| *** ralonsoh_ is now known as ralonsoh | 08:10 | |
| opendevreview | Rajesh Tailor proposed openstack/nova master: Add support for showing image properties in server show response https://review.opendev.org/c/openstack/nova/+/939649 | 08:22 |
|---|---|---|
| mikal | melwitt: I only just saw your comments on https://review.opendev.org/c/openstack/nova/+/924844 -- I'll work through them tomorrow. | 09:15 |
| andrewbonney | Morning. If anyone has time I'd appreciate a review of the bug fix in https://review.opendev.org/c/openstack/nova/+/919961 | 09:51 |
| opendevreview | Michael Still proposed openstack/nova master: libvirt: direct SPICE console object changes https://review.opendev.org/c/openstack/nova/+/926876 | 10:03 |
| opendevreview | Michael Still proposed openstack/nova master: libvirt: direct SPICE console database changes https://review.opendev.org/c/openstack/nova/+/926877 | 10:03 |
| opendevreview | Michael Still proposed openstack/nova master: libvirt: allow direct SPICE connections to qemu https://review.opendev.org/c/openstack/nova/+/924844 | 10:03 |
| opendevreview | Michael Still proposed openstack/nova master: libvirt: Add extra spec for sound device. https://review.opendev.org/c/openstack/nova/+/926126 | 10:03 |
| opendevreview | Michael Still proposed openstack/nova master: libvirt: Add extra specs for USB redirection. https://review.opendev.org/c/openstack/nova/+/927354 | 10:03 |
| opendevreview | Michael Still proposed openstack/nova master: Protect older compute managers from sound model requests. https://review.opendev.org/c/openstack/nova/+/940770 | 10:03 |
| mikal | sean-k-mooney: https://review.opendev.org/c/openstack/nova/+/940770 is a first attempt at the compute service version bump thing as discussed this morning. | 10:03 |
| *** ralonsoh_ is now known as ralonsoh | 10:31 | |
| opendevreview | Balazs Gibizer proposed openstack/nova master: Bump InstanceNUMACell version after data migration https://review.opendev.org/c/openstack/nova/+/940604 | 14:20 |
| gibi | sean-k-mooney: dansmith: add the extra tests cases and extended the logic to handle the half migrated cases ^^ | 14:25 |
| MengyangZhang[m] | <sean-k-mooney> "Mengyang Zhang: i suspect the..." <- Since Libvirt introduced these burst parameters in 2.4.0, which was released in 2016, do we still have to worry about this? Nova needs to support n-2 versions, but all versions should already support this burst feature. | 14:35 |
| dansmith | gibi: personally I'd prefer that as two patches, even though I know it's more backport work, but I'll defer to the others | 14:39 |
| gibi | OK I can split it | 14:42 |
| dansmith | gibi: wait for the others to opine, I'm just saying that's _my_ opinion | 14:48 |
| sean-k-mooney | MengyangZhang[m]: its not about the libvirt verion. its about the nova compute version deployed on each host | 14:48 |
| sean-k-mooney | MengyangZhang[m]:in 2025.1 we need to supprot a mix of comptue with (2024.1, 2024.2 and 2025.1) because its a skip level upgrade release | 14:49 |
| sean-k-mooney | in 2025.2 we only need to suprot a mix of 2025.1 and 2025.2 nova computes | 14:49 |
| sean-k-mooney | we need to make sure cold/live migration works acorss those mixed versions | 14:49 |
| sean-k-mooney | so we need a min compute service version check to prevent using the feature until the cloud is fully upgraded or explcit schduler supprot to make sure the isntance only get deploy to the upgraded nodes | 14:50 |
| MengyangZhang[m] | understood, make sense | 14:57 |
| gibi | dansmith: understood | 14:58 |
| sean-k-mooney | the min compute service check is simpler todo, in either case cinder need to be prepared for nova to reject the attachment of a volume with brust qos if the compute node is too old. i belive that is already the case however so that does not really change the existing workflow | 14:59 |
| *** ralonsoh_ is now known as ralonsoh | 15:15 | |
| opendevreview | Sylvain Bauza proposed openstack/nova master: Add a new ImagePropertiesWeigher https://review.opendev.org/c/openstack/nova/+/940642 | 16:58 |
| gmann | bauzas: this is ready, rbac job cleanup we talked yesterday https://review.opendev.org/c/openstack/nova/+/940737 | 17:07 |
| gibi | there is an easy review (just adding more spice to the nova-next job) https://review.opendev.org/c/openstack/nova/+/937275 | 17:08 |
| bauzas | gibi: gmann: reviewing your meats | 17:11 |
| gibi | thanks folks | 17:15 |
| gmann | gibi: if you are around, please chekc this https://review.opendev.org/c/openstack/nova/+/940737 | 17:44 |
| gmann | gibi: bauzas and very trivial typo one https://review.opendev.org/c/openstack/nova/+/937744 | 17:45 |
| gibi | done and done :) | 17:48 |
| gmann | thanks guys | 17:50 |
| mikal | Morning. | 18:34 |
| mikal | sean-k-mooney: I'm taking a look at tempest just now based on melwitt's comments on https://review.opendev.org/c/openstack/nova/+/924844/21. Is it just me or are there now SPICE tempest tests at all? https://github.com/openstack/tempest/blob/master/tempest/config.py#L541 | 18:48 |
| mikal | s/now/no/ | 18:48 |
| melwitt | mikal: there are not and when I looked it looked like they were removed in the past https://github.com/openstack/tempest/commit/d5bda712599a329acb865f9a9a41ad55824b1781 I don't know if it | 18:49 |
| mikal | Hmmm, I wonder why the test cases were removed. | 18:50 |
| melwitt | *if it's realistic and/or reasonable to test spice there but it was something I was curious about | 18:50 |
| dansmith | yeah, the justification for removal in those seems weird, although I'm not sure if they were testing spice itself or just that we could get a spice url from the api | 18:51 |
| mikal | sean-k-mooney and I have have gone back and forth on this a bit. I would like to add tempest tests, but its hard to test much meaningfully there right now -- we'd be basically testing if a TCP port was open I think. | 18:51 |
| dansmith | we have (or used to have?) a VNC test that would at least try to connect and get the RFB00 banner from the socket to make sure it was at least plausibly running | 18:51 |
| mikal | I think I could write meaningful tests by having a simple python spice client, but I haven't managed to find the time to do that yet and it would take longer than I have before the feature freeze this cycle. | 18:51 |
| melwitt | yeah I was thinking of vnc as well, for example https://github.com/openstack/tempest/blob/a552cb93aea10fd62648f1b8d74b16a7b3e801dc/tempest/api/compute/servers/test_novnc.py#L28 | 18:52 |
| mikal | Yeah, it seems a bit websockety, but it is checking for a header: https://github.com/openstack/tempest/blob/master/tempest/api/compute/servers/test_novnc.py#L195 | 18:53 |
| sean-k-mooney | dansmith: ya that still exits as shown above ^ | 18:54 |
| sean-k-mooney | we dont currently have spice in any of our nova jobs btu we could add it to any of them | 18:54 |
| sean-k-mooney | the issue woudl be testing with kerbside really | 18:55 |
| mikal | Where are the tempest config options set? Somewhere in the nova repo? | 18:55 |
| dansmith | I think the benefit of such a test should not be understated.. there's a lot of plumbing that is verified by merely confirming that banner string | 18:55 |
| sean-k-mooney | my understainding is mikal has developed kolla-ansible supprot | 18:55 |
| sean-k-mooney | but we dont have devstack support | 18:55 |
| mikal | Yeah, I have a prototype deployer for kolla-ansible (prototype because I am fairly sure I got HA wrong but I don't have enough machines in the lab to test right now). | 18:56 |
| sean-k-mooney | dansmith: that testing is really testing the nova-spiceproxy | 18:56 |
| sean-k-mooney | which is unused with spice direct | 18:56 |
| sean-k-mooney | that siad i guess we coudl write the test to conenct directly to spice | 18:56 |
| sean-k-mooney | like kerbside would | 18:56 |
| dansmith | ...I mean, there's still spice at the other end right? | 18:56 |
| sean-k-mooney | that woudl confirm the data we include in teh api is correct | 18:56 |
| melwitt | would spice direct not present a html login page or something like vnc does or? | 18:56 |
| mikal | I can give that a go, but I'd probably need some hand holding. | 18:56 |
| sean-k-mooney | melwitt: the kerbside proxy might | 18:57 |
| sean-k-mooney | but nothign in nova will | 18:57 |
| mikal | melwitt: no, spice-direct just gives you connection details to the TCP ports on the hypervisor. There is then a proxy which sits in front which is not currently proposed in openstack land. | 18:57 |
| melwitt | ok, i see what you're saying now | 18:57 |
| gmann | we do have devstack support ? but it is not tested anywhere i think https://github.com/openstack/devstack/blob/master/lib/nova#L719 | 18:57 |
| sean-k-mooney | so we can test that those conenction detail are right as an admin/service user in tempest | 18:58 |
| mikal | spice-direct by itself meets shadow.tech's needs (I hope), and then I have a whole bunch of other bits to propose later once this first bit lands. | 18:58 |
| sean-k-mooney | gmann: that only for spice with nova-spiceproxy | 18:58 |
| sean-k-mooney | its enouch to test part of it | 18:58 |
| sean-k-mooney | jsut not the kerbside part wich is probaly ok | 18:58 |
| sean-k-mooney | mikal: shadow.tech can actully work with the nova-spiceproxy too | 18:59 |
| sean-k-mooney | they dont actully need spice-dicret they want teh sound and usb supprot your adding at the end | 18:59 |
| sean-k-mooney | they need the usb ports to do rediectoin so gaming periferals to the vm which works over the nova websocket proxy | 19:00 |
| sean-k-mooney | they have there own custom spice client that can work with that | 19:00 |
| sean-k-mooney | spice-direct is proably faster so they might move to that in the future, they currently deploy one nova-spice proxy per compute node to keep latency down | 19:01 |
| mikal | Huh. | 19:01 |
| mikal | Well, I'll go spelunking in tempest today and see what I can work out. | 19:03 |
| mikal | sean-k-mooney: I'm a bit confused about your comments on the sound model patch. I am not sure if I should read them as saying that https://review.opendev.org/c/openstack/nova/+/940770/1 isn't actually needed any more. | 19:04 |
| sean-k-mooney | anyway i think that off topic. the minium useful tempets test would use the new api microverion to get the connecction info kerbside would use adn then check the banner | 19:04 |
| sean-k-mooney | it looks like i didt push my late set of comemtns one sec | 19:04 |
| sean-k-mooney | ok i ment to push them this morning | 19:05 |
| sean-k-mooney | mikal: that is still needed | 19:05 |
| sean-k-mooney | it shoudl be merged into the previous change | 19:05 |
| sean-k-mooney | but the comptue service bump is needed | 19:05 |
| mikal | Yeah, I split it out so it was easier to see what was happening. I can squash it back into the sound patch. It seems you're largely happy, you just want that hardware.py helper used? | 19:07 |
| sean-k-mooney | yep that was my main ask | 19:07 |
| mikal | Ok, so I'll do that and take a look at tempest today then. | 19:08 |
| sean-k-mooney | reuse the helper btu the rest looks ok | 19:08 |
| sean-k-mooney | its unrelate but you ahve a nice performance improment in that too by only doing the min verison check once | 19:09 |
| mikal | Honestly fixing that and the affected unit tests was by far the slowest bit, but I didn't like how we looked up the version over and over. | 19:10 |
| sean-k-mooney | im debating if that should be its own patch because its a ligit performance bug | 19:10 |
| sean-k-mooney | your new check for the sound modle whoudl be with the patch that intoduced it. the refactor of the single lookup could be its own patch but im not going to ask you to split it now. other might but i think its ok to incoperate in a signel patch | 19:12 |
| mikal | That wouldn't be hard to do, but it would nicer if it was before the sound patch. I could split the series into two series? objects-db-impl for spice-direct, and then performance-sound-usb? | 19:13 |
| mikal | That would also make a cherry pick cleaner for shadow.tech if they were that way inclined. | 19:14 |
| mikal | But honestly right now you were right yesterday -- I am mostly focussed on getting those API changes landed, because rebasing them is a huge pain. | 19:15 |
| sean-k-mooney | ya it woudl be nice ot put the perfomance improment as a sepreate patch before the sound patch | 19:15 |
| sean-k-mooney | its up to you | 19:16 |
| mikal | Ok. Well, I am going to focus on the tempest thing and we'll see how far I get and if I have time to rearrange the later patches today or not. | 19:16 |
| sean-k-mooney | ack | 19:16 |
| mikal | Where are those tempest flags set for a nova gate run? I can't see that in the nova repo. Is there a third repo involved or something? | 19:16 |
| sean-k-mooney | we do not have a job in nova ath uses spice as far as i recall | 19:19 |
| sean-k-mooney | so we need to chose one. add the devstack/tempest flag to use spice and have that depend on the tempest change that adds the new test | 19:20 |
| sean-k-mooney | so one patch to tempest to add the new test, and either update patch 3 (the api patch) or add a dedicated ci patch to the nova series | 19:21 |
| mikal | Oh I hadn't thought of that. So if we turn on SPICE we have to turn off VNC and therefore wouldn't be testing that. | 19:21 |
| sean-k-mooney | i think i would be ok with a testing patch at the end of the serise if melwitt and dansmith are ok with that | 19:21 |
| sean-k-mooney | mikal: ya but we have several jobs | 19:21 |
| sean-k-mooney | so we can change it in one of the many jobs teh nova-ovs-hybrid-plug job | 19:22 |
| dansmith | yeah a test at the end is good to me.. even if we can't land it with the rest, being able to see it working (or almost working or whatever) is data for review | 19:22 |
| sean-k-mooney | mikal: i would modify this https://github.com/openstack/nova/blob/master/.zuul.yaml#L144 | 19:23 |
| melwitt | mikal: tempest config are set in https://github.com/openstack/nova/blob/d267ede98f1b0aa25630b1efe07db5a8b478d473/.zuul.yaml#L136 for example | 19:23 |
| mikal | Ok, so the moving parts are: add a test to tempest; add plumbing in devstack to enable that test; fiddle with nova gate jobs to use a devstack with the spice test turned on? | 19:23 |
| sean-k-mooney | we use it to mainly test with ml2/ovs but its a good candiate for non standard configs | 19:23 |
| sean-k-mooney | ya i might be able to help by putting up a poc patch of just enabling spice in that job | 19:24 |
| sean-k-mooney | you coudl modify it and make it depend on your tempest patch when that is ready | 19:25 |
| mikal | Ok, well I'll take a swing at this today but its sounding like more than I'll get done in a single day to be honest. | 19:25 |
| opendevreview | Merged openstack/nova master: [Trivial] Fix the typo error https://review.opendev.org/c/openstack/nova/+/937744 | 19:29 |
| opendevreview | sean mooney proposed openstack/nova master: Disable the heal instance info cache periodic task https://review.opendev.org/c/openstack/nova/+/939476 | 19:50 |
| sean-k-mooney | bauzas: i did a spell check pass on ^ otherwise it the same as the last version | 19:51 |
| opendevreview | Merged openstack/nova master: Cleanup RBAC jobs in check/gate pipeline https://review.opendev.org/c/openstack/nova/+/940737 | 20:25 |
| opendevreview | sean mooney proposed openstack/nova master: [WIP] move nova0ovs-hybrid-plug to deploy with spice https://review.opendev.org/c/openstack/nova/+/940835 | 20:36 |
| opendevreview | Doug Goldstein proposed openstack/nova stable/2024.2: ironic: Fix ConflictException when deleting server https://review.opendev.org/c/openstack/nova/+/940836 | 20:45 |
| sean-k-mooney | mikal: https://review.opendev.org/c/openstack/nova/+/940835 might work but i expect it to fail, if it does i think https://review.opendev.org/c/openstack/devstack/+/940838 might fix it. im going to finish for today but if you want to rebase those patches, addign a depend on and setting NOVA_SPICE_ENABLED later feel free. | 21:02 |
| sean-k-mooney | doing a quick search with hound i dont think we have any job anywhere that enables spice so i dont know if there is anything else broken | 21:03 |
| sean-k-mooney | s/any job/any devstack job/ | 21:04 |
| mikal | sean-k-mooney: thanks, I'll take a look soon. | 22:10 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!