| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2021-02-01 | |||
| 14:42:41 | dansmith | since we define the job, we're holding them hostage :) | |
| 14:42:56 | dansmith | (or anyone else ^) | |
| 14:43:27 | sean-k-mooney | ya that makes sense | |
| 14:44:21 | gibi | dansmith: looking | |
| 14:44:27 | sean-k-mooney | dansmith: i assume we are not doing that policy file upstae in [post-config| /etc/glance/policy.json] | |
| 14:44:29 | sean-k-mooney | for some reason | |
| 14:44:58 | sean-k-mooney | the default policy file shoudl be empty | |
| 14:45:29 | sean-k-mooney | also i guess at somepoint that need to be a policy.yaml file | |
| 14:46:04 | lyarwood | would post-config bounce the service? | |
| 14:46:17 | sean-k-mooney | no but it runs before its started | |
| 14:46:22 | sean-k-mooney | so it does not need too | |
| 14:46:43 | sean-k-mooney | post-config is after all the files are generated but before the services are started for the first time | |
| 14:47:03 | sean-k-mooney | excluddign rabbit and mariadb which start before post-config for reasons | |
| 14:47:04 | lyarwood | ah understood | |
| 14:47:32 | sean-k-mooney | if that was doing something more complicated post-cofnig would not work | |
| 14:47:52 | sean-k-mooney | but for this i think it would and they could then override it by inheriting form the job if they needed too | |
| 14:48:13 | sean-k-mooney | as a quick fix however that will work hence +1 | |
| 14:50:13 | openstackgerrit | Lee Yarwood proposed openstack/nova master: Add regression test for bug #1908075 https://review.opendev.org/c/openstack/nova/+/766976 | |
| 14:50:15 | openstack | bug 1908075 in OpenStack Compute (nova) "Nova allows a non-multiattach volume to be attached to multiple instances *if* its volume state is reset by an admin" [Undecided,New] https://launchpad.net/bugs/1908075 - Assigned to Lee Yarwood (lyarwood) | |
| 14:50:15 | openstackgerrit | Lee Yarwood proposed openstack/nova master: api: Reject volume attach requests when an active bdm exists https://review.opendev.org/c/openstack/nova/+/768472 | |
| 14:50:15 | openstackgerrit | Lee Yarwood proposed openstack/nova master: fup: Merge duplicate volume attachment checks https://review.opendev.org/c/openstack/nova/+/773380 | |
| 14:52:16 | gibi | dansmith: done | |
| 15:03:32 | openstackgerrit | sean mooney proposed openstack/nova master: [FUP] this change removes an unused paramater https://review.opendev.org/c/openstack/nova/+/773409 | |
| 15:03:40 | dansmith | gibi: thanks | |
| 15:03:59 | sean-k-mooney | FYI that ^ is really trivial if we can merge that quickly | |
| 15:04:51 | sean-k-mooney | lyarwood: i asked you about squash merging it into the backport of the original patch. are you still ok with that | |
| 15:05:10 | sean-k-mooney | lyarwood: if so ill repin those now | |
| 15:08:30 | lyarwood | sean-k-mooney: yup that's fine, lets land this first and then respin the backports | |
| 15:08:53 | sean-k-mooney | ok ill hold off then untill its merged on master | |
| 15:13:49 | sean-k-mooney | dansmith: are you still finding gerrit slow. currently im having the same multi second post behavior you were having when submitting comments | |
| 15:14:03 | dansmith | sean-k-mooney: yeah | |
| 15:14:15 | dansmith | sean-k-mooney: apparently that's just "how it is" with the new version of gerrit | |
| 15:14:19 | sean-k-mooney | ok so its not on my end so. | |
| 15:14:21 | sean-k-mooney | ack | |
| 15:14:27 | dansmith | suspected because they now use git as their db (!) | |
| 15:14:55 | sean-k-mooney | i wonder if that menas its more disk io sensitive then it used to be | |
| 15:15:30 | sean-k-mooney | im sure storing the data in git has proably resulted in more updates for that | |
| 15:15:33 | dansmith | it is, a lot | |
| 15:16:27 | sean-k-mooney | we need to get intel,samsung or micron to donate some fast ssds to infra :) | |
| 15:56:41 | openstackgerrit | Merged openstack/nova master: libvirt: Use specific user when probing encrypted rbd disks during extend https://review.opendev.org/c/openstack/nova/+/772869 | |
| 16:00:45 | openstackgerrit | Lee Yarwood proposed openstack/nova stable/victoria: libvirt: Use specific user when probing encrypted rbd disks during extend https://review.opendev.org/c/openstack/nova/+/773443 | |
| 16:01:07 | openstackgerrit | Lee Yarwood proposed openstack/nova stable/ussuri: libvirt: Use specific user when probing encrypted rbd disks during extend https://review.opendev.org/c/openstack/nova/+/773444 | |
| 16:02:01 | supamatt | sean-k-mooney: I'd talk to Koxia and see if they want to sponsor our lab with storage ssd's lol | |
| 16:11:53 | openstackgerrit | Balazs Gibizer proposed openstack/nova stable/ussuri: Warn when starting services with older than N-1 computes https://review.opendev.org/c/openstack/nova/+/770764 | |
| 16:11:54 | openstackgerrit | Balazs Gibizer proposed openstack/nova stable/ussuri: Add upgrade check about old computes https://review.opendev.org/c/openstack/nova/+/770765 | |
| 16:26:27 | ganso | Hi Nova folks! Quick question: is there any way to migrate a VM that is in an affinity group without using the deprecated --live option? I have 2 VMs in the same host, they're in the same affinity group, and trying to move either (without --live) results in "No valid host". My main concern is, when that option is removed will there be no way to migrate those VMs? | |
| 16:28:14 | supamatt | ganso: try with --os-compute-api-version 2.11 | |
| 16:30:32 | ganso | supamatt: I'm actually already using --os-compute-api-version 2.30 to use the --host parameter. I still get "No valid host" | |
| 16:31:30 | supamatt | okay so try 2.11 | |
| 16:32:40 | ganso | supamatt: same error | |
| 16:33:05 | supamatt | you may want to check the nova scheduler logs and see whats going on | |
| 16:33:49 | ganso | supamatt: already did, "Filter ServerGroupAffinityFilter returned 0 hosts" | |
| 16:34:14 | ganso | which brings me back to my first question | |
| 17:59:36 | sean-k-mooney | ganso: you would need to force the migration | |
| 18:00:06 | sean-k-mooney | ganso: i dont think there is any supported way beyond that | |
| 18:00:06 | ganso | sean-k-mooney: with the deprecated option --live <host> ? | |
| 18:00:16 | sean-k-mooney | ganso: hat release of nova are you deploying | |
| 18:00:21 | sean-k-mooney | *what | |
| 18:00:36 | ganso | sean-k-mooney: stein | |
| 18:00:45 | sean-k-mooney | im pretty sure you would need to disable the compute node late affinity check | |
| 18:00:57 | sean-k-mooney | and i think you need to pass a --force option | |
| 18:01:39 | ganso | sean-k-mooney: it works with --live <host>. But my concern is that this option is deprecated. Has it been removed from a later release or planned to be removed? | |
| 18:01:40 | sean-k-mooney | ganso: so you would have to set this to true https://docs.openstack.org/nova/latest/configuration/config.html#workarounds.disable_group_policy_check_upcall | |
| 18:02:26 | sean-k-mooney | ganso: it has been removed in a later microversion yes | |
| 18:02:30 | sean-k-mooney | its still in the api | |
| 18:02:37 | sean-k-mooney | but not with the latest version | |
| 18:03:21 | sean-k-mooney | ganso: basically there no fully supported way to move vms in a server affinity group | |
| 18:03:35 | sean-k-mooney | we cant do a group migration for example | |
| 18:04:09 | sean-k-mooney | the best you could do without forcing the migration would likely be to shelve all vms in the server group | |
| 18:04:26 | sean-k-mooney | then unsleve them | |
| 18:04:54 | sean-k-mooney | thats a specualtion and not something i have tested by the way | |
| 18:05:11 | sean-k-mooney | i guess if you shelved all but 1 of the vms you could live migrate it too and then unshelve the rest | |
| 18:05:32 | ganso | sean-k-mooney: but then the VMs have to be shut down. If the main point was do have a live-migration, it cannot be shelved | |
| 18:05:35 | sean-k-mooney | but server affinity basically disbales move operations | |
| 18:06:03 | ganso | sean-k-mooney: yea that's exactly what I realized :\ | |
| 18:06:25 | ganso | sean-k-mooney: thanks! | |
| 18:06:36 | sean-k-mooney | so using the deprecated options you can force the migration but then you are violating the policy | |
| 18:06:45 | sean-k-mooney | since you can move them all at once | |
| 18:08:28 | sean-k-mooney | ganso: by the way we removed the ablity to do force live migration in 2.68 https://docs.openstack.org/nova/latest/reference/api-microversion-history.html#id62 | |
| 18:08:36 | sean-k-mooney | and it was added in 2.30 | |
| 18:09:23 | ganso | sean-k-mooney: as long as v2 is not removed (I guess this will take a while), it will still be able to be used I assume | |
| 18:09:37 | ganso | just need to specify a lower microversion in the CLI | |
| 18:10:24 | sean-k-mooney | so this is the release note for the removal https://github.com/openstack/nova/blob/master/releasenotes/notes/remove-live-migrate-evacuate-force-flag-cb50608d5930585c.yaml | |
| 18:11:33 | sean-k-mooney | i belive we cap the ablity to force migration with 2.30-2.67 | |
| 18:13:36 | sean-k-mooney | ganso: http://paste.openstack.org/show/802192/ | |
| 18:14:26 | sean-k-mooney | so you need to do "nova --os-compute-api-version=2.30 live-migration --force <server> <host>" | |
| 18:14:44 | ganso | sean-k-mooney: yes, it will still work as long as I need a funcionality >2.68 in the same API call | |
| 18:14:51 | ganso | as long as I *don't need | |
| 18:14:57 | sean-k-mooney | ganso: i belive we intentionally do not support it with the openstsck client | |
| 18:15:33 | sean-k-mooney | ganso: well in very rare cases we also drop all compatiablity for older microversions | |
| 18:15:39 | ganso | sean-k-mooney: I've been using the openstack client, and it is slightly different there. The parameter name is "--live" instead of "--force". Where "--live" uses that force behavior, and "--live-migration" uses the non-forced behavior | |
| 18:16:04 | sean-k-mooney | ah right correct | |
| 18:17:03 | sean-k-mooney | v4 | |
| 18:17:10 | sean-k-mooney | we had a v3 already which we then deleted | |
| 18:17:41 | sean-k-mooney | ganso: we only drop all support if we are fully removing a subsystem | |
| 18:18:07 | sean-k-mooney | ganso: for example when we removed nova-networks or xenapi code entirely | |
| 18:18:50 | ganso | ah great then it is going to be fine \o/ | |
| 18:19:16 | sean-k-mooney | yep untill we do a v3/4 which wont be anytime soon if ever | |
| 18:19:39 | sean-k-mooney | that said we are getting close to 2.100 so at somepoint we proably should | |
| 18:23:00 | openstackgerrit | Merged openstack/nova master: Restart glance after updating policy in job https://review.opendev.org/c/openstack/nova/+/773192 | |