| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-10-09 | |||
| 18:48:01 | sdague | melwitt: is there an appropriate plugin phase for the ceph plugin to do it? | |
| 18:48:05 | mriedem | melwitt: so this is progress right? http://logs.openstack.org/63/509663/4/experimental/gate-tempest-dsvm-py35-full-devstack-plugin-ceph-ubuntu-xenial-nv/2b03b3a/logs/devstacklog.txt.gz#_2017-10-06_05_26_37_113 | |
| 18:48:11 | mriedem | sdague: apparently not | |
| 18:48:17 | mriedem | keystone is special as it's loaded early in stack.sh | |
| 18:48:35 | sdague | where is the ceph plugin source? | |
| 18:48:37 | melwitt | sdague: there used to be post-config. but now keystone is started before post-config and keystone start is the thing that's failing in the ceph job. so I need a new phase between pip install and keystone start | |
| 18:48:48 | mriedem | sdague: https://github.com/openstack/devstack-plugin-ceph | |
| 18:50:05 | melwitt | mriedem: that's the py35 job which won't work until the ceph plugin is installing python3-rados and python3-rbd. I have a patch up for that too | |
| 18:50:16 | melwitt | mriedem: but you want to take a look at the py27 job, it passes with that hack | |
| 18:50:16 | mriedem | oh right | |
| 18:50:31 | melwitt | I got confused by the same thing last week | |
| 18:50:45 | mriedem | i know i reported a bug for ceph libradoes in glance-store in lp somewhere | |
| 18:50:50 | mriedem | for py3 | |
| 18:51:16 | mriedem | ah here it is passing http://logs.openstack.org/63/509663/4/experimental/gate-tempest-dsvm-full-devstack-plugin-ceph-ubuntu-xenial/af8e25f/ | |
| 18:51:18 | melwitt | ah yup, it's that. glance-store weirdly sort of silently ignores import errors even in the rbd file | |
| 18:51:24 | sdague | melwitt: why can't it run at stack_install ? | |
| 18:51:26 | mriedem | melwitt: yup | |
| 18:51:51 | melwitt | sdague: is stack_install after pip install and before keystone start? if it is, then it can run there | |
| 18:52:33 | sdague | https://github.com/openstack-dev/devstack/blob/a29329b55983d69c3966e1699f985be8cdf6e59d/stack.sh#L914 | |
| 18:52:49 | mriedem | ah yup | |
| 18:52:53 | sdague | it's definitely before keystone start | |
| 18:53:11 | mriedem | so need something like this https://github.com/openstack/devstack-plugin-ceph/blob/master/devstack/plugin.sh#L6 | |
| 18:53:17 | mriedem | elif [[ "$1" == "stack" && "$2" == "install" ]]; then | |
| 18:53:23 | mriedem | good call | |
| 18:53:32 | melwitt | omg, thanks sdague | |
| 18:53:47 | melwitt | sweet | |
| 18:53:56 | sdague | melwitt: no prob | |
| 18:54:07 | melwitt | I need better devstack-fu | |
| 18:54:32 | sdague | no worries, it's a lot of code | |
| 18:55:09 | sdague | if no appropriate phase existed, I would have created one for it, because the whole plugin abstraction completely breaks if you have plugin names in the main code :) | |
| 18:55:35 | sdague | we've added ones in the past when we realized there was a missing place folks needed to do something | |
| 18:56:08 | sdague | I think I've got +2 on the ceph plugin repo too, so let me know when that's posted and I can help get it in | |
| 18:57:26 | melwitt | awesomesauce | |
| 19:04:34 | openstackgerrit | sean mooney proposed openstack/nova master: WIP reorder destination host setup https://review.openstack.org/510658 | |
| 19:11:01 | melwitt | sdague: posted here https://review.openstack.org/#/c/510660 will need to await test results to be 100% sure it's gonna owrk | |
| 19:11:03 | melwitt | *work | |
| 19:11:38 | sdague | melwitt: cool, I starred it, will come back around when the tests are in | |
| 19:11:50 | melwitt | thx | |
| 19:17:22 | mriedem | so uh https://review.openstack.org/#/c/394808/ | |
| 19:17:32 | mriedem | as far as i can tell, | |
| 19:17:44 | mriedem | that was added in ocata but the libvirt live migration flag it's setting is only available starting in libvirt 2.3.0 | |
| 19:17:52 | mriedem | http://libvirt.org/git/?p=libvirt.git;a=blob;f=docs/news-2016.html.in#l437 | |
| 19:18:03 | mriedem | our minimum is 1.2.9 | |
| 19:18:56 | sean-k-mooney | we can override this from the config no? | |
| 19:19:01 | mriedem | no | |
| 19:20:02 | sdague | mriedem: probably just put a conditional around that? | |
| 19:20:04 | sean-k-mooney | oh so the live_migration_flag option in the libvirt section of the nova-cpu.conf is jsut the default set ? | |
| 19:20:29 | mriedem | oh this might just be some other libvirt backend | |
| 19:21:15 | mriedem | yeah nvm, super old http://libvirt.org/git/?p=libvirt.git;a=commit;h=632be336898c9efef258e1f2d1c3df199a98a4aa | |
| 19:26:57 | sean-k-mooney | so live_migration_flag was apparently removed in kilo ... i should probably update our internal wiki to not tell you to use those when livemigrating | |
| 20:11:44 | openstackgerrit | Matt Riedemann proposed openstack/nova master: libvirt: remove old code in post_live_migration_at_destination https://review.openstack.org/510680 | |
| 20:12:45 | openstackgerrit | Merged openstack/nova master: Remove dead code of api.fault notification sending https://review.openstack.org/505164 | |
| 20:12:58 | mriedem | sean-k-mooney: having to re-learn live migration to review this port binding spec again of course | |
| 20:45:18 | claudiub | mriedem: hellou. just an fyi, and I'm going to send a mail to the MLs and a patch for it, we're deprecating the support for Windows / Hyper-V Server 2012; the acceptable versions will be 2012 r2 and 2016. | |
| 20:45:22 | mriedem | man we do not do a good job of documenting the valid values for filtering instances by status | |
| 20:45:31 | mriedem | claudiub: ok | |
| 20:55:05 | openstackgerrit | melanie witt proposed openstack/nova master: Target context for build notification in conductor https://review.openstack.org/509967 | |
| 20:55:06 | openstackgerrit | melanie witt proposed openstack/nova master: Elevate existing RequestContext to get bandwidth usage https://review.openstack.org/509968 | |
| 20:55:06 | openstackgerrit | melanie witt proposed openstack/nova master: WIP Improve the CellDatabases test fixture and usage https://review.openstack.org/508432 | |
| 20:55:07 | openstackgerrit | melanie witt proposed openstack/nova master: Fix target_cell usage for scatter_gather_cells https://review.openstack.org/510691 | |
| 20:57:25 | efried | mriedem sdague Sorry, I'm back now. Little lunchtime injury to deal with. Anything I need to know right now or will me email/dashboard tell all? | |
| 20:58:17 | melwitt | mriedem, dansmith: found a couple of cells bugs while working on CellDatabases. bottom two in the stack ^ | |
| 20:59:36 | openstackgerrit | Merged openstack/nova master: Always put 'uuid' into sort_keys for stable instance lists https://review.openstack.org/510140 | |
| 21:00:29 | mriedem | efried: the latter | |
| 21:00:41 | efried | mriedem Cool, thanks. | |
| 21:01:29 | efried | mriedem And to answer your question, the distance between those two patches was significant and we didn't want to collapse them together, though we could have. But you're absolutely correct that the second one just undoes what the first one adds wrt those auth options. | |
| 21:07:37 | mriedem | melwitt: don't tell me you can't get bandwidth usage with multiple cells?! | |
| 21:07:40 | openstackgerrit | Matt Riedemann proposed openstack/nova master: api-ref: fix server status values in GET /servers docs https://review.openstack.org/510696 | |
| 21:09:21 | dansmith | melwitt: not possible. | |
| 21:09:27 | melwitt | mriedem: lol. that one is not about a bug | |
| 21:10:19 | melwitt | dansmith: true, bugs in cells is unpossible | |
| 21:10:48 | dansmith | melwitt: I feel like we should be able to test that bottom one somehow | |
| 21:10:58 | dansmith | er, | |
| 21:11:03 | dansmith | I mean the scatter gather one | |
| 21:12:17 | melwitt | dansmith: yeah, I'm pondering on it | |
| 21:13:10 | melwitt | at first I was thinking there's not a good way to test it but there's probably a way | |
| 21:14:30 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Update "SHUTOFF" description in API guide https://review.openstack.org/510697 | |
| 21:15:16 | melwitt | just ... want ... to ... save ... comment ... in .... gerrit ... | |
| 21:15:24 | melwitt | nnnnggghh | |
| 21:15:35 | mriedem | it is pretty slow right now | |
| 21:15:55 | melwitt | yeah, some ppl are talking about it in #infra | |
| 21:16:10 | melwitt | probably gonna have to restart it again | |
| 21:16:37 | dansmith | yeah, super slow | |
| 21:16:46 | dansmith | melwitt: a couple questions in the other patch | |
| 21:18:40 | openstackgerrit | Merged openstack/nova master: use already loaded BDM in instance. |
|
| 21:21:33 | melwitt | dansmith: in the test I was trying to assert that the context for the notify was targeted to the same cell the instance was created in. though I'm not sure how we know it's cell1 (there's already an assertion in that test that knows instance is created in cell1) | |
| 21:22:55 | dansmith | I really wish we had something in the context that identified the cell, because several tests go through obscure hoops to try to verify the cell | |
| 21:24:23 | dansmith | if you're going to leave it that way I think you need a bunch more comments explaining all the assumptions, how the fixture works, etc | |
| 21:24:38 | openstackgerrit | Merged openstack/nova master: use already loaded BDM in instance. |
|
| 21:25:05 | dansmith | personally I think it'd be better to just (re-)mock target_cell so you can get a stable __enter__ result and ensure that its yielded context is used for the instance | |
| 21:25:39 | melwitt | dansmith: it's not relying on the fixture giving the same context for subsequent targets (I removed all of that in the top patch and this is me working backward fixing the bugs I found) | |
| 21:26:04 | dansmith | I mean the same oslo_db context | |
| 21:26:07 | dansmith | or whatever it's called | |
| 21:26:09 | melwitt | I just wanted to verify it's targeted to the right thing | |
| 21:26:20 | dansmith | the thing we yield from our global db facade hacky thing | |
| 21:26:24 | melwitt | yeah, I took out that stuff in the last patch | |
| 21:26:58 | melwitt | I left only default_ctxt_mgr. unless that's what you're talking about | |
| 21:27:10 | dansmith | is that how this works then? | |
| 21:27:25 | dansmith | because you're asserting that something from a freshly created context is the same as an earlier-created one | |
| 21:28:39 | melwitt | oh, I think I understand what you're saying. you're talking about how in the fixture we only create a new context manager once per cell | |