| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-04-05 | |||
| 12:03:29 | artom | yasemin_, from the command line on the controller node | |
| 12:10:02 | yasemin_ | atom, how is the command ? or can you suggest any doc ? | |
| 12:16:16 | artom | yasemin_, the command is 'nova-manage cell_v2 discover_hosts' :) | |
| 12:16:26 | frickler | yasemin_: https://docs.openstack.org/developer/nova/cells.html#setup-of-cells-v2 | |
| 12:17:13 | artom | frickler, I think his issue is that he added a compute, and the compute isn't mapped to a cell | |
| 12:18:30 | yasemin_ | artom, :) sorry | |
| 12:18:50 | yasemin_ | result = self._query(quer | |
| 12:18:50 | yasemin_ | /usr/local/lib/python2.7/dist-packages/pymysql/cursors.py:166: Warning: (3090, u"Changing sql mode 'NO_AUTO_CREATE_USER' is deprecated. It will be removed in a future release.") | |
| 12:19:03 | frickler | artom: yes, and the step you mention is correct, I was just adding the doc component for it | |
| 12:19:10 | artom | I'm pretty sure there are plans to run discover_hosts automatically, either in a periodic task, or upon manual compute node addition, but those haven't materialized yet, so it has to be run manually | |
| 12:20:19 | frickler | artom: one can enable it with this I think: https://review.openstack.org/#/c/453004/1/doc/install-guide/source/nova-compute-install.rst | |
| 12:20:34 | yasemin_ | i think , it should be automatically :) | |
| 12:24:05 | openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Use plain routes list for '/servers' endpoint instead of stevedore https://review.openstack.org/445864 | |
| 12:35:42 | markus_z | oomichi: Please have a look at https://review.openstack.org/#/c/453220/ | |
| 12:40:04 | efried | sdague Remember that thing the other day where threading was killinme and you pointed me to recent eventlet changes? | |
| 12:40:30 | efried | sdague https://review.openstack.org/#/c/453301/ is part of the solution. You're a requirements bigwig, right? | |
| 12:40:37 | efried | Would you mind having a look? | |
| 12:46:19 | jaypipes | jianghuaw_: morning (or evening) :) | |
| 12:46:38 | jaypipes | jianghuaw_: so I've been wondering about something in the VGPU spec... | |
| 12:46:44 | sdague | efried: +2 | |
| 12:46:53 | efried | sdague Thank you sir. | |
| 12:46:59 | jaypipes | are you and Bob Ball available for a Google hangout this morning to discuss a couple things? | |
| 12:51:38 | gmann | Nova API meeting in 10 min on #openstack-meeting-4 | |
| 12:54:06 | markus_z | alex_xu: Please have a look at https://review.openstack.org/#/c/453220/ | |
| 12:58:19 | jaypipes | mdbooth: "searchlight should load it from nova" <-- hmm, pretty sure Searchlight isn't going to want to do that :) | |
| 12:58:30 | jaypipes | mdbooth: though I agree with your general statement. | |
| 12:58:37 | jaypipes | about long-term maintenance of the code. | |
| 12:59:09 | mdbooth | jaypipes: I briefly poked around in searchlight's nova plugin, and it contains a bunch of assumptions about state transitions | |
| 12:59:24 | mdbooth | Things which we don't even do a good job of following just within compute manager | |
| 12:59:25 | jaypipes | mdbooth: like I said, I agree with your general statement :) | |
| 12:59:51 | jaypipes | mdbooth: I'm just saying I don't believe searchlight will want to make a dependency on any nova Python package. | |
| 12:59:56 | mdbooth | Duplicating in a proverbial locked cabinet behind a door marked 'beware of the leopard' seems unwise | |
| 13:00:16 | jaypipes | mdbooth: unless we created a different wheel entirely for this, housed in the nova source tree, but packaged separately. | |
| 13:00:25 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: Add snapshot id to the snapshot notifications https://review.openstack.org/453077 | |
| 13:00:37 | mdbooth | jaypipes: Right, I wasn't really suggesting that. More that if they defined a plugin api, they could just load a bunch of plugins | |
| 13:00:50 | jaypipes | ya | |
| 13:00:52 | mdbooth | And the nova plugin would be part of nova | |
| 13:00:55 | jaypipes | totally makes sense. | |
| 13:01:01 | mriedem | searchlight wouldn't import nova, it would load a stevedore extension right? | |
| 13:01:11 | mriedem | all runtime shenanigans | |
| 13:01:22 | mdbooth | mriedem: Yeah, totally at runtime | |
| 13:01:28 | jaypipes | mriedem: yeah, but how would the Nova plugin code "get" to searchlight? | |
| 13:01:37 | jaypipes | mriedem: you need to ship the code from something... | |
| 13:02:15 | jaypipes | mriedem: that's all I'm referring to here. the mechanism for delivering that code into a place searchlight knows about... | |
| 13:02:50 | jaypipes | alex_xu: around? I'd love to make progress on https://review.openstack.org/#/c/376200/ today if you can. | |
| 13:04:21 | bauzas | johnthetubaguy: reviewing it now | |
| 13:09:19 | yasemin_ | artom, which one is true for compute nodes, placement-api or placement-client? | |
| 13:11:06 | dane-fichter | jaypipes: have you looked at the newest draft of the cert validation spec? | |
| 13:11:18 | jaypipes | dane-fichter: no sir, not yet. | |
| 13:11:56 | dane-fichter | jaypipes: ok. I think peter posted it yesterday but it was after y'all had already kind of established your queues for spec review day | |
| 13:12:51 | jaypipes | dane-fichter: will take another look today. | |
| 13:12:59 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: Add snapshot id to the snapshot notifications https://review.openstack.org/453077 | |
| 13:13:03 | dane-fichter | jaypipes: thanks jay! | |
| 13:13:08 | yasemin_ | frickler, artom, which one is true for compute nodes, placement-api or placement-client? | |
| 13:28:36 | mriedem | claudiub: for volume attach, does hyperv honor the user-supplied device name (if one is provided)? | |
| 13:36:51 | artom | mriedem, wait, in a Windows guest? | |
| 13:36:56 | artom | As in, 'D:' ? | |
| 13:37:39 | mriedem | hyperv can run linux guests right? | |
| 13:37:46 | artom | yasemin_, I'm not sure I understand your question, how can placement-api be true on a compute node? | |
| 13:37:46 | mriedem | but that's a good question - what if it's a windows guest | |
| 13:38:03 | mriedem | yasemin_: for devstack? placement-client | |
| 13:38:10 | mriedem | as a subnode | |
| 13:38:12 | mriedem | in devstack | |
| 13:38:19 | yasemin_ | artom, yes for devstack compute node enable services | |
| 13:38:44 | artom | mriedem, huh, TIL hyperv isn't limited to Windows guests | |
| 13:38:50 | mriedem | yasemin_: https://github.com/openstack-infra/devstack-gate/blob/master/features.yaml#L269 | |
| 13:39:12 | mriedem | yasemin_: that's when d-g is setting up a subnode | |
| 13:39:13 | artom | yasemin_, compute node only needs placement-client, methinks | |
| 13:40:06 | yasemin_ | okey | |
| 13:40:50 | EmilienM | hello here, can you review this patch please? https://review.openstack.org/#/c/453338 - it would help TripleO CI to test upgrades from Ocata to Pike. Thanks a lot! | |
| 13:42:09 | jaypipes | melwitt: pls see my comments on https://review.openstack.org/#/c/453030/... | |
| 13:42:29 | jaypipes | melwitt: it's already been approved, but I have some requests for you in there. | |
| 13:43:16 | jaypipes | melwitt: we should really be using GET /usages, not GET /allocations for the quota-counting call. | |
| 13:46:01 | claudiub | mriedem: hi. nope, we can't honor that. | |
| 13:46:30 | claudiub | mriedem: and ofc, hyper-v can run linux guests. :) | |
| 13:47:14 | mriedem | claudiub: thanks | |
| 13:47:17 | mriedem | i'm not sure about vmware | |
| 13:47:22 | mriedem | i think xen might be the only one | |
| 13:49:10 | bauzas | jaypipes: heh, split-brain is not exactly happening in systems with leader elections | |
| 13:49:19 | bauzas | it can happen there, but it's just a case | |
| 13:49:39 | bauzas | my own definition of split-brain is when you have 2 datasets being inconsistent | |
| 13:49:53 | bauzas | which is larger than yours :) | |
| 13:56:32 | melwitt | jaypipes: okay, thanks. I can propose an amendment to the spec | |
| 13:58:28 | openstackgerrit | Maciej Szankin proposed openstack/nova-specs master: Add remove-nova-cert spec https://review.openstack.org/448882 | |
| 14:02:40 | gmann | sdague: on extension removal, as i updated the base patch in that series, i ll rebase others if needed that might need re approval | |
| 14:03:04 | gmann | johnthetubaguy: can you check this base patch https://review.openstack.org/#/c/450834/ | |
| 14:05:57 | dane-fichter | mriedem: when you get a chance, could you take a look at the newest cursive patch? it passes the experimental check. | |
| 14:06:03 | dane-fichter | mriedem: https://review.openstack.org/#/c/351232/ | |
| 14:07:14 | mriedem | yeah | |
| 14:15:41 | mriedem | jaypipes: on https://review.openstack.org/#/c/453030/ to be clear, are you saying both GET /allocations *and* GET /usages, or just the latter as a new handler? | |
| 14:17:01 | melwitt | is the placement-api-ref on docs.o.o anywhere? I can't find it | |
| 14:17:03 | jaypipes | mriedem: implement both in the REST API but have the quota-handling call only the GET /usages | |
| 14:18:38 | mriedem | jaypipes: why both | |
| 14:18:39 | mriedem | ? | |
| 14:18:46 | mriedem | why do we need two handlers for the same thing? | |
| 14:18:56 | jaypipes | mriedem: well, one is the detail view, the other is the aggregate view. | |
| 14:19:24 | mriedem | i think we need a respin of the spec then with the rest api impact section filled out with those details | |
| 14:19:38 | mriedem | if we're adding a new handler | |
| 14:19:57 | melwitt | mriedem: I'm working on an amendment for the spec to address everyone's comments | |
| 14:20:29 | sdague | melwitt: I'm not sure that publish loop ever fully got setup | |