| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-02-14 | |||
| 17:45:23 | mriedem | BobBall: when can we expect multinode xenserver ci for live migration testing? | |
| 17:45:32 | mriedem | virtuozzo already beat you guys to that | |
| 17:48:24 | BobBall | Oh shocking! multinode is proving to be tricky because of the dodgyness of the nodes we are using (Ugly details are XenServer in VM, with devstack running in a nested VM and fudging the MAC/IP so nodepool talks directly to devstack...). Current focus is on upgrading to Xenial, but then we will switch back to trying to get multinode working. | |
| 17:49:23 | BobBall | Can't promise timings, but I would definitely expect multinode working during Pike, probably by P-1 but I'd be happier to be held to P-2 ;) | |
| 17:49:50 | mriedem | i won't hold you to it, but it makes getting stuff like this through faster | |
| 17:49:54 | mriedem | when it's actually tested | |
| 17:50:41 | mriedem | sdague: can you hit these backports for the cells v1 failure? https://review.openstack.org/#/q/Id7e8351026ca86694fc9814002b96260f6ba10f9,n,z | |
| 17:50:55 | BobBall | Indeed... I want to make sure that we don't regress again because it's plain embarassing. | |
| 17:51:43 | sdague | mriedem: yeh, I'm just going to fast approve them as we know it's a gate break | |
| 17:51:55 | mriedem | yeah | |
| 17:52:19 | lbragstad | mriedem are you planning on having a specific etherpad for your rbac/policy discussions? | |
| 17:52:33 | mtreinish | sdague: here's your backwards incompat fix: https://github.com/mtreinish/stestr/commit/97a52706c92cedd033484009e7ce409a0679ba41 | |
| 17:52:49 | mriedem | lbragstad: i'm not privy to any of that | |
| 17:53:01 | mriedem | johnthetubaguy: are we having a big rbac/policy pow-wow at the PTG? | |
| 17:53:07 | sdague | mtreinish: ++ | |
| 17:53:27 | johnthetubaguy | mriedem: I think lbragstad was looking at scheduling something | |
| 17:53:59 | lbragstad | johnthetubaguy mriedem i have policy/rbac penciled in for wednesday afternoon (but that's open to change if needed) | |
| 17:54:30 | mriedem | lbragstad: we were going to do resource providers/placement on wed afternoon, and the ironic team was going to be in on that too | |
| 17:54:47 | lbragstad | mriedem makes sense - i can work things around | |
| 17:54:53 | mriedem | thursday afternoon or friday morning probably works better for us | |
| 17:55:11 | mriedem | can you throw something on https://etherpad.openstack.org/p/nova-ptg-pike ? | |
| 17:55:16 | lbragstad | mriedem ok - i'll rework some things | |
| 18:09:51 | cfriesen | what is the distinction between doc/api_samples/servers/v2.X and nova/tests/functional/api_sample_tests/api_samples/servers/v2.X ? are there instructions somewhere about adding functional tests? | |
| 18:10:59 | sdague | cfriesen: the api samples tests include templates | |
| 18:11:02 | mriedem | cfriesen: the api-samples functional tests use the doc api samples as sample input data | |
| 18:11:34 | sdague | and when run there is a 3 way comparison between - returned results from the server, templates filled in, and static documentation that's included in our api-ref | |
| 18:11:51 | mriedem | i believe you're supposed to write the functional test first with the template, and then run tox -r -e api-samples to generate the doc api samples | |
| 18:11:54 | sdague | which ensures the documentation is up to date with the code | |
| 18:12:59 | cfriesen | sdague: mriedem: thanks. will give that a try. | |
| 18:17:12 | dansmith | sdague: have there ever been plans to make the functional tests start services (such as nova-compute) as processes somehow? I'm not sure how the fake rpc driver works and that may sever our connectivity.. | |
| 18:17:36 | dansmith | sdague: asking because RSN we're going to start bumping up against problems switching global db state in the cells fixtures | |
| 18:17:46 | dansmith | and by RSN I mean.. in stuff I was working on yesterday | |
| 18:21:13 | dansmith | mriedem: melwitt, bauzas, and I all have a conflict for tomorrow's cells meeting.. do you want to run it or cancel? | |
| 18:21:53 | mriedem | at 1PST? | |
| 18:21:57 | mriedem | i have a conflict too | |
| 18:22:08 | dansmith | no, tomorrow it's 9PST | |
| 18:22:34 | mriedem | i have a conflict then too :) | |
| 18:22:37 | mriedem | 4 meetings tomorrow | |
| 18:22:48 | dansmith | okay then | |
| 18:24:25 | sdague | dansmith: there have not ever been those plans | |
| 18:24:44 | sdague | and I would not expect the fake_rpc driver to work in that case, because it's just shared memory state iirc | |
| 18:40:54 | mriedem | test_stamp_pattern is also breaking the ceph job on mitaka :( | |
| 18:48:18 | openstackgerrit | Vivek Agrawal proposed openstack/nova-specs master: Spec for Veritas HyperScale Nova filter https://review.openstack.org/433855 | |
| 18:55:31 | mriedem | gah | |
| 18:55:32 | mriedem | so, | |
| 18:55:54 | mriedem | apparently if you have nova/instances lying around in mitaka, the libvirt unit tests hang trying to execute an unmocked chown | |
| 18:56:16 | cburgess | cute | |
| 18:56:28 | mriedem | i'm glad i wasted the last hour figuring that out | |
| 19:00:41 | openstackgerrit | Vivek Agrawal proposed openstack/nova-specs master: Spec for Veritas HyperScale Nova filter https://review.openstack.org/433855 | |
| 19:04:59 | sdague | mriedem: I remember cleaning up a few of those in newton | |
| 19:05:51 | mriedem | i assume it's something in the libvirt tests that uses the $state_path to create those and never cleans up | |
| 19:06:10 | mriedem | i reported it https://bugs.launchpad.net/nova/+bug/1664669 but not actively working it | |
| 19:06:10 | openstack | Launchpad bug 1664669 in OpenStack Compute (nova) "Leftover nova/instances from test runs can make tests fail locally" [Low,Confirmed] | |
| 19:07:59 | openstackgerrit | Sean Dague proposed openstack/nova-specs master: update validate-project-with-keystone for pike https://review.openstack.org/433862 | |
| 19:08:14 | mriedem | oh sdague | |
| 19:08:21 | sdague | mriedem: there were a few fall throughs | |
| 19:08:29 | mriedem | you're supposed to copy it, not move it | |
| 19:08:34 | mriedem | plus other things | |
| 19:08:43 | sdague | ah, no worries, I'll redo it | |
| 19:09:00 | sdague | sorry, it's been a while since I did one of those moves forward | |
| 19:09:05 | mriedem | rusty | |
| 19:09:24 | openstackgerrit | Vivek Agrawal proposed openstack/nova-specs master: Spec for Veritas HyperScale Nova filter https://review.openstack.org/433855 | |
| 19:10:26 | sdague | mriedem: ... what are the "plus other things" | |
| 19:11:55 | openstackgerrit | Christopher Martin proposed openstack/nova master: Stop using mox in test_compute_cells.py https://review.openstack.org/389750 | |
| 19:13:14 | mriedem | sdague: the previously-approved thing in the commit message | |
| 19:13:25 | openstackgerrit | Sean Dague proposed openstack/nova-specs master: update validate-project-with-keystone for pike https://review.openstack.org/433862 | |
| 19:13:28 | sdague | ok | |
| 19:24:27 | mriedem | sdague: did you rewrite a bunch of stuff in that spec? the diff from ocata is pretty bi | |
| 19:24:28 | mriedem | *big | |
| 19:24:56 | mriedem | or a mass line wrap? | |
| 19:25:50 | sdague | mriedem: I tweaked the line wrap | |
| 19:26:46 | mriedem | ok approved, along with the bp | |
| 19:26:47 | mriedem | thanks | |
| 19:40:13 | sdague | cool, I'll see about digging in to that in the morning | |
| 19:42:09 | dansmith | sdague: yeah, I expect you're right about fake rpc.. I'm not really sure what to do as I think we're going to struggle with the amount of global db state switching we need to do here | |
| 19:42:10 | dansmith | but anyway | |
| 19:43:00 | sdague | dansmith: is there an oslo.db fix that would make that less painful? | |
| 19:43:17 | sdague | I guess I don't fully understand the issue with the global db state switching | |
| 19:43:18 | dansmith | sdague: I dunno, it seems rather fundamental | |
| 19:43:45 | dansmith | sdague: the mere definition of db-accessing routines through the enginefacade stuff maintains global state in the decorators | |
| 19:43:53 | sdague | ug | |
| 19:44:11 | dansmith | it's kinda the entire design | |
| 19:45:02 | claudiub|2 | mriedem: hi. i might have found a bug regarding pci devices. let's say i have an instance with a flavor with 2 pci devices. i cannot resize that instance to a flavor only having one pci device instead. http://paste.openstack.org/show/598884/ | |
| 19:45:34 | claudiub | just an fyi. do you know anything about it? | |
| 19:45:45 | mriedem | claudiub: have you asked moshele about that? i feel like that was fixed in newton, but might be thinking of something else | |
| 19:45:54 | claudiub | just found out. | |
| 19:45:56 | mriedem | i may be thinking of cold migrate rather than resize | |
| 19:47:06 | claudiub | hm, afaik moshele should be online around this time. will bug him more about this. :) | |
| 19:49:57 | moshele | claudiub: resize should work | |
| 19:50:36 | claudiub | moshele: well, it didn't for me. :) | |
| 19:50:57 | moshele | claudiub: did you try on the resize on the same host or it also try to move it to another hypervisor | |
| 19:51:17 | claudiub | moshele: same host resize | |
| 19:51:37 | claudiub | i only have 1 compute node | |
| 19:51:49 | moshele | claudiub: and you allow to resize on the same host? | |
| 19:52:32 | moshele | claudiub: you will need at least 3 pci device for this to work (2 for the old guest and 1 for the new guest) | |
| 19:52:53 | claudiub | moshele: yep, allow_resize_to_same_host = True | |
| 19:53:24 | claudiub | moshele: why, i basically want to detach one of the PCI devices. | |
| 19:53:32 | claudiub | and I only have 2 | |
| 19:53:43 | claudiub | I should be able to detach it. | |
| 19:55:06 | moshele | claudiub: becuase of this https://github.com/openstack/nova/blob/master/nova/compute/resource_tracker.py#L244-L247 | |