| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-08-01 | |||
| 18:46:21 | dansmith | tha's mah bad | |
| 18:46:29 | dansmith | I have to fix unit tests on my top patch anyway | |
| 18:46:46 | dansmith | or, hmm, maybe that's the fail | |
| 18:46:54 | mriedem | i'm sure it is | |
| 18:47:28 | dansmith | I really don't think I resolved a conflict here, | |
| 18:47:33 | dansmith | but it looks like it would have so maybe I did | |
| 18:48:55 | dansmith | hey look'a the'r | |
| 18:49:07 | openstackgerrit | Dan Smith proposed openstack/nova master: Make ResourceTracker.stats node-specific https://review.openstack.org/587636 | |
| 18:49:08 | openstackgerrit | Dan Smith proposed openstack/nova master: Hook resource_tracker to remove stale node information https://review.openstack.org/587922 | |
| 18:49:54 | mriedem | lgtm | |
| 18:56:44 | mriedem | dansmith: one question in https://review.openstack.org/#/c/587922/ | |
| 18:59:02 | mriedem | i also wonder if it's worth backporting https://review.openstack.org/#/c/584204/ | |
| 18:59:23 | mriedem | it's a super latent issue and allocations in placement would still show root_gb usage for bfv instances in queens | |
| 18:59:25 | mriedem | so maybe not worth it | |
| 18:59:51 | dansmith | mriedem: answwered | |
| 19:00:02 | openstackgerrit | Matt Riedemann proposed openstack/nova stable/queens: Make ResourceTracker.stats node-specific https://review.openstack.org/587976 | |
| 19:05:42 | sean-k-mooney | dansmith: remove_node is called within the update_available_resource perodic task. if we delete the hypervior form the compute then read it after the periodic task. are they running in seperate threads? | |
| 19:06:01 | sean-k-mooney | sorry that got messed up | |
| 19:06:30 | sean-k-mooney | do the periodic tasks run in a seperate full os thread form teh rest of the compute agent or a green thread | |
| 19:07:24 | dansmith | everything is a green thread | |
| 19:07:31 | dansmith | with the exception of the libvirt event feeder thread | |
| 19:08:18 | sean-k-mooney | oh ok then ya we are not doing any io in remvoe_node so we wont yield so we should not race | |
| 19:38:34 | melwitt | how do we specify a microversion to openstackclient? trying to use osc-placement and most of the abilities are in later microversions | |
| 19:40:30 | melwitt | I guess it wouldn't be documented by openstackclient because it doesn't officially support it, just passes through to novaclient | |
| 19:41:06 | openstack | bug 1784705 in OpenStack Compute (nova) queens "ResourceTracker.stats can leak across multiple ironic nodes" [High,In progress] https://launchpad.net/bugs/1784705 - Assigned to Matt Riedemann (mriedem) | |
| 19:41:06 | openstackgerrit | Matt Riedemann proposed openstack/nova stable/pike: Add recreate test for RT.stats bug 1784705 https://review.openstack.org/588036 | |
| 19:41:07 | openstackgerrit | Matt Riedemann proposed openstack/nova stable/pike: Make ResourceTracker.stats node-specific https://review.openstack.org/588037 | |
| 19:41:15 | mriedem | melwitt: totally different | |
| 19:41:30 | mriedem | novaclient does version negotiation with the server on the cli to see which version to use, | |
| 19:41:37 | mriedem | and uses the max supported between the client and server for the CLI | |
| 19:41:48 | mriedem | osc requires that the user opt into the microversion per request | |
| 19:41:53 | mriedem | so, | |
| 19:42:02 | mriedem | openstack --os-compute-api-version 2.37 server create .... | |
| 19:42:11 | melwitt | right... but as far as what option to pass, I thought osc passes through any CLI option and one of those will tell novaclient | |
| 19:42:28 | mriedem | osc will create novaclient with the version specified yes, | |
| 19:42:37 | melwitt | just found that but it doesn't work for requesting a later placement API microversion? | |
| 19:42:39 | mriedem | because novaclient's python API bindings default to no microversion, so 2.1 | |
| 19:42:50 | mriedem | novaclient != placement... | |
| 19:42:58 | mriedem | there is no 2.1 in placement | |
| 19:43:00 | mriedem | there is 1.0 | |
| 19:43:08 | mriedem | osc-placement doesn't have anything to do with novaclient | |
| 19:43:26 | dansmith | the passthrough is of the version, | |
| 19:43:29 | dansmith | not the CLI opt | |
| 19:43:34 | dansmith | since osc is its own client | |
| 19:43:39 | melwitt | oh, right. so how do we do microversions with it? it's saying I can't 'resource class list' because 1.0 doesn't support it. but it's in the command list | |
| 19:43:53 | dansmith | there's an option to specify the version | |
| 19:44:18 | dansmith | --os-compute-api-version I think | |
| 19:44:26 | mriedem | openstack --os-placement-api-version 1.7 resource class list | |
| 19:44:30 | mriedem | something like that, | |
| 19:44:37 | mriedem | it's not 'compute' because it's not nova | |
| 19:44:39 | mriedem | in the service catalog | |
| 19:44:49 | mriedem | the --os-<service-type>-api-version is based on the service catalog | |
| 19:45:06 | mriedem | https://docs.openstack.org/osc-placement/latest/cli/index.html#resource-class-list | |
| 19:45:09 | mriedem | This command requires at least –os-placement-api-version 1.2. | |
| 19:45:19 | mriedem | so, openstack --os-placement-api-version 1.2 resource class list | |
| 19:45:37 | mriedem | if you know you're at some max version of placement, you could just set an env var | |
| 19:45:43 | mriedem | export OS_PLACEMENT_API_VERSION=1.7 | |
| 19:45:45 | dansmith | yeah, she said novaclient so I assumed she was hitting nova | |
| 19:45:49 | dansmith | but placement for placement obviously | |
| 19:45:49 | melwitt | thanks, I just couldn't find that option --os-placement-api-version anywhere | |
| 19:45:54 | melwitt | yeah, sorry | |
| 19:45:54 | mriedem | and then don't have to worry about per-call mv | |
| 19:46:08 | mriedem | melwitt: should be in the help output for the command | |
| 19:46:17 | mriedem | and https://docs.openstack.org/osc-placement/latest/user/index.html#microversion-usage | |
| 19:46:28 | melwitt | I got confused when I started thinking about passthrough of microversions when using osc | |
| 19:46:41 | melwitt | and it made me think of novaclient and the --os-compute-api-version thing | |
| 19:47:00 | mriedem | yeah, it would be the same if there was a python-placementclient thing | |
| 19:47:11 | mriedem | but we've intentionally always avoided that | |
| 19:47:15 | melwitt | ok yeah, I see now 'openstack resource class list -h' | |
| 19:47:29 | mriedem | b/c we have osc for cli and openstacksdk for sdk | |
| 19:48:05 | melwitt | "Return a list of all resource classes. This command requires at least --os- | |
| 19:48:06 | melwitt | placement-api-version 1.2." | |
| 19:48:28 | melwitt | thanks | |
| 20:18:16 | openstackgerrit | Merged openstack/nova master: Nix unused raise_if_custom_resource_class_pre_v1_1 https://review.openstack.org/575847 | |
| 20:42:32 | openstack | bug 1784705 in OpenStack Compute (nova) queens "ResourceTracker.stats can leak across multiple ironic nodes" [High,In progress] https://launchpad.net/bugs/1784705 - Assigned to Matt Riedemann (mriedem) | |
| 20:42:32 | openstackgerrit | Matt Riedemann proposed openstack/nova stable/ocata: Add recreate test for RT.stats bug 1784705 https://review.openstack.org/588076 | |
| 20:42:33 | openstackgerrit | Matt Riedemann proposed openstack/nova stable/ocata: Make ResourceTracker.stats node-specific https://review.openstack.org/588077 | |
| 21:03:42 | mriedem | ooo goody 5 new bugs in triage | |
| 21:04:48 | melwitt | nooo | |
| 21:04:57 | mriedem | well one is against mitaka so.... | |
| 21:05:49 | melwitt | yeah, just opened that one | |
| 21:06:21 | melwitt | I've noticed a fair bit of bugs being added to nova that used to be on other projects lately | |
| 21:06:45 | melwitt | or something. really old bugs coming back as new | |
| 21:15:59 | melwitt | mnaser: have you ever used evacuate with ceph-backed instances? looking at this bug and wondering if evacuate doesn't work with ceph period or if it's only if you've enabled the ceph "exclusive-lock" feature on a disk https://bugs.launchpad.net/nova/+bug/1781878 | |
| 21:16:00 | openstack | Launchpad bug 1781878 in OpenStack Compute (nova) "VM fails to boot after evacuation when it uses ceph disk" [Undecided,New] | |
| 21:16:55 | melwitt | the way the bug is worded, I'm not certain whether they're saying evacuate is broken with ceph out-of-the-box | |
| 21:17:09 | mnaser | i am almost certain this is a manifestation of another bug | |
| 21:17:13 | melwitt | so I was wondering if you've ever done evacuate successfully | |
| 21:17:14 | mnaser | of a poorly upgraded ceph cluster | |
| 21:17:15 | mnaser | one second | |
| 21:17:19 | melwitt | k | |
| 21:17:20 | mnaser | yes, evacuate works all the time, but one second | |
| 21:18:00 | openstackgerrit | Eric Fried proposed openstack/nova master: Compute: Handle reshaped provider trees https://review.openstack.org/576236 | |
| 21:18:08 | mnaser | melwitt: http://lists.ceph.com/pipermail/ceph-users-ceph.com/2017-September/020722.html | |
| 21:18:10 | mnaser | alongside doc there | |
| 21:18:39 | melwitt | thanks | |
| 21:18:42 | mnaser | saaame exact behaviour afaik as original reported issue | |
| 21:19:36 | melwitt | oh, this is familiar ... I remember an internal bug we had about this | |
| 21:19:55 | mnaser | a lot of people were affected by this because the step to tell you to do this was forgotten in the original ceph upgrade docs | |
| 21:20:23 | melwitt | a-ha | |