14:03:47 <jbernard> #startmeeting cinder 14:03:47 <opendevmeet> Meeting started Wed May 14 14:03:47 2025 UTC and is due to finish in 60 minutes. The chair is jbernard. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:03:47 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:03:47 <opendevmeet> The meeting name has been set to 'cinder' 14:03:51 <jayaanand_> hi 14:03:53 <simondodsley> o/ 14:03:55 <agalica> o/ 14:03:58 <Luzi> o/ 14:04:02 <jbernard> o/ 14:04:04 <hvlcchao1> o/ 14:04:09 <eharney> hi 14:04:17 <gireesh> o/ 14:04:18 <jbernard> #link https://etherpad.opendev.org/p/cinder-flamingo-meetings 14:04:35 <rosmaita> o/ 14:05:43 <whoami-rajat> hi 14:06:04 <jungleboyj> o/ 14:06:52 <jbernard> welcome everyone, lets get started 14:07:00 <jbernard> rosmaita: you have some annoucements 14:07:03 <jbernard> #topic annoucements 14:07:29 <rosmaita> yeah, looks like maybe possibly the gates will get under control today 14:07:39 <rosmaita> info is on the agenda 14:08:20 <simondodsley> exciting - been waiting for this one 14:08:30 <rosmaita> keep your fingers crossed 14:08:38 <rosmaita> the other CI news is about the NFS CI job 14:08:39 <simondodsley> thanks rosmaita for sorting this out 14:09:11 <rosmaita> tosky has a patch up to add the cinder-tempest-plugin tests to the devstack-plugin-nfs job 14:09:28 <rosmaita> there's a question on his patch about whether the swap size change will work 14:09:35 <Sai> o/ 14:09:44 <rosmaita> so i put up a patch to run the job 10 times 14:09:49 <rosmaita> it's still running 14:10:11 <rosmaita> but that should give us some info ... key thing is, we need to be running the cinder-tempest-plugin tests on NFS 14:10:46 <rosmaita> but hopefully we can get https://review.opendev.org/c/openstack/devstack-plugin-nfs/+/898965 merged soon 14:10:58 <simondodsley> is there a simple way 3rd party CIs can force this to run on our side? 14:11:46 <rosmaita> by "this' do you mean the NFS job, or the cinder-tempest-plugin tests? 14:11:53 <simondodsley> NFS 14:12:31 <simondodsley> I'm happy to run the NFS generic plugin on our CI but need to understand how to configure Zuul so it uses NFS shares I provide 14:12:53 <simondodsley> This would also provide some validation on using the generic driver on 3rd party paltforms 14:13:16 <rosmaita> that would be nice ... but to answer your question, i'm not sure 14:13:16 <eharney> definitely possible, if you have NFS shares, you'd just need to configure the generic driver and its related options in cinder 14:13:45 <rosmaita> what Eric said 14:14:02 <eharney> https://opendev.org/openstack/devstack/src/branch/master/lib/cinder_backends/nfs shows what needs to be set 14:14:43 <simondodsley> gret - I'll look into that 14:15:06 <rosmaita> sounds good! 14:15:10 <rosmaita> that's all from me 14:16:07 <whoami-rajat> I've proposed some patches to better address the new location API issue but it needs attention from glance rather than cinder https://review.opendev.org/q/topic:%22fix-location-apis%22 14:17:42 <rosmaita> whoami-rajat: those would allow us to run our tests with 'do_secure_hash = true" ? 14:18:04 <whoami-rajat> with the tempest change merged, YES 14:18:46 <rosmaita> ok, but for now, we can just happily run with 'do_secure_hash = false' and the ceph job will pass 14:19:14 <whoami-rajat> yep, I also have a patch to disable it in glance since it might cause issues in deployments as well 14:19:56 <rosmaita> that secure hash seems to be a performance hit 14:20:55 <eharney> the test failures aren't because it's a perf issue 14:22:33 <eharney> the test fail b/c the image is in "active" while there are still tasks happening that interfere with other operations 14:23:45 <rosmaita> yeah, but isn't the "tasks still happening" taking up time & cpu ? 14:24:04 <eharney> yes 14:24:06 <rosmaita> i mean, if they were instantaneous, we wouldn't see this 14:24:10 <rosmaita> ok 14:24:36 <whoami-rajat> if we disable hashing then we are not really testing it anywhere 14:24:50 <eharney> in short i think there are design problems that that tempest test exposed, and we need to document those (as a bug) before we start merging workarounds into tempest tests 14:27:57 <whoami-rajat> bug: https://bugs.launchpad.net/glance/+bug/2110581 14:27:57 <whoami-rajat> fix: https://review.opendev.org/c/openstack/glance/+/949668 14:27:57 <whoami-rajat> doc: https://review.opendev.org/c/openstack/glance/+/949672 14:29:21 <eharney> i think there is another bug in that it is unclear what the expected behavior is for an image that's in "active" (and therefore usable) but that still has ongoing hash calculation tasks running -- it seems like it shouldn't be in the active state 14:29:29 <eharney> but we don't have to sort all that out here 14:31:24 <whoami-rajat> it was by design this way but the cases like delete were not considered so turned out to be problematic 14:31:25 <whoami-rajat> yep, let us meet with glance and we can discuss 14:31:39 <eharney> delete is not the scariest part of this design IMO 14:33:54 <jbernard> ok, thanks for getting all of this sorted 14:34:31 <jbernard> the glance meeting is tomorrow, maybe this is a good topic for their agenda 14:35:09 <jbernard> do we have any specific plans to introduce this at that meeting, or are the relevant persons already aware? 14:36:53 <eharney> there still isn't a bug fully describing the problem so i'm not assuming that all the relevant material is being discussed yet 14:38:49 <jbernard> ok, unless someone wants to volunteer, i can jump in there tomorrow and at least raise the issue 14:39:28 <hemna> mep 14:40:47 <jbernard> ok, next on the list 14:41:07 <jbernard> tobias-urdin has an rbd data pool patch up 14:41:18 <jbernard> https://review.opendev.org/c/openstack/cinder/+/914930 14:42:07 <jbernard> asking if it should honor the configured data pool when cloning, or use the one from the original volume 14:42:45 <eharney> in the situation where you've configured a data pool on an existing deployment that didn't have one? 14:43:20 <jbernard> or if the configured data pool doesn't match the data pool of the volume being cloned 14:44:06 <eharney> good question 14:44:09 <jbernard> yeah 14:44:16 <eharney> it's also not obvious to me how this impacts our capacity stats collection 14:44:27 <eharney> because the patch seems to ignore that currently 14:44:52 <jbernard> i lean toward honoring the configuration defined, but ive only been thinking about it for a minute 14:45:19 <jbernard> eharney: that's a good questiion, that patch definately needs some input 14:46:59 <jbernard> eharney: ive added us as reviewers, so i should show up in your queue 14:47:05 <jbernard> s/i/it 14:48:47 <whoami-rajat> I'm missing the use case of the patch, is the purpose to have RBD image data in a separate data pool and the metadata in the volumes pool? 14:49:45 <eharney> yes, it doesn't really spell out the use case very well, but one of the reasons that's desired is to enable use of erasure-coded pools 14:49:56 <jbernard> yep 14:50:41 <whoami-rajat> ok, don't we also need to document the performance impact of using erasure coded pool? I'm sure the parity calculation should be pretty CPU intensive 14:51:11 <eharney> it will definitely have a perf impact (on the ceph cluster side) 14:52:17 <jbernard> related to tobias-urdin, his microversion patches are all posted and ready for review 14:52:30 <jbernard> https://etherpad.opendev.org/p/cinder-flamingo-reviews#L18 14:52:58 <whoami-rajat> yes, so deployers/tooling needs to configure more resources to the nodes deploying the OSDs hosting the data pool 14:53:21 <whoami-rajat> to me this patch is very abstract given the impact it can have on deployments 14:54:01 <jbernard> whoami-rajat: these would be good review comments ;) 14:54:21 <eharney> it's geared toward someone who already decided they wanted to setup erasure-coded pools, but yeah, more documentation would definitely help 14:55:32 <jbernard> i want to get this one in before the bell, 14:55:48 <jbernard> mhen, Luzi have some cycles to work on image encryption 14:56:00 <jbernard> https://review.opendev.org/q/topic:%22LUKS-image-encryption%22 14:56:03 <whoami-rajat> not for a specific customer but a lot of deployments adapt new features thinking it's suited for their use case without knowing the full context so I'm more inclined towards documenting this better 14:56:35 <Luzi> well starting with next week we will have a few hours per week to work on the image encryption again 14:57:14 <Luzi> it would be nice to finally come to terms with this :D 14:57:16 <jbernard> those patches need review (as do others), if anyone is looking for things in immediate need of attention 14:57:23 <jayaanand_> we have multiple customers looking for in-use NFS volume extension. there is a proposal to fix generic NFS issue related bug https://bugs.launchpad.net/cinder/+bug/1870367. can we implement NetApp driver in similar line. 14:57:28 <jbernard> https://etherpad.opendev.org/p/cinder-flamingo-reviews 14:58:03 <jbernard> eharney: i meant to relay jayaanand_'s quetions from earlier, i think you might know best this answer 14:59:07 <eharney> there is a patch up currently implementing this in the netapp nfs driver, right? 14:59:17 <eharney> https://review.opendev.org/q/topic:%22bp/extend-volume-completion-action%22 14:59:18 <eharney> in there ^ 14:59:51 <jayaanand_> it didn't went pass review 14:59:56 <agamboa> hey rosmaita, we took a look at your comments and addressed them for this merge: https://review.opendev.org/c/openstack/cinder/+/901318/ 15:00:06 <eharney> i'm not sure i understand the question. this area is still being worked on 15:00:15 <rosmaita> agamboa: ok, will take a look 15:00:35 <jayaanand_> https://review.opendev.org/c/openstack/cinder/+/873889 15:01:52 <jayaanand_> also bluprint is re-proposed https://review.opendev.org/c/openstack/cinder-specs/+/922374 15:02:11 <eharney> i think the intent is to complete 873889 and land it 15:02:49 <jayaanand_> ok 15:02:59 <jayaanand_> thank you! 15:03:50 <jbernard> we're just a bit over time, last call everyone 15:04:20 <agalica> I have a quick question if we're in the open questions section 15:04:34 <agalica> If we were to support a new type of storage array which in turn requires a new volume type for one of the required features do we require two patches or just one? 15:04:55 <agalica> My suspicion is 2, but I wanted to confirm since the new storage device requires said feature 15:05:16 <jbernard> agalica: you mean a new volume driver? 15:05:25 <jbernard> agalica: do you have an code up that we can look at? 15:05:56 <agalica> not at this time as it's a future. I mean a storage device that we currently do not support in the driver. We are adding support for it, but it requires a volume type feature as well. 15:06:02 <whoami-rajat> added my comment to the RBD patch 15:06:14 <jbernard> whoami-rajat: thank you! 15:06:22 <tobias-urdin> sorry late to the meeting 15:06:44 <jbernard> tobias-urdin: no worries 15:07:14 <jbernard> agalica: i personally am not sure, i say post the code and we can discuss 15:07:22 <whoami-rajat> agalica, volume types can have driver specific attributes in extra specs so a special volume type should not require code changes but would be better to understand your use case more in detail 15:07:25 <agalica> So, for example, supporting storage type X requires feature Y which has not been implemented. In order to support X, we must support Y. Does this mean that Y requires its own patch and then we can have a patch for storage type X? Or can both of these things happen at once? If it helps, most likely the only changes required will be these new supporting features 15:08:11 <agalica> ok, thanks guys. 15:10:19 <jbernard> ok, lets wrap up, thank you everyone! 15:10:23 <jbernard> #endmeeting