Posted	Nick	Remark
#openstack-nova - 2022-09-01
16:29:31	dansmith	or just use instance lock on the compute node to hold off on any other requests.. I guess that's enough
16:29:32	sean-k-mooney	i think we need to disucss this again next cycle
16:29:53	bauzas	the next PTG will be virtual
16:30:05	bauzas	anyone can join, budged saved
16:30:24	sean-k-mooney	ya but we can reopne the sepc and discuss it again before htat
16:30:31	bauzas	oh yeah
16:30:58	sean-k-mooney	i see to ways we coudl go with the instance action by the way
16:31:05	sean-k-mooney	one is one for regeneratin cofnig drive
16:31:11	sean-k-mooney	the other is more targed for this feature
16:31:21	sean-k-mooney	an instance action to update user_data
16:31:30	sean-k-mooney	that either succeed or does not
16:31:49	sean-k-mooney	and have that be the only wya to update it
16:31:57	sean-k-mooney	so no updating it via server put
16:32:25	dansmith	yeah, the user should not see "regenerate config drive"
16:32:41	dansmith	so it should be "update my user_data" at the external surface for sure
16:32:52	sean-k-mooney	right
16:32:55	sean-k-mooney	so if we do that
16:33:16	sean-k-mooney	that dedicated instance action can "do the right thing" regardess of how the instance is booted
16:33:31	dansmith	right
16:33:42	sean-k-mooney	and returna 409 conflict if its not supprot by the driver or whatever
16:33:55	jhartkopf	sean-k-mooney: Oh that'd be bad news. Was hoping to get at least the stuff without config drives merged this cycle.
16:34:12	sean-k-mooney	but it can be atomic
16:34:40	JayF	o/, I have a handful of ironic driver patches I'm still trying to get backported thru. They are all clean backports and passing tests.
16:34:42	JayF	https://review.opendev.org/c/openstack/nova/+/854257 (no live migration tests on ironic driver changes -> yoga) https://review.opendev.org/c/openstack/nova/+/821351 (ignore plug_viles on ironic driver -> ussuri) https://review.opendev.org/c/openstack/nova/+/853546 (minimize window for a resource provider to be lost -> train)
16:34:42	bauzas	sean-k-mooney: dansmith: yeah, the user-surfaced API is the userdata, not the configdrive
16:35:01	bauzas	so the instance action has to be related to it
16:35:39	sean-k-mooney	JayF: did we land the skip of the migration job yet
16:35:47	JayF	sean-k-mooney: that's the yoga backport above
16:35:52	JayF	sean-k-mooney: the version -> master is in
16:35:55	sean-k-mooney	an cool
16:36:03	sean-k-mooney	i was onder if the master on had landed
16:36:05	bauzas	jhartkopf: the problem is that we can't introduce this userdata update by splitting without configdrive as we just agreed on a brand new API
16:36:24	JayF	oh, I've already checked all these -- previous branches have merged, these should be good to go. e.g. I'd approve these if they were ironic stable changes
16:36:32	bauzas	for consistency reasons, we have to hide the configdrive details from the user request
16:36:43	sean-k-mooney	bauzas: that and my +2 on the spec was contingent on it not depending on if you use configdrive or not
16:36:46	JayF	and elod helped me clean up the only piece around nova stable policy that's more strict than Ironic :D
16:37:33	bauzas	sean-k-mooney: that kind of design problems happening late in the cycle are common
16:38:14	bauzas	jhartkopf: this isn't at all about your implementation, this is rather about the fact we discovered some design problems we weren't considering at the spec review time
16:38:16	gibi	we should be better at doing this kind of design discussion before the FF day
16:38:37	gibi	but I don't know how
16:38:43	bauzas	honestly, this is the only blueprint that fails this for this cycle AFAICT
16:38:51	bauzas	gibi: we had precedents
16:39:07	bauzas	some spec approved and a design question raising late
16:39:08	jhartkopf	bauzas: I understand that, but it's really unfortunate though
16:39:52	bauzas	jhartkopf: I don't disagree with you and I understand how much this can be frustrating
16:40:01	artom	bauzas, wait, so ephemral disk, PCI, and manial will all make it?
16:40:12	artom	*Manilla
16:40:20	gibi	bauzas: I don't think we can detect all the design issues at spec time. But still I think we should be better than doing it at the day of FF
16:40:21	sean-k-mooney	not sure about manilla
16:40:22	bauzas	jhartkopf: I just hope you can understand that our API resources are important to us and we really care about the UX
16:40:39	sean-k-mooney	ephmeral also likely not
16:40:51	bauzas	artom: no, I was mentioning a design issue being raised
16:40:53	artom	Ah
16:41:38	bauzas	ephemerals and PCI are on their way but they're big and need a lot of back and forths
16:41:48	gibi	the only easy answer for me know is that I should have asked dansmith earlier to review the series
16:41:53	bauzas	but they hadn't required a redesign
16:42:11	bauzas	gibi: well, if you read the spec, I had concerns about the restart
16:42:37	dansmith	gibi: this is why I'm against yearly releases.. because this stuff, very unfortunately, only comes to a head when there's a deadline...
16:42:40	gibi	so the I should have asked you to review the series too
16:42:41	bauzas	we just didn't went further thinking about configdrive problems
16:42:59	gibi	dansmith: I totally agree about having more meaningful deadline would help
16:43:07	sean-k-mooney	dansmith: because thats the forcing function that get a lot of use to review a tthe same time
16:43:14	sean-k-mooney	dansmith: in the absence of an inperson PTG
16:43:19	dansmith	and I only reviewed it because it was in front of something else that I know we needed to land, which was up against the deadline
16:43:27	bauzas	and I was on PTO for 3.5 weeks
16:43:30	bauzas	so blame me too
16:43:33	dansmith	WE ALL SUCK
16:43:36	dansmith	there ^ :P
16:43:40	sean-k-mooney	:)
16:43:59	gibi	I don't want to blame, I want to find something I or we can do different next time
16:44:27	jhartkopf	this ^
16:44:30	bauzas	everyone has different review sensitiveness
16:44:54	bauzas	for me, I mostly care about upgrades and controler/compute relationship
16:45:01	bauzas	but I miss other things
16:45:19	bauzas	the point is, we're humans
16:45:37	bauzas	artom: you're Quebecian
16:45:43	bauzas	there is a default fail
16:46:02	gibi	I don't want to believe that this is OK as is. :)
16:46:06	artom	You want a Maurice Richard riot v2.0? Because that's how you get a Maurice Richard riot v2.0.
16:46:08	gibi	I want to improve!
16:46:48	sean-k-mooney	well i guss where this should have been hashed out comes back to the spec
16:46:56	bauzas	yup ^
16:47:15	sean-k-mooney	and the first red flag should have been when the quetion of how to regenerate teh config drive came up
16:47:21	sean-k-mooney	after the spec was merged
16:48:26	gibi	dansmith: we would need a bit more than deadlines we would need to inject deadlines that forces a patch series to be ignored for 2 months or so
16:48:45	dansmith	to not be?
16:49:08	sean-k-mooney	we did actully have such a deadline in the past
16:49:19	gibi	dansmith: to be. If the deadline only means I can try the same deadline two weeks from now then that does not force me to push for this particular deadline
16:49:20	dansmith	I think part of the problem is that jhartkopf has to wait 6mon (minimum) before he has a consumable release with this in it
16:49:24	sean-k-mooney	it was called the non-priorty feature propsal deadline
16:49:30	dansmith	gibi: yeah that's true
16:49:39	sean-k-mooney	which requried any feature not marked as a priorty to be propsed by m2
16:49:47	dansmith	gibi: two weeks is probably too infrequent to be pressure-inducing
16:50:01	gibi	dansmith: true, I defaulted to scrum
16:50:04	dansmith	we should probably retro this later, I have other m3 stuff I need to do
16:50:56	gibi	sean-k-mooney: that fulfills the big-enough-pain-to-miss requirement. :)
16:51:44	gibi	sean-k-mooney: I guess it introduced another issue, selecting priorities. But that was maybe less of a problem
16:52:07	gibi	and we did see that having priorities would help with the rebuild bfv series
16:52:16	gibi	* would have helped
16:52:50	gibi	so I'm not against discussing the reintroduction of non-priority feature freeze as one possible option

Earlier   Later