14:00:53 <abhishekk> #startmeeting glance 14:00:53 <opendevmeet> Meeting started Thu Apr 21 14:00:53 2022 UTC and is due to finish in 60 minutes. The chair is abhishekk. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:00:53 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:00:53 <opendevmeet> The meeting name has been set to 'glance' 14:01:08 <abhishekk> #topic roll call 14:01:14 <abhishekk> #link https://etherpad.openstack.org/p/glance-team-meeting-agenda 14:01:15 <dansmith> p/ (yay DST) 14:01:17 <abhishekk> o/ 14:01:26 <rosmaita> o/ 14:01:27 <mrjoshi> o/ 14:01:42 <abhishekk> pranali and cyril are not around 14:01:59 <abhishekk> just wait couple of minutes more before we start 14:02:15 <jokke_> o/ 14:02:17 <pslestang> o/ 14:02:19 <alistarle> o/ 14:02:28 <abhishekk> lets start 14:02:41 <abhishekk> #topic release/periodic jobs update 14:02:48 <abhishekk> Milestone 1 is 4 weeks away 14:02:54 <abhishekk> most of the specs are up for review 14:03:05 <abhishekk> we will walk through them in next section 14:03:27 <abhishekk> Periodic job, all is green except newly added fips job which has failed 14:03:45 <dansmith> I thought it was working? 14:04:21 <abhishekk> it is failing intermittently 14:04:36 <abhishekk> current execution failed while the previous one has passed 14:04:38 <dansmith> okay, do you have a link handy to a failed run? 14:04:46 <dansmith> if not I'll go dig it up 14:05:02 <abhishekk> https://zuul.opendev.org/t/openstack/build/88b4b25744234bc2ad89cb97525970a7 14:05:07 <dansmith> thanks 14:05:12 <abhishekk> np 14:05:16 <abhishekk> moving ahead 14:05:24 <dansmith> more volume fails ugh 14:05:30 <abhishekk> yes 14:05:35 <abhishekk> #topic Specs for review 14:05:58 <abhishekk> I have added 1st draft for API for Instant caching of an Image 14:06:09 <abhishekk> #link https://review.opendev.org/c/openstack/glance-specs/+/838642 14:06:29 <abhishekk> Expanding store details - https://review.opendev.org/c/openstack/glance-specs/+/835606 14:06:29 <abhishekk> Delete API for metadef resource types - https://review.opendev.org/c/openstack/glance-specs/+/818192 14:06:29 <abhishekk> spec-lite: ability to purge all rows by glance-manage - https://review.opendev.org/c/openstack/glance-specs/+/819622 (one +2) 14:06:36 <dansmith> ah cool I need to look at that 14:07:02 <abhishekk> Last one is we agreed on last PTG but it is moved to this cycle 14:07:25 <abhishekk> So all members, kindly review the spec and give your suggestions 14:07:35 <abhishekk> moving to next topic 14:07:47 <abhishekk> #topic Secure RBAC 14:08:07 <abhishekk> We are yet to get updates from Policy Popup team about our queries 14:08:36 <abhishekk> meanwhile taking reference of rosmaita's work, pranali has prepared a glance policy matrix for us 14:08:44 <abhishekk> #link https://review.opendev.org/c/openstack/glance/+/838857 14:09:06 <abhishekk> Kindly go through it and verify whether we are mapping it correctly or not 14:09:47 <rosmaita> nice 14:09:48 <abhishekk> moving forward; 14:10:06 <abhishekk> slowly we are learning from you :D 14:10:25 <abhishekk> #topic glance-download import method 14:10:35 <abhishekk> #link https://review.opendev.org/c/openstack/glance-specs/+/836132 14:10:55 <abhishekk> pslestang, alistarle There are bunch of suggestions on the spec 14:11:05 <dansmith> I was not aware we were re-discussing this at this meeting or I would have re-reviewed the spec earlier 14:11:23 <rosmaita> yeah, i have not looked at it yet either 14:11:46 <alistarle> Yeah we answer couple of the comment 14:11:47 <pslestang> abhishekk: yes we read them 14:11:54 <alistarle> thanks for the review btw 14:11:55 <abhishekk> sorry for that, this was decided last meeting but dansmith you were not around 14:12:11 <dansmith> ah, okay yeah sorry 14:12:15 <abhishekk> and I failed to update it this week 14:13:20 <abhishekk> alistarle, pslestang summary is only copying data from another glance is not useful 14:14:09 <abhishekk> So if you guys think we can have a virtual discussion for this topic, sometime next week 14:14:52 <alistarle> So you all agreed that copying metadata is a minimal requirement for the spec right ? 14:15:04 <dansmith> as I said before, I'm also happy to put my money where my mouth is and help with getting the metadata bits right, although I think alistarle's comment made it sound like he was pretty on top of it 14:15:21 <dansmith> alistarle: it is, IMHO.. at least hw_* and os_* things 14:15:50 <abhishekk> os_* but not os_glance_*, right? 14:16:02 <dansmith> it was good to get feedback from sean because he does a lot of nova stuff with required traits and numa instances and things that are likely to have those bits set and not work without them 14:16:22 <abhishekk> ++ 14:16:23 <dansmith> abhishekk: yeah, the os_ things that control stuff in nova like whether or not it's a windows image, which requires other things to be turned on in the hypervisor, etc 14:16:41 <abhishekk> ack 14:17:05 <dansmith> alistarle: pslestang: there's not code posted for this yet right? 14:17:16 <pslestang> not right now 14:17:29 <dansmith> okay 14:17:35 <abhishekk> sean has also pointed out some swift example 14:17:38 <pslestang> but we have some code that works that I can push as a WIP 14:17:54 <abhishekk> that will be good to see 14:18:08 <dansmith> abhishekk: the swift temporary url thing, yeah, that's exactly that I was proposing for being able to do this without federation 14:18:08 <pslestang> abhishekk: acked 14:18:39 <pslestang> I think it will overlap jokke_ stuff 14:18:42 <abhishekk> dansmith, yeah, will have a look at it and we can see how feasible it is for us 14:18:47 <pslestang> about import/export 14:18:51 <alistarle> dansmith: yeah we talk a bit about it in the spec, but I think it is far more complicated to achieve 14:19:09 <dansmith> abhishekk: I'm not saying we need to hang up the glance-download stuff on export/temp-url, to be clear 14:19:22 <abhishekk> yep, me neither 14:19:28 <dansmith> alistarle: export you mean? yeah, I'm not saying we hold this up for that 14:20:12 <dansmith> alistarle: but with that sort of temp-url export, all the work you do here will be applicable in that case too, just via a dedicated url instead.. the image show and download will work the same, just no token required 14:20:32 <jokke_> Like I've mentioned before ... the export would really have been useful addition ot this work. Definitely not the other way around 14:20:59 <alistarle> Oh yeah I see, yes we totally agree, but like plestang said it is more a topic related to export work we talk with jokke_ 14:21:26 <alistarle> But we'll rely on it when it will be ready for sure 14:21:36 <dansmith> there's also a little bit of terminology confusion here: 14:21:47 <jokke_> yeah. Also the swift use case was in the initial Interoperable Image Import spec, never implemented as no-one saw real push/value for that 14:22:06 <abhishekk> really? 14:22:11 <dansmith> the temp-url thing I was referring to was more like "take-out" which provides a way to have a non-trusted user access a single image without a user account.. 14:22:25 <dansmith> "export" being more what jokke_ was talking about where we bundle image and metadata in one file 14:22:37 <jokke_> yeah, that was basically Rackspace and once they moved away from active development, no-one else stepped up saying it would be useful 14:22:38 <dansmith> export can be used by a regular user or via the take-out, 14:22:49 <dansmith> and glance-download could use export if not authenticated via federation 14:23:23 <abhishekk> ack 14:23:30 <dansmith> sorry 14:23:37 <dansmith> glance-download could use take-out if not federated 14:23:39 <alistarle> Hmm yeah agree, it is 3 different topic indeed, but who can be interconnected 14:23:42 <dansmith> man, I need to write it down somewhere :P 14:23:55 <pslestang> dansmith: ok agree with that and thanks for your clarification 14:23:56 <abhishekk> definitely 14:24:12 <dansmith> alistarle: I can write up a summary of the three not on the spot in irc if you want, 14:24:23 <dansmith> and we can refer to that when we get confused again, if it would help 14:24:27 <dansmith> just as a sort of potential road map 14:24:30 <abhishekk> ++ 14:24:56 <alistarle> Yep, especially if we want to implement them in following cycle 14:25:11 <abhishekk> Also I think we can have one follow up virtual session to finalize the work items 14:25:14 <dansmith> yeah, if nothing else, just a "if we're going to do this, we should do it this way" sort of thing 14:25:23 <dansmith> like jokke_'s draft export spec that isn't planned to be done right away 14:26:17 <abhishekk> makes sense 14:26:23 <alistarle> Great, so what is the status for this glance-download spec then ? 14:26:57 <jokke_> I think abhishekk had plan to get it worked on once the spec is good to go (so yeah in that light within this group, not in next few cycles) 14:27:10 <dansmith> I think there's a lot of comments in there to address, put in more detail about the namspaces we're going to copy, behavior of if the format is a mismatch, 14:27:24 <dansmith> and I think it would be good to document what the client behavior is going to be, as sean brought up confusion there 14:27:33 <dansmith> alistarle: ^ 14:28:15 <abhishekk> jokke_, I will be happy to work on it, if the approach is finalized before Milestone 2 14:28:15 <alistarle> Ok so let's talk about this metadata, because it was not clear in the PTG, and we needed to talk about it here 14:28:53 <abhishekk> go ahead 14:29:12 <jokke_> yeah, I think there is lots of mixup happening in the spec comments due to that too 14:30:24 <alistarle> last status during the PTG was to not include the metadata at all, to keep it simple as a first step 14:30:35 <jokke_> ++ 14:31:03 <alistarle> And in the spec it seems we need to include them, so the question are: 14:31:03 <dansmith> that's not how we left it when I was there... 14:31:25 <abhishekk> alistarle, that was not the decision, I said I will decide whether we should include hw_ properties or not in later week after PTG 14:32:27 <alistarle> abhishekk: yes you're right 14:33:07 <abhishekk> yeah, and looking at comments from nova team, just copying the data is not enough and not a good user experience 14:33:18 <dansmith> since we had that discussion I've done some more (but not a comprehensive) survey of the critical properties, and hw_ os_ and probably trait: are pretty critical to copy I think 14:33:46 <alistarle> If we need to include the hw_ properties, do you think we need to include it inside the glance-download plugin ? make it configurable in the json ? or enforce which metadata we copy in the glance configuration ? 14:34:31 <dansmith> I think always copying those namespaces is fine, no config or option in the request is fine and the simplest to implement 14:34:41 <jokke_> I think the comments on the spec tell quite clear story how confusing the metadata part makes this whole feature. And also it being extended on every round of discussion making it even more mess in everyone's heads 14:34:49 <dansmith> as noted I would *like* to also take a list of other namespaces to copy, but not critical like hw_ os_ etc 14:35:57 <alistarle> dansmith: we also need to update the container_format and disk_format to match the source image right ? 14:36:03 <jokke_> and no, it definitely should not be done by the internal plugin that does the data transfer 14:36:04 <dansmith> there's a lot of back and forth in the spec, but I don't think I see any arguing or confusion over the metadata like hw_* being important to copy 14:36:18 <dansmith> alistarle: update or reject, I'm not sure I have a strong opinion either way 14:36:57 <abhishekk> I think update should be the right approach 14:37:04 <alistarle> dansmith: we already do it in the image conversion plugin, so I think update is OK, and we can even re-update if needed later in the image conversion plugin 14:37:06 <dansmith> alistarle: sounded like you preferred update, which is fine with me.. definitely easier for the user 14:37:35 <dansmith> alistarle: it also means the client can just create the image with any value for those, and let the plugin override them so they're correct, which is much nicer for the user 14:37:47 <dansmith> like, bare/raw and let it be updated to match 14:38:15 <alistarle> Yep, that looks good to me 14:38:44 <dansmith> cool 14:38:53 <pslestang> and do we add an option to let the customer add his own metadata? 14:39:12 <dansmith> pslestang: add his own keys or his own namespaces? 14:39:43 <alistarle> it can be a regex of keys 14:39:52 <alistarle> so it match keys and namespaces 14:39:58 <jokke_> pslestang: no 14:40:22 <dansmith> alistarle: sure that works... 14:40:47 <abhishekk> why no, any specific reason? 14:42:06 <pslestang> jokke_: I defnitely think that we should rely on your export spec, but for the first version of glance-download we need to have a solution that is ok for everyone 14:43:02 <jokke_> I think we should either consistently let the user to take care of their metadata or import all we can. Expecting user to write regexp as input to client is turning this same mess the old tasks api was already where we expected json as input from user and that got abandoned as whole due to that 14:43:58 <alistarle> Ok so let's hardcode the namespace and that's it, right ? 14:44:08 <dansmith> we're not expecting a user to write a regex, 14:44:21 <dansmith> we're allowing them to if they have other keys to copy, other than the ones we know need to be copied 14:44:26 <jokke_> pslestang: remember that this is stable API we're talking about. We do it one way now, we're stuck with it. We don't break the api consistently and excuse that to microversions 14:45:55 <pslestang> jokke_: yeah understood 14:46:10 <alistarle> jokke: I agree we you, we need to take care of the API, so if we are not sure of the solution, let's not touch it 14:46:56 <alistarle> And as we agree the definitive solution will be related to export stuff or other patch, I propose to hardcode it, so it is easier to move in future patch 14:47:45 <dansmith> alistarle: I think there's a majority of people here who agree on most of the elements of a direction, so I'd say revise the spec to get us closer to that, address/resolve all the existing comments and we can go from there 14:48:02 <abhishekk> +1 14:48:24 <dansmith> it'll be easier to iterate and discuss a concrete thing than a thing plus a diff of a bunch of comments and discussion 14:48:39 <abhishekk> and also if you think you are stuck in some part feel free to ping us 14:49:19 <alistarle> great, clear for me 14:49:38 <abhishekk> cool 14:49:55 <abhishekk> moving to open discussion 14:50:00 <abhishekk> #topic Open discussion 14:50:19 <abhishekk> #link https://review.opendev.org/c/openstack/glance/+/838822 14:50:24 <abhishekk> Update migration constant 14:50:45 <abhishekk> Kindly review the patch, easy one and important one 14:51:27 <abhishekk> Also I think pranali has forget to include export draft in review list 14:51:42 <abhishekk> #link https://review.opendev.org/c/openstack/glance-specs/+/837127 14:51:48 <jokke_> abhishekk: is that ^^ the alembic thingie we need to do every cycle? 14:51:55 <abhishekk> jokke_, yep 14:52:48 <jokke_> gating 14:52:54 <abhishekk> cool, thank you 14:53:15 <abhishekk> dansmith, I am also not aware but looks like alembic migration depends on that 14:53:31 <dansmith> abhishekk: I don't think the other projects have to do this sort of thing 14:53:42 <jokke_> it's part of how the migration scripts gets picked up iirc 14:53:56 <abhishekk> they might have implemented the script execution differently 14:54:25 <abhishekk> Something in our glance-db-manage needs to be fixed 14:54:33 <dansmith> abhishekk: clearly they did, I'm just saying it seems like maybe something we could resolve 14:54:39 <jokke_> I think glance was the first one that got implemented and some of the inconveniences were quite quickly picked up and done differently elsewhere 14:54:46 <abhishekk> ++ 14:54:58 <abhishekk> dansmith, agree 14:55:33 <abhishekk> Nothing else form me 14:55:40 <abhishekk> last 5 minutes 14:55:51 <jokke_> I think it's one of the things that 10min of work every year has not yet warranted anyone dig deep enough into it to change the beaviour :P 14:56:04 <dansmith> right :) 14:56:46 <abhishekk> :D, lets change it this cycle 14:57:17 <abhishekk> anything else guys? 14:57:33 <jokke_> I'm pretty sure you could tie it to pbr somehow 14:57:37 <jokke_> not from me 14:58:03 <abhishekk> cool, lets wrap up 14:58:06 <abhishekk> thank you all! 14:58:12 <abhishekk> have a nice weekend o/ 14:58:34 <abhishekk> #endmeeting