| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2020-10-27 | |||
| 09:07:48 | bauzas | lyarwood: so, no need to use another release ? | |
| 09:07:50 | kashyap | (Thanks also to Mel Witt and DanPB for the root cause analysis :-)) | |
| 09:08:30 | lyarwood | bauzas: hopefully not | |
| 09:30:02 | openstackgerrit | Lee Yarwood proposed openstack/nova master: releasenotes: Add fixes note for bug #1882521 https://review.opendev.org/758971 | |
| 09:30:02 | openstack | bug 1882521 in Cinder "Failing device detachments on Focal" [Undecided,New] https://launchpad.net/bugs/1882521 | |
| 09:30:02 | lyarwood | stephenfin: ^ updated the releasenote | |
| 09:39:01 | Diavel | Hi guys, I wanted to ask for your opinion. Does it make sense to add a notification for compute nodes when an instance metadata gets updated -- see here https://github.com/openstack/nova/blob/master/nova/compute/manager.py#L4232? Currently only the controller sends a notification `instance.update` with `InstanceUpdatePayload` which is fine but imho it could also be useful to have it sent by compute nodes. The thing is I am working on | |
| 09:39:01 | Diavel | a tool which consumes amqp messages and runs on every compute node, to simplify life it consumes only messages targeted/emitted for/from the specific compute node. With this the code of the tool I am doing could stay pretty clean. | |
| 09:39:13 | openstackgerrit | Lee Yarwood proposed openstack/nova master: libvirt: Only ask tpool.Proxy to autowrap vir* classes https://review.opendev.org/759831 | |
| 09:42:07 | gibi | Diavel: Hi! the instance metadata change handled fully in the controller. There is an RPC cast to the compute but that was only there for the XenAPI virt driver that is recently removed so the RPC will go away in the future | |
| 09:43:31 | Diavel | gibi: that means that in the future versions if I wanna keep a watch over instance metadata updates I have no other option than to watch controller amqp messages? | |
| 09:43:32 | gibi | Diavel: also duplicating the same notification feels bad to me | |
| 09:43:58 | gibi | Diavel: I assume this is not the only notification that is emited by the controller only | |
| 09:45:38 | gibi | so overall it is good idea to watch for the controller too | |
| 09:46:08 | gibi | you can configure nova to emit the notification to a specific message bus independently from th RPC config | |
| 09:46:27 | gibi | so you can create a config where every nova service uses the same message bus for notifications | |
| 09:46:34 | gibi | and then you only need to listen to that bus | |
| 09:46:36 | gibi | in a single place | |
| 09:47:31 | Diavel | yeah that's what I am doing.. actually I have configured it so that every compute uses its own message bus and that is consumed by that tool which I am working on | |
| 09:47:42 | Diavel | but to be honest only the metadata update event is missing | |
| 09:47:55 | Diavel | otherwise everything else I have tested is sent by compute | |
| 09:51:31 | gibi | Diavel: just top of my head instance tags and update on instance attributes like display_name are also handled only on the controller | |
| 09:53:54 | Diavel | gibi: ok got it. Thanks for clarifying this. :) | |
| 09:55:46 | stephenfin | lyarwood: good sleuthing | |
| 10:01:22 | lyarwood | stephenfin: danpb and melwitt found it tbh | |
| 10:01:40 | stephenfin | Fair | |
| 10:01:49 | stephenfin | I'm just happy it's not _my_ mypy stuff causing issues :) | |
| 10:02:17 | lyarwood | haha indeed | |
| 10:12:02 | gibi | stephenfin: it is someone else's mypy stuff but it is still some typing issue :D | |
| 10:12:39 | gibi | so I guess this is still a negative thing for the typing effort | |
| 10:23:33 | bauzas | stephenfin: I prefer my pie rather than mypy :p | |
| 10:24:22 | bauzas | (a quince pie) | |
| 12:27:31 | sean-k-mooney | is the gate currently broken by the libvirt issue by the way? | |
| 12:27:36 | sean-k-mooney | before i recheck something | |
| 12:27:47 | gibi | sean-k-mooney: yes it is | |
| 12:28:02 | gibi | nova-compute fails to start | |
| 12:28:32 | gibi | sean-k-mooney: we hope https://review.opendev.org/#/c/759831/ will solve it | |
| 12:28:56 | sean-k-mooney | ... ok so there is no point in me asking for reviews i have 2 patches i need to get merged soon | |
| 12:29:13 | sean-k-mooney | ya i saw that earlier i was not sure if it had merged | |
| 12:29:17 | sean-k-mooney | and i lost the link | |
| 12:29:20 | sean-k-mooney | thanks :) | |
| 12:30:02 | sean-k-mooney | return tuple([cls[1] for cls in classes if cls[0].startswith("vir")]) should work provided they keep to there nameing scheme | |
| 12:31:03 | sean-k-mooney | well when the gate is green again can people take a look at https://review.opendev.org/#/c/751642/7 and ill recheck https://review.opendev.org/#/c/602432/22 later | |
| 12:31:16 | gmann | gibi: thanks | |
| 13:12:27 | sean-k-mooney | lyarwood: gibi stephenfin just spoke to fungi and frickler in #opendev we might need to make grenade non voting for https://review.opendev.org/#/c/759831/ to merge or merge it to victoria first and forward port it | |
| 13:13:22 | lyarwood | sean-k-mooney: oh fun | |
| 13:13:49 | lyarwood | sean-k-mooney: I didn't realise this was also broken in V | |
| 13:13:52 | fungi | usually you merge to the oldest affected branch first and work your way up to the newest | |
| 13:14:09 | fungi | at least if upgrade testing is impacted | |
| 13:14:25 | sean-k-mooney | lyarwood: i saw it broken on a grenade job fro one of frickler patches it failed in the inital stack i think | |
| 13:14:37 | lyarwood | sean-k-mooney: did we also bump eventlet etc there? | |
| 13:15:19 | lyarwood | the other way of fixing this is to revert requirements.txt bumps | |
| 13:15:56 | sean-k-mooney | actully looking at it again it might be the scond stacking | |
| 13:16:06 | sean-k-mooney | https://zuul.opendev.org/t/openstack/build/89443efd462f4d7fb2e470568546b232 | |
| 13:16:16 | lyarwood | hopefully that's the case | |
| 13:16:34 | fungi | yeah, if that's the case then just fixing in master should be sufficient | |
| 13:16:36 | sean-k-mooney | it https://storage.bhs.cloud.ovh.net/v1/AUTH_dcaab5e32b234d56b626f72581e3644c/zuul_opendev_logs_894/759850/1/check/designate-grenade-pdns4/89443ef/controller/logs/screen-n-cpu.txt | |
| 13:16:53 | sean-k-mooney | there is a grenade.sh log | |
| 13:16:55 | sean-k-mooney | https://storage.bhs.cloud.ovh.net/v1/AUTH_dcaab5e32b234d56b626f72581e3644c/zuul_opendev_logs_894/759850/1/check/designate-grenade-pdns4/89443ef/controller/logs/grenade.sh_log.txt | |
| 13:17:11 | sean-k-mooney | so that means the first stacking passed right | |
| 13:17:17 | sean-k-mooney | e.g. victoia | |
| 13:17:28 | sean-k-mooney | so this is just a master issue | |
| 13:17:40 | lyarwood | coolio | |
| 13:18:19 | sean-k-mooney | well im not sure about that | |
| 13:18:33 | sean-k-mooney | e.g. i dont know if we alwasy have a grenade.sh log or not in that job | |
| 13:19:05 | sean-k-mooney | ok we are good | |
| 13:19:07 | sean-k-mooney | https://storage.bhs.cloud.ovh.net/v1/AUTH_dcaab5e32b234d56b626f72581e3644c/zuul_opendev_logs_894/759850/1/check/designate-grenade-pdns4/89443ef/controller/logs/old/devstacklog.txt | |
| 13:19:16 | sean-k-mooney | we have a succeful inital devstack run | |
| 13:19:25 | sean-k-mooney | so ya its failing on the master upgrade | |
| 13:20:31 | openstackgerrit | Lee Yarwood proposed openstack/nova master: releasenotes: Add fixes note for bug #1882521 https://review.opendev.org/758971 | |
| 13:20:31 | openstack | bug 1882521 in OpenStack Compute (nova) "Failing device detachments on Focal" [High,Fix released] https://launchpad.net/bugs/1882521 - Assigned to Lee Yarwood (lyarwood) | |
| 13:21:51 | lyarwood | sean-k-mooney: cool, so everything is green in CI at the moment, I'll continue watching it. | |
| 13:26:14 | gibi | fingers crossed | |
| 13:38:54 | bauzas | sean-k-mooney: agreed with you on the "vir" prefix, I'm a bit afraid they could change their naming | |
| 13:39:12 | bauzas | but let's merge the change now and discuss for it on a FUP | |
| 13:39:43 | sean-k-mooney | yep if i was very concerned about that i would have -1d | |
| 13:40:02 | bauzas | sean-k-mooney: i did in on PS3, but after thinking on it, I finally provided a +2 | |
| 13:40:25 | sean-k-mooney | i think we can cross that bridge when it comes up | |
| 13:41:02 | lyarwood | right, I really doubt it's ever going to change tbh | |
| 13:41:13 | lyarwood | even if it did checking for _ wouldn't be enough | |
| 13:41:39 | lyarwood | so yeah lets cross that bridge if it ever comes up | |
| 13:42:09 | bauzas | sure, but i wonder whether we should verify that only LibvirtError should be a private class and maybe if not then stopping the compute service when restarting | |
| 13:42:16 | bauzas | lyarwood: ^ | |
| 13:43:26 | bauzas | so in case libvirt supports a new public class that is not having a vir prefix by a new major release (hopefully), then the compute service would stop to run and we would get an exception explaining it | |
| 13:44:19 | bauzas | maybe a overthought so, but here the problem is that I don't see any API situation for python-libvirt | |
| 13:44:40 | bauzas | we only have https://libvirt.org/python.html that is documented | |
| 14:09:15 | openstackgerrit | Lee Yarwood proposed openstack/nova-specs master: WIP/DNM - Image defined ephemeral storage encryption https://review.opendev.org/752284 | |
| 14:23:33 | lyarwood | bauzas: sorry my irc client disconnected silently, yeah we can make this more robust but I don't think the return on that effort is going to be high tbh | |
| 14:24:10 | bauzas | lyarwood: I wish libvirt would provide a better robust API | |
| 14:24:17 | bauzas | for the python bindings | |
| 14:24:25 | bauzas | like, having a public scheme | |
| 14:24:25 | lyarwood | bauzas: I'm not sure why you think it isn't | |
| 14:24:47 | lyarwood | bauzas: oh, well it's basically https://libvirt.org/html/libvirt-libvirt-domain.html | |
| 14:24:59 | bauzas | lyarwood: the problem ties to python, by default you can introspect any object | |
| 14:25:13 | lyarwood | bauzas: through the two main classes described in your previous link | |
| 14:25:21 | bauzas | lyarwood: yeah, but that's not the python binding API, right? | |
| 14:26:38 | lyarwood | bauzas: true, we could use better documentation | |
| 14:26:40 | bauzas | totally unrelated, I really would love Gerrit supporing images or memes https://review.opendev.org/#/c/759828/2/specs/wallaby/approved/nova-support-webvnc-with-password-anthentication.rst@175 | |
| 14:26:56 | bauzas | I mean the Gerrit UI | |
| 14:27:19 | bauzas | hyperlinks would stay plain text for gerrit text clients | |