14:00:15 <rosmaita> #startmeeting cinder
14:00:15 <opendevmeet> Meeting started Wed Jan  8 14:00:15 2025 UTC and is due to finish in 60 minutes.  The chair is rosmaita. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:00:15 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:00:15 <opendevmeet> The meeting name has been set to 'cinder'
14:00:24 <jungleboyj> o/
14:00:25 <rosmaita> #topic roll call
14:00:26 <whoami-rajat> Hi
14:00:33 <jungleboyj> Happy New Year!
14:00:35 <flelain> Hello everyone!
14:00:44 <flelain> Happy New Year!
14:00:47 <andre> how can I joing the meeting? in order to watch it?
14:00:48 <whoami-rajat> happy new year!
14:01:00 <andre> happy new year =)
14:01:08 <eharney> hi
14:01:10 <sp-bmilanov> Happy New Year! :)
14:01:12 <abishop> o/
14:01:42 <abishop> andre: you've joined the meeting by being here!
14:01:52 <jhorstmann> o/
14:02:30 <flelain> andre: weird the first time, isn't it?! But pretty efficient you'll see :)
14:02:43 <akawai> o/
14:02:51 <ccokeke[m]> O/
14:02:57 <andre> oh previously it was video call
14:03:26 <abishop> occasionally video, but typically just irc
14:03:57 <rosmaita> usually video is the last meeting of each month
14:04:06 <rosmaita> hello everyone
14:04:23 <rosmaita> jon will be delayed, we are having bad weather here and the schools have a delayed opening
14:04:39 <rosmaita> jungleboyj will laugh but it is very cold here at -7C
14:05:08 <rosmaita> here's a link to the agenda
14:05:17 <rosmaita> #link https://etherpad.opendev.org/p/cinder-epoxy-meetings
14:05:29 <rosmaita> #topic announcements
14:05:35 <jungleboyj> Lol.  -16 C here.  :-)
14:05:49 <jungleboyj> But -7 is cold for you guys.  :-)
14:05:57 <rosmaita> yes indeed!
14:06:13 <rosmaita> been that way for a few days now
14:06:21 <jungleboyj> Unfortunately, I am getting old and don't handle the cold as well as I used to.
14:06:38 <rosmaita> :D
14:06:47 <rosmaita> ok, that's the weather report
14:07:05 <rosmaita> not on the agenda, but welcome back everyone
14:07:12 <hemna> yough
14:07:16 * whoami-rajat shivering in 7 degrees
14:07:22 <rosmaita> this week is Epoxy Milestone 2
14:07:28 <jungleboyj> lol
14:07:56 <hemna> only 21F here.  so...hot
14:07:58 <rosmaita> we're supposed to release os-brick so that other teams can start working with any Epoxy changes we have
14:08:30 <rosmaita> hemna: 18F here last i checked, which is weird since i am farther south than you
14:08:46 <rosmaita> weird for the northern hemisphere, i guess
14:08:48 <hemna> but you are in the "mountains"
14:08:54 <vdhakad> hi
14:08:56 <rosmaita> good point
14:08:57 * sp-bmilanov has a question about the os-brick release, but that's in the agenda in the reviews section
14:09:49 <rosmaita> sp-bmilanov: yes, that was one of the changes i think we should get in
14:09:52 <Vivek_> Hi
14:10:10 <rosmaita> ok, so my point about os-brick release is that no functional changes have merged since the last release
14:10:28 <rosmaita> which would make a release right now kind of pointless
14:10:47 <whoami-rajat> this is an important change required by the dell nvme-fc driver
14:10:48 <rosmaita> so i would like to identify a few key patches that it makes sense to get in there
14:10:50 <whoami-rajat> #link https://review.opendev.org/c/openstack/os-brick/+/923345
14:10:59 <rosmaita> whoami-rajat: thanks
14:11:19 <rosmaita> ok, lets make a list on the agenda starting at line 75
14:12:10 <MengyangZhang[m]> hello, is there a link to the cinder meeting? This is my first time joining
14:12:38 <Luzi> o/
14:13:02 <rosmaita> MengyangZhang[m]: you are in the cinder meeting!  here is a link to the agenda: https://etherpad.opendev.org/p/cinder-epoxy-meetings
14:13:28 <rosmaita> whoami-rajat: there seems to be controversy over https://review.opendev.org/c/openstack/os-brick/+/923345 , or am i looking at the wrong patch?
14:14:18 <whoami-rajat> rosmaita, simondodsley can correct me but my understanding is, Simon is asking for FCZM support and the author just wants to add basic NVME-FC support
14:15:49 <whoami-rajat> not sure what changes are required in the connector to support it, i thought FCZM is a driver specific thing but i could be wrong
14:16:08 <MengyangZhang[m]> rosmaita: have we talked about my proposal of Add Burst Length Support to Cinder QoS?
14:16:28 <rosmaita> MengyangZhang[m]: not yet
14:17:12 <whoami-rajat> also the NVMe-FC support is crucial for other vendors to enable support for NVMe over FC, currently we only support -TCP and -RDMA but FC seems to be an important alternative to support
14:17:27 <whoami-rajat> hence the patch is important to get in IMO
14:17:35 <vdhakad> whoami-rajat +1
14:18:19 <rosmaita> well, what concerns me is this comment from simon: "These changes take no account of FCZM required items like initiator_target_maps which
14:18:19 <rosmaita> makes it very difficult, if not impossible, for other vendors who do support FCZM to implement NVMe-FC based on the
14:18:19 <rosmaita> os-brick changes implemented."
14:18:44 <abishop> does the existing patch provide a partial implemenation that can be enhances later with FCZM? Or would the existing patch need to be significantly reworked?
14:21:37 <rosmaita> abishop: this comment makes it sound like it could be added fairly easily:
14:21:37 <rosmaita> https://review.opendev.org/c/openstack/os-brick/+/923345/comment/c020d040_4eea715d/
14:21:49 <whoami-rajat> i think it provides the basic infra to connect to NVMe devices via FC, surely dell has tested it with their new driver so apart from FCZM changes, it should work
14:22:40 <flelain> Could we get another vendor to test it out and shares his opinion?
14:22:50 <whoami-rajat> i missed that comment but it does confirm my understanding of the patch
14:23:13 <whoami-rajat> flelain, currently no vendor has nvme-fc driver to test it apart from dell
14:24:41 <whoami-rajat> also to vendors, does anyone use the Cinder zone manager or the zoning is configured outside of openstack deployment?
14:24:46 <rosmaita> my worry at the moment is that the patch has been sitting for 6 weeks; the community should decide whether to accept it as is, or require the changes
14:25:01 <rosmaita> hemna: iirc, you were a fczm user at some point?
14:25:06 <whoami-rajat> personally, I haven't seen any deployment using the zone manager
14:25:24 <rosmaita> i haven't either
14:26:59 <jungleboyj> Been a long time since I have heard it mentioned.
14:27:17 <abishop> if fczm can be added later without significantly impacting the current code then I feel we should move forward with the current patch
14:27:27 <rosmaita> abishop: ++
14:27:36 <abishop> "better is the enemy of good"
14:27:53 <jungleboyj> __
14:28:01 <jungleboyj> Ooops.  ++
14:28:59 <whoami-rajat> I agree with abishop
14:29:06 <whoami-rajat> i can take a look at that patch
14:29:14 <flelain> same expression in French abishop lol ++
14:29:36 <rosmaita> ok, great, and i will make sure jbernard is aware of this discussion as well
14:30:21 <rosmaita> ok, sorry that took so much time ... any other key os-brick patches ?
14:31:50 <rosmaita> ok, thanks everyone
14:32:06 <rosmaita> #topic add burst length support to cinder qos
14:32:13 <rosmaita> MengyangZhang[m]: i think that is you?
14:32:44 <MengyangZhang[m]> yes
14:34:12 <himanshu> hi
14:34:39 <MengyangZhang[m]> The change is really on nova side and I have created a nova spec https://review.opendev.org/c/openstack/nova-specs/+/932653. It was reviewd by nova team but since it is a cross project effort, they would like to hear opinions on cinder side first.
14:35:41 <rosmaita> my understanding is that on the cinder side, we just pass the new keys/values along as we currently do, so not much to do on our side
14:35:52 <MengyangZhang[m]> Correct
14:36:05 <rosmaita> but the nova team doesn't want to make changes unless they know that we will be passing them the new keys and values!
14:36:49 <MengyangZhang[m]> Correct
14:37:37 <rosmaita> MengyangZhang[m]: do you have a cinder spec in addition to the blueprint you linked?
14:37:41 <MengyangZhang[m]> And nova just pass them to libvirt really. The burst length parameters are already supported in libvirt
14:37:44 <whoami-rajat> don't we have a restricted set of QoS settings that we allow? or any variable can be set as QoS?
14:38:14 <rosmaita> i think it's restricted, probably by the request schema
14:38:15 <abishop> it sounds like the question is whether cinder needs to add support for new key (names)
14:38:22 <MengyangZhang[m]> rosmaita: I don't since I wasn't sure what the code change would be on cinder side
14:38:50 <rosmaita> MengyangZhang[m]: ok, basically, that's what whoami-rajat and abishop are asking, whether we need to do anything
14:39:25 <whoami-rajat> also is it just related to front-end QoS or it impacts back-end QoS as well?
14:41:44 <MengyangZhang[m]> It's just related to front-end policy and I don't think there's a restriction for the keys. We have already implemented a nova patch to support setting burst length parameters in my company. The burst length parameters are just stored in connection_info field and are passed to nova.
14:43:12 <rosmaita> i think you are correct about no restriction for key names
14:43:29 <abishop> I think the nova team just needs confirmation that no cinder changes are required in order to support the new keys
14:43:50 <whoami-rajat> MengyangZhang[m], which cinder driver are you testing this with?
14:45:20 <MengyangZhang[m]> 8.3.0 i think
14:46:09 <whoami-rajat> MengyangZhang[m], i mean the backend driver, RBD, netapp iscsi/fc, pure etc?
14:48:17 <whoami-rajat> i don't see any restriction for adding front-end qos specs to connection info
14:48:19 <whoami-rajat> #link https://github.com/openstack/cinder/blob/master/cinder/volume/manager.py#L1853-L1854
14:48:31 <whoami-rajat> so this should work
14:48:50 <rosmaita> MengyangZhang[m]: you didn't need to patch cinder, did you?
14:49:04 <MengyangZhang[m]> I need to patch nova
14:49:10 <whoami-rajat> though it's still good to document the new options on cinder side
14:49:45 <rosmaita> i agree with whoami-rajat, looks like no cinder changes are needed other than documentation improvement
14:50:32 <MengyangZhang[m]> * patch nova since the latest version doesn't support passing the burst length parameters to libvirt. The relevant code is here https://github.com/openstack/nova/blob/23733b7f9b0e0db774b7eccc21daa604c01b69da/nova/virt/libvirt/volume/volume.py#L65
14:50:35 <whoami-rajat> yep, doesn't seem to be restricted by schema either
14:50:37 <whoami-rajat> #link https://github.com/openstack/cinder/blob/master/cinder/api/schemas/qos_specs.py#L19-L37
14:51:34 <rosmaita> ok, so to summarize:
14:51:45 <rosmaita> 1. no cinder spec required, because no changes needed
14:52:07 <rosmaita> 2. whoami-rajat and rosmaita will leave comments on the nova spec saying that cinder is ok with the change
14:52:33 <rosmaita> 3. someone should take a look at improving the cinder docs around qos specs at some point
14:52:40 <rosmaita> -- end ---
14:53:01 <rosmaita> sound good?
14:53:22 <jungleboyj> Sounds like a good plan to me.  :-)
14:53:29 <abishop> ++
14:53:57 <rosmaita> MengyangZhang[m]: any questions?
14:54:00 <whoami-rajat> one thing i forgot to mention
14:54:13 <andre> I have a question, will there be any drivers deleted during this cycle?
14:54:16 <MengyangZhang[m]> That's all for me. Thanks for the review!!
14:54:25 <rosmaita> whoami-rajat: what did you forget?
14:54:27 <whoami-rajat> if there are per GB options in the new spec, then it does require cinder change https://github.com/openstack/cinder/blob/master/cinder/volume/manager.py#L1841-L1851
14:55:18 <whoami-rajat> MengyangZhang[m], can you confirm if that's the case or not? ^
14:55:50 <MengyangZhang[m]> We don't touch the per GB options
14:56:09 <MengyangZhang[m]> and this is to set the burst length
14:56:29 <whoami-rajat> ack, then we are good
14:56:51 <rosmaita> great!
14:57:19 <MengyangZhang[m]> Thank you all!
14:57:32 <andre> I have a question, will there be any drivers deleted during this cycle?
14:58:12 <rosmaita> not sure, that would be a question for the PTL ... we do have several deprecated drivers, though, that would be eligible to be removed
14:58:48 <rosmaita> and we probably should take a look at the CI results and deprecate some more drivers that aren't responding
14:59:08 <rosmaita> #link http://cinderstats.ivehearditbothways.com/
14:59:09 <whoami-rajat> rosmaita, regarding the os-brick release, isn't it during R-6 (Feb 17 - Feb 21) for the non-client release
14:59:22 <Luzi> I just want to give a heads up, mhen and i each have maybe 2 hours/week from now on to help working on image encryption and volume type metadata - so we will be able to address smaller review comments, but no big changes.
14:59:36 <andre> when will it happen? My CI is broken so I would like to know how much time I have to fix it
14:59:59 <rosmaita> whoami-rajat: yes, that's the Epoxy release ... the one tomorrow is just so that other projects that use os-brick don't get surprises
15:00:10 <rosmaita> Luzi: thanks, noted
15:00:20 <rosmaita> andre: looking
15:00:39 <whoami-rajat> rosmaita, ack, thanks
15:00:56 <whoami-rajat> also not sure about the spec deadline but projects keep it at M-2
15:01:23 <rosmaita> yes, nova spec freeze is this week
15:01:33 <abishop> andre: I think your driver is safe because we know you're working on fixing CI. It's the abandoned drivers that we worry about
15:01:57 <rosmaita> andre: which driver are you working on ?
15:02:01 <whoami-rajat> asking because I've a cinder spec up for review :D
15:02:02 <whoami-rajat> #link https://review.opendev.org/c/openstack/cinder-specs/+/931581
15:02:34 <andre> rosmaita: https://wiki.openstack.org/wiki/ThirdPartySystems/Open-E_CI
15:02:52 <abishop> #link https://review.opendev.org/c/openstack/cinder-specs/+/937808 is mine
15:03:33 <rosmaita> andre: ok, what abishop says is correct, plus we will give advance notice, probably at the midcycle meeting
15:03:42 <andre> abishop: yes, but things can happen, it is better not to leave to 'chance'
15:04:00 <andre> rosmaita: thanks, I will work on figuring things out
15:04:07 <rosmaita> sounds good!
15:04:34 <rosmaita> we are over time ... thanks for attending everyone, please prioritize reviewing specs (since i think the freeze is this week for cinder too)
15:04:49 <rosmaita> and don't forget the list of review requests on the agenda
15:04:50 <sp-bmilanov> andre: I don't see you in #openstack-cinder, let me know if you have issues with your CI
15:05:06 <rosmaita> sp-bmilanov: thanks for volunteering!
15:05:39 <sp-bmilanov> rosmaita: :)
15:05:39 <rosmaita> andre: #openstack-cinder is the channel to discuss cinder outside of meeting times
15:05:50 <rosmaita> #endmeeting