| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2020-07-27 | |||
| 12:28:42 | bauzas | and then you need to upgrade directly before restarting the services | |
| 12:29:14 | bauzas | once you upgrade to Victoria, then you need to upgrade the DB by nova-manage | |
| 12:29:30 | lyarwood | bauzas: no, you upgrade through each release | |
| 12:29:30 | bauzas | and once you're done, then you restart all the compute services | |
| 12:29:40 | lyarwood | bauzas: for the db migrations | |
| 12:29:48 | bauzas | oh yeah of course you're right | |
| 12:29:58 | lyarwood | bauzas: the classic FFU approach is to stop all openstack services on S | |
| 12:30:05 | lyarwood | bauzas: roll through the db migrations to V | |
| 12:30:09 | bauzas | but during this time, do you have libvirtd running ? | |
| 12:30:10 | lyarwood | bauzas: bring the control plane up on V | |
| 12:30:15 | lyarwood | bauzas: and then computes on V | |
| 12:30:23 | bauzas | on computes | |
| 12:30:26 | lyarwood | bauzas: and yeah assuming there's no underlying host upgrade libvirtd will continue to run | |
| 12:30:59 | bauzas | lyarwood: okay, so, we *could* expect libvirtd to be around when we want to nova-manage a specific compute for reshapes | |
| 12:31:06 | bauzas | that's then acceptable | |
| 12:31:20 | sean-k-mooney | stephenfin: not yes no | |
| 12:31:25 | bauzas | if so, that's simplier | |
| 12:31:25 | lyarwood | bauzas: are you running nova-manage on the compute itself | |
| 12:31:36 | bauzas | lyarwood: I'm about to provide a new spec for it | |
| 12:31:41 | bauzas | for reshapes | |
| 12:32:03 | bauzas | lyarwood: anyway, I'll upload my new revision and you could review it | |
| 12:32:15 | lyarwood | bauzas: yeah happy to | |
| 12:35:05 | sean-k-mooney | * not yet | |
| 12:38:32 | lyarwood | jsuchome: the change LGTM now, did we talk about getting this tested somewhere in the gate previously? | |
| 12:38:54 | lyarwood | jsuchome: would you be able to wire it up in the ceph job? | |
| 12:39:07 | sean-k-mooney | stephenfin: im not sure how there patch and that bug are related | |
| 12:40:00 | sean-k-mooney | i need to check but i did not think we used the devname | |
| 12:40:15 | sean-k-mooney | we should be using just the pci address in nova | |
| 12:41:02 | openstackgerrit | Sylvain Bauza proposed openstack/nova-specs master: Offline Reshape tool spec https://review.opendev.org/742908 | |
| 12:42:02 | jsuchome | lyarwood: I think that was that test job you created previously ... I lost the link | |
| 12:43:21 | jsuchome | lyarwood: I think this one https://review.opendev.org/#/c/727225/ ? | |
| 12:43:59 | sean-k-mooney | hum i guess we do use the device name | |
| 12:44:27 | jsuchome | lyarwood: for release nodes we have this I think https://review.opendev.org/#/c/728095/ ... | |
| 12:50:15 | lyarwood | jsuchome: yeah I wonder if we want to enable this all of the time in the ceph job? | |
| 12:51:11 | jsuchome | lyarwood: unfortunatelly I do not really know how this test infrustructure works ... I remember the job always failed but it does not seem to be related to the change | |
| 12:51:31 | lyarwood | jsuchome: k, let me just wire it up quickly | |
| 12:51:48 | lyarwood | jsuchome: ack on the other change having a releasenote for this | |
| 12:51:54 | lyarwood | jsuchome: as long as it's somewhere I'm okay | |
| 12:52:57 | jsuchome | you even commented back in May about the possible reason for the job failure (https://review.opendev.org/#/c/727225/) | |
| 12:54:51 | bauzas | lyarwood: the spec I raised to you https://review.opendev.org/742908 | |
| 12:55:01 | bauzas | (the offline reshape one) | |
| 12:56:44 | lyarwood | bauzas: ack thanks queued | |
| 12:59:56 | artom | What's the reverse of Depends-On again? | |
| 12:59:59 | artom | Used-by? | |
| 13:00:58 | openstackgerrit | Lee Yarwood proposed openstack/nova master: Add ability to download Glance images into the libvirt image cache via RBD https://review.opendev.org/574301 | |
| 13:00:59 | openstackgerrit | Lee Yarwood proposed openstack/nova master: WIP zuul: Enable [glance]/allowed_direct_url_schemes within nova-ceph-multistore https://review.opendev.org/743220 | |
| 13:01:00 | lyarwood | jsuchome: ^ I had to rebase that change to pick up nova-ceph-multistore | |
| 13:01:46 | lyarwood | hmm acually that job uses ceph for everything so I'm not sure if this is going to exercise it correctly | |
| 13:01:56 | lyarwood | now I remember why I had to create that other job | |
| 13:02:15 | lyarwood | well at least we can show it doesn't break nova-ceph-multistore now | |
| 13:02:30 | artom | Ah, Needed-by | |
| 13:04:54 | jsuchome | lyarwood: ok, let's see how it's gonna fail :-) | |
| 13:12:33 | mnaser | has anyone seen this before? | |
| 13:12:38 | mnaser | select count(*) from key_pairs where user_id='b1807b548e8a422ab74a05fe52f42ea1'; => 100 | |
| 13:12:46 | mnaser | nova quota-show --user b1807b548e8a422ab74a05fe52f42ea1 --detail => in_use = 0 | |
| 13:12:56 | mnaser | i'm going to go start digging into this now but in case someone has just.. ran into it | |
| 13:13:35 | artom | mnaser, stab in the dark, but I assume you've checked soft-deleted stuff? | |
| 13:14:15 | mnaser | artom: looks like keypairs are not soft deleted | |
| 13:14:17 | mnaser | no deleted_at column | |
| 13:14:30 | artom | mnaser, ack - as I said, stab in the dark | |
| 13:14:35 | artom | Sorry I can't be of more help :( | |
| 13:14:48 | mnaser | yeah i'm going to dig into where the api gets those numbers from | |
| 13:14:53 | mnaser | i dont think the api counts | |
| 13:15:24 | artom | mnaser, melwitt might have an idea, when she comes online in a few hours | |
| 13:16:34 | mnaser | https://github.com/openstack/nova/blob/master/nova/api/openstack/compute/quota_sets.py#L81-L82 | |
| 13:18:23 | mnaser | ok i think i see the issue here | |
| 13:18:30 | mnaser | i need to provide _both_ tenant id and user_id | |
| 13:18:38 | mnaser | because it tries to get quota of user in context of a specific tenant | |
| 13:19:12 | mnaser | and i assume this was trying to check user_id in scope of admin tenant | |
| 13:23:41 | mnaser | https://github.com/openstack/nova/blob/eeeb964a5f65e6ac31dfb34b1256aaf95db5ba3a/nova/quota.py#L178-L182 | |
| 13:23:42 | mnaser | oh | |
| 13:25:04 | sean-k-mooney | didnt we stop applying quotas for keypairs as part of unified limits? | |
| 13:47:41 | openstackgerrit | Stephen Finucane proposed openstack/nova master: libvirt: Drop support for UML https://review.opendev.org/743230 | |
| 13:47:41 | openstackgerrit | Stephen Finucane proposed openstack/nova master: libvirt: Drop support for Xen https://review.opendev.org/743231 | |
| 13:52:34 | kashyap | stephenfin: Oh, wauw - dropping Xen altogether? | |
| 13:52:57 | kashyap | "there doesn't appear to be | |
| 13:52:59 | kashyap | anyone using it" | |
| 13:53:15 | stephenfin | We agreed to drop XenAPI this cycle during the PTG, but that patch is just focusing on libvirt+xen | |
| 13:53:18 | kashyap | ... is a bold claim, at least for Xen :-) | |
| 13:53:28 | kashyap | stephenfin: I see. Missed that bit. | |
| 13:53:42 | kashyap | Yeah, but I agree, though - on ruthless pruning | |
| 13:54:18 | kashyap | We maintain what's actively used; this way it gives a clear picture of WTF is _really_ maintained | |
| 13:54:43 | kashyap | Thanks for doing the tedious donkey work. | |
| 15:03:22 | melwitt | mnaser: fyi in_use for keypairs, and server_group_members have always been 0, it's a legacy thing. I didn't change the behavior during the re-architecture implementation in pike bc I wasn't bumping the microversion. api behavior was meant to stay the same before and after, as much as possible | |
| 15:32:59 | openstackgerrit | Merged openstack/nova master: Implement extend_volume for libvirt NFS volume driver https://review.opendev.org/739077 | |
| 15:49:39 | openstackgerrit | Stephen Finucane proposed openstack/nova stable/stein: objects: Update keypairs when saving an instance https://review.opendev.org/743264 | |
| 15:53:38 | openstackgerrit | Stephen Finucane proposed openstack/nova stable/rocky: objects: Update keypairs when saving an instance https://review.opendev.org/743268 | |
| 15:57:06 | openstackgerrit | Stephen Finucane proposed openstack/nova stable/queens: objects: Update keypairs when saving an instance https://review.opendev.org/743271 | |
| 16:32:28 | stephenfin | kashyap: Do you understand the various live migration parameters? Any chance you could look at this tomorrow, if so? https://review.opendev.org/#/c/741473/ | |
| 16:33:17 | kashyap | stephenfin: Yeah, there's several of them; and we've deprecated a few in the past, introduced some new after that (including yourself :)) | |
| 16:33:21 | kashyap | stephenfin: Will look; have it opened | |
| 16:33:51 | kashyap | Which reminds me, `live_migration_tunnelled` should be deprecated ... we should have a note in the code somewhere. /me scours | |
| 16:34:52 | kashyap | Ha, I recall documenting it here in this commit - https://opendev.org/openstack/nova/commit/9160fe5098 | |
| 16:35:01 | kashyap | Although opendev.org is slow as molasses here for me | |
| 16:36:09 | kashyap | # Allowing both "tunnelling via libvirtd" (which will be | |
| 16:36:09 | kashyap | # deprecated once the MIN_{LIBVIRT,QEMU}_VERSION is sufficiently | |
| 16:36:09 | kashyap | # new enough) and "native TLS" options at the same time is | |
| 16:36:10 | kashyap | # nonsensical. | |
| 16:38:45 | kashyap | Ah, I can deprecate them now, actually... | |
| 16:46:29 | kashyap | (But for now, I've added a quick comment; tomm, I'll look at the interactions b/n different flags) | |