauth master: add support for auth_receipts and multi-method auth https://review.opendev.org/675049
20:43:35 mordred Shrews: this is quite the big beastie
20:43:45 mordred Shrews: but yay!
20:44:40 mordred Shrews: I believe with DLO objects we just pass the "delete your manifests" parameter and deleting the manifest object will cause the segment objects to be deleted
20:45:42 Shrews mordred: but that requires actually uploading the manifest, yeah?
20:46:16 mordred oh - joy. yes
20:46:21 mordred Shrews: you have the smart brain
20:46:21 openstackgerrit Adrian Turjak proposed openstack/keystoneauth master: add support for auth_receipts and multi-method auth https://review.opendev.org/675049
20:46:25 Shrews can't delete it if you can't upload it :-P
20:53:10 efried mordred: I loaded up your patch from yesterday and was able to get nova talking to cyborg just fine.
20:53:23 efried I've got the setup running in a public cloud, do you want to poke at anything?
20:53:23 mordred efried: woot
21:02:47 efried mordred: ...but afaict it doesn't look any different than the ksa version.
21:03:24 mordred efried: I think as long as it worked like you expected that's good enough for me
21:04:33 mordred Shrews: if you have a sec in between bouts of yelling at object-store apis ... https://review.opendev.org/#/c/675135/ is easy-ish
21:05:36 mordred Shrews: (and this is the patch efried is mentioning that he tested in the scrollback just above that makes cyborg and placement integration in nova work better)
21:06:04 efried mordred: We expected that to get rid of a bunch of warnings to placement as well, let me see if I can verify that...
21:06:39 mordred efried: if it doesn't, something is horribly wrong
21:27:59 efried mordred: Okay, I confirmed. Went back through the logs before I applied your patch and found those UnsupportedServiceVersion warnings, and they then went away once the patch was applied.
21:28:35 mordred efried: woot!
21:28:48 mordred senseless warnings in logs are like fungus
21:28:56 mordred you gotta get rid of them before the spread
21:29:40 efried left a +1 accordingly.
21:29:55 efried mordred: though a guy wonders if there should be some unit test for this....
21:30:23 mordred efried: some days I feel like you and Shrews are the same person what with your "shouldn't there be a test for this" ;)
21:31:05 efried I can think of worse company to be in.
21:31:29 efried I mean, I'm happy having tested this live, BUT I only tested it with a nice clean default devstack service catalog and conf
21:31:31 mordred efried: oh - I don't know if you saw - patch one on on the path to getting you a reusable fixture is actually working now
21:31:43 efried nice
21:31:58 mordred efried: people use openstack in some configuration that isn't a nice clean default devstack?
21:32:20 efried ...I didn't put it through the paces of endpoint_override, weird service catalog entries, non-default interfaces, weird version requests....
21:35:00 mordred yeah. well - so - the nice thing here (and we should do all of those things) - but the nice thing about this patch is it basically says "if we don't know anything, just defer 100% to the existing and already tested elsewhere adapter creation logic"
21:35:26 mordred so all the other combos *should* be covered already
21:35:43 mordred but that's me rationalizing a scenario that clearly needs some coverage :)
21:47:48 Shrews mordred: i insist on a test for the test as well

Earlier   Later