| *** akekane is now known as abhishekk | 04:53 | |
| *** akekane_ is now known as abhishekk | 09:19 | |
| opendevreview | Maksim Malchuk proposed openstack/nova stable/xena: Fix to implement 'pack' or 'spread' VM's NUMA cells https://review.opendev.org/c/openstack/nova/+/829804 | 10:54 |
|---|---|---|
| *** dasm|off is now known as dasm | 13:38 | |
| gibi | bauzas: fyi I cannot join to the tomorrow's nova meeting | 14:33 |
| bauzas | gibi: ack | 14:34 |
| bauzas | gibi: you may be interested in merging the stable/xena branch from https://review.opendev.org/c/openstack/nova/+/864932/4 :) | 15:06 |
| sean-k-mooney | once the xena version are merged ill update the wallaby versions with the new cherry-picked lines | 15:08 |
| gibi | bauzas: sorry I have to get back to this | 15:08 |
| bauzas | sean-k-mooney: ahha, I was about to ping you for https://review.opendev.org/c/openstack/nova/+/865334 | 15:08 |
| dansmith | bauzas: it would be great if we could have this merged before I disappear for the year: https://review.opendev.org/c/openstack/nova-specs/+/863152 | 15:11 |
| bauzas | dansmith: ack, I'll look | 15:11 |
| sean-k-mooney | dansmith: i think you said you had a poc of that somewhere. did you push it for reveiw | 15:14 |
| sean-k-mooney | i know you will be on PTO soon but we can at least start on teh review while your away | 15:15 |
| dansmith | sean-k-mooney: yes: https://review.opendev.org/q/topic:bp%252Fstable-compute-uuid | 15:15 |
| dansmith | as I mentioned, there are two tests in there that are failing because they assert that we recreate compute nodes, which this makes us not do | 15:15 |
| sean-k-mooney | oh ok i missed that then ill try and review it before i finish for the year | 15:15 |
| dansmith | so I left them as failing so that you can see that, and will go back and update the tests (of course) | 15:16 |
| sean-k-mooney | ya you mentioned that the current fixutre for restarting the compute service was problematic | 15:16 |
| sean-k-mooney | ack | 15:16 |
| dansmith | yes, but I think I've fixed that | 15:16 |
| dansmith | the failing tests are just because the test confirms that we re-create a compute node, but with this code, we re-find the old one | 15:17 |
| dansmith | it's been sitting a month at this point so I don't fully remember, and I'll be even more hazy in Jan, but will pick it up then :) | 15:17 |
| sean-k-mooney | ack i think in the long run that is proably more correct anyway (refinding them) | 15:17 |
| dansmith | that's the point of the work, yeah :) | 15:18 |
| dansmith | (well one of the points) | 15:18 |
| dansmith | it avoids the stale/broken placement allocations and such | 15:19 |
| sean-k-mooney | well i think (would have to check) we were reusing the uuid or soemthign liek tha tso that the palcement side was correct | 15:19 |
| bauzas | dansmith: sent to the gate | 15:19 |
| sean-k-mooney | in the funtional tests | 15:19 |
| sean-k-mooney | but thanks for the context | 15:19 |
| bauzas | dansmith: thanks for having added some paragraphs | 15:20 |
| dansmith | bauzas: thanks | 15:20 |
| dansmith | sean-k-mooney: in the ironic case we are I think, not sure about the regular case | 15:21 |
| dansmith | sean-k-mooney: in the case of a host rename, we don't find the old provider by name (because our name changed) so we re-create everything right? | 15:21 |
| sean-k-mooney | i think the way the current recreate worked we were not actully creating a new compute node record in the db | 15:21 |
| sean-k-mooney | * how it worked in teh functional tests | 15:22 |
| dansmith | okay I'm talking about in real life | 15:22 |
| sean-k-mooney | ya so in real life the lookup can fail in one of two ways | 15:23 |
| sean-k-mooney | either the host value does not match adn we create a new compute service entry | 15:23 |
| sean-k-mooney | or the hypervior_hostname does not match and we delete and recreate teh compute node record | 15:23 |
| dansmith | neither will match on a libvirt host that has renamed right? | 15:24 |
| sean-k-mooney | so if the host value changes the service entry gets recreated and if the hypervior_hostname mismatches its the compute node | 15:24 |
| sean-k-mooney | dansmith: if you have notihgn set in the config then yes | 15:24 |
| dansmith | ah, right | 15:25 |
| sean-k-mooney | ooo at least hardcodes the host value | 15:25 |
| sean-k-mooney | so if the hostname change and the config is not updated then it will find the service | 15:25 |
| sean-k-mooney | then delete the compute node and recreate it | 15:25 |
| dansmith | sean-k-mooney: yeah I guess we need to make sure we have those cases covered in the tests.. "host renamed, no config set" and "host renamed, config set to something" | 15:27 |
| sean-k-mooney | i think we can emulate that by mocking the driver method to get the host name and config flags at least at a unit test level | 15:28 |
| sean-k-mooney | for functional tests we could proably do it too but only with one compute node sicne the config is shared | 15:28 |
| sean-k-mooney | but yes i agree i dont know if we need to test all 4 combination but proably woudl not hurt | 15:29 |
| admin1 | is there a good rescue image (os ) for openstack .. that kind of boots directly without login/pass for rescue ? | 15:30 |
| sean-k-mooney | well by default the same image the vm is booted with is used for rescue | 15:32 |
| sean-k-mooney | but any of the images you woudl use on a phsyical host will be good | 15:33 |
| sean-k-mooney | like KNOPPIX | 15:33 |
| opendevreview | Merged openstack/nova-specs master: Add stable-compute-uuid spec https://review.opendev.org/c/openstack/nova-specs/+/863152 | 15:33 |
| sean-k-mooney | it actully would proably have made sense to allow a keypair to be specifed for rescue | 15:34 |
| sean-k-mooney | although the assumtion was the tenant would do rescue so that why we reuse the one the vm is booted with | 15:35 |
| sean-k-mooney | admin1: if you are doing this on behalf of a tenatn what i would recommend is creating a iamge yourself and baking in a set of ssh keys and then usign that to rescue tenant vms | 15:36 |
| opendevreview | Merged openstack/nova stable/xena: refactor: remove duplicated logic https://review.opendev.org/c/openstack/nova/+/864932 | 17:03 |
| opendevreview | Merged openstack/nova stable/xena: Record SRIOV PF MAC in the binding profile https://review.opendev.org/c/openstack/nova/+/864933 | 17:04 |
| opendevreview | Merged openstack/nova stable/xena: Remove double mocking https://review.opendev.org/c/openstack/nova/+/864934 | 17:04 |
| opendevreview | Merged openstack/nova stable/xena: Remove double mocking... again https://review.opendev.org/c/openstack/nova/+/864935 | 17:04 |
| opendevreview | Merged openstack/nova stable/xena: Add compute restart capability for libvirt func tests https://review.opendev.org/c/openstack/nova/+/864936 | 17:04 |
| opendevreview | Merged openstack/nova stable/xena: enable blocked VDPA move operations https://review.opendev.org/c/openstack/nova/+/864937 | 17:04 |
| sean-k-mooney | :) | 17:56 |
| sean-k-mooney | ok ill update the wallaby ones now | 17:56 |
| melwitt | sean-k-mooney: fyi this is the strange py27 failure on the stable/train backport that I could not figure out what is going on https://storage.bhs.cloud.ovh.net/v1/AUTH_dcaab5e32b234d56b626f72581e3644c/zuul_opendev_logs_80f/866090/1/check/openstack-tox-py27/80f58ef/testr_results.html | 18:02 |
| sean-k-mooney | yep i have that open in a tab somewhere | 18:03 |
| melwitt | ok cool :) | 18:03 |
| sean-k-mooney | im wondering if this is related to a call to super or something like that | 18:04 |
| sean-k-mooney | im not exactly sure where TypeError: __init__() takes at least 2 arguments (1 given) | 18:05 |
| sean-k-mooney | is coming form or if this is a tully the issue | 18:05 |
| melwitt | hmm, I will check | 18:06 |
| sean-k-mooney | my guess is either this is related to the python 3 only verison fo super or we require an adtional positional arg in train for a class | 18:09 |
| sean-k-mooney | the one arg that was given is presumabley self | 18:09 |
| sean-k-mooney | althugh it could be the context | 18:10 |
| melwitt | I feel like it has to be something to do with creation of the ClientException objects in the unit tests, there's really not much in the patch | 18:13 |
| melwitt | *not much else | 18:14 |
| sean-k-mooney | whats weired is it only fails on py27 | 18:14 |
| sean-k-mooney | i expect most or all of that to not care | 18:14 |
| melwitt | yeah, same. so weird | 18:15 |
| -opendevstatus- NOTICE: The lists.opendev.org and lists.zuul-ci.org sites will be offline briefly for migration to a new server | 20:01 | |
| zigo | Hi there! About https://bugs.debian.org/1025110 to me it feels like novaclient is calling suparser.addparser twice with the same subparser name, which now raises an error in Py 3.11. | 21:01 |
| zigo | Though I'm having trouble finding the 2 places ... | 21:02 |
| zigo | (I'm not familliar with novaclient...) | 21:02 |
| zigo | Can someone have a look? | 21:02 |
| zigo | In novaclient/tests/unit/test_shell.py, I can see shell._find_actions() called twice with fake_actions_module as argument ... | 21:04 |
| zigo | That's both in test_load_versioned_actions() and test_load_actions_with_versioned_args(). | 21:05 |
| zigo | So this looks like a test issue, not a novaclient issue. | 21:06 |
| zigo | https://bugs.launchpad.net/python-novaclient/+bug/1998856 | 21:10 |
| *** dasm is now known as dasm|off | 22:51 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!