Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2022-06-15
16:02:15 bauzas dansmith: I'm afraid of this being not trivial and requiring some hard work, including testing
16:02:16 dansmith sean-k-mooney: right, this is to solve that problem, and let nova present actual credentials to do that thing, instead of letting unprivileged people do it, but only through the internal api endpoint
16:02:25 dansmith sean-k-mooney: basically real auth instead of host-based auth
16:02:59 dansmith bauzas: yeah, I don't think it'll be super hard since we only do this in a couple places, but it's deep, so not trivial
16:03:14 bauzas whoami-rajat: before saying "not too hard", it's quite self-contained
16:03:20 bauzas ah, burned by dansmith
16:03:25 bauzas but there be dragons
16:03:51 bauzas also
16:04:13 dansmith I really don't think there will much burning or dragons, but I haven't gone digging myself to see exactly what would be involved, so I'm hedging :)
16:04:14 bauzas I guess you'll stick with the existing auth'd roundrobins for a couple of releases ?
16:04:32 bauzas rolling upgrades is a bit of a thing
16:04:46 dansmith bauzas: I think what we do is: if the credentials are configured, use the new way, if not, assume the old way is still active
16:04:58 bauzas all good then
16:05:00 dansmith because some people could still be running glance from folsom and be perfectly happy
16:05:26 dansmith so we don't want to introduce a hard requirement.. we could force a move later, but I think it has to be a decent window of overlap
16:05:33 bauzas I'll chime into the review, thanks for the heads up
16:05:34 whoami-rajat Folsom??? but as dansmith said, we will have backward compatibility
16:05:52 dansmith whoami-rajat: sarcasm so extreme to make it obvious :P
16:06:07 bauzas whoami-rajat: as someone who played with Folsom, glance was perfectly good at that time
16:06:13 dansmith haha
16:06:25 dansmith honestly, openstack has been downhill since folsom anyway :P
16:06:26 whoami-rajat :D
16:06:40 whoami-rajat wow, glance has been stable forever!
16:06:45 bauzas things were easier honestly
16:06:57 bauzas nova-net, glance and nova-volume were easy to manage
16:07:18 bauzas but I'm diverting
16:07:52 dansmith okay so to circle back,
16:08:27 dansmith are there any concerns with the general approach that should hold up glance merging the spec and working on this new interface in zed so that we can start converting nova and cinder in Aardvark?
16:08:55 bauzas I saw the first +2
16:08:59 bauzas I'll review the spec now
16:09:35 bauzas but from a 10.000ft level, no I think we're on a safe bet
16:10:01 dansmith bauzas: ack, could you look at it today before you go? if so I can throw my +2 on there
16:10:08 bauzas dansmith: on it
16:10:10 dansmith bauzas: thanks
16:10:44 whoami-rajat thanks dansmith and bauzas for the discussion
16:10:53 dansmith ++
16:27:07 gibi I read back I have no objection
16:28:03 whoami-rajat thanks gibi
16:32:44 bauzas I reviewed the patch
16:32:54 opendevreview Artom Lifshitz proposed openstack/nova stable/train: fake: Ensure need_legacy_block_device_info returns False https://review.opendev.org/c/openstack/nova/+/843958
16:32:55 opendevreview Artom Lifshitz proposed openstack/nova stable/train: tests: work around malformed serial XML https://review.opendev.org/c/openstack/nova/+/844743
16:32:56 opendevreview Artom Lifshitz proposed openstack/nova stable/train: functional: Use tempdir for CONF.instances_path https://review.opendev.org/c/openstack/nova/+/844750
16:32:57 opendevreview Artom Lifshitz proposed openstack/nova stable/train: Add a regression test for bug 1939545 https://review.opendev.org/c/openstack/nova/+/843959
16:32:59 opendevreview Artom Lifshitz proposed openstack/nova stable/train: compute: Ensure updates to bdms during pre_live_migration are saved https://review.opendev.org/c/openstack/nova/+/843960
16:33:01 opendevreview Artom Lifshitz proposed openstack/nova stable/train: fup: Make connection_info returned by CinderFixture unique per attachment https://review.opendev.org/c/openstack/nova/+/844744
16:33:03 opendevreview Artom Lifshitz proposed openstack/nova stable/train: func: Add live migration rollback volume attachment tests https://review.opendev.org/c/openstack/nova/+/844745
16:33:05 opendevreview Artom Lifshitz proposed openstack/nova stable/train: fup: Assert state of connection_info during LM rollback in func tests https://review.opendev.org/c/openstack/nova/+/844746
16:33:12 bauzas my only concern which isn't blocking is about making sure we provide a smooth upgrade path for our operators
16:33:46 bauzas asking them to change their config before upgrading is a pain
16:34:02 bauzas so, please, don't make it mandatory in one cycle
16:34:16 bauzas dansmith: ^ this reminds me the tick-tock cadence
16:34:54 dansmith bauzas: yeah absolutely. like I said, I think we need a pretty decent window
16:34:57 bauzas if all of this becomes Aa
16:35:07 dansmith definitely longer than even a single tick-tock
16:35:11 bauzas then we can't remove the old client usage in nova in Bb
16:35:18 dansmith for sure
16:36:13 bauzas anyway, this is a nova trhing
16:36:24 bauzas no need to paper it out in the glance spec
16:36:35 bauzas but just sayin', you could keep the old API for a bit
16:36:57 bauzas whoami-rajat: ^
16:41:21 whoami-rajat bauzas, yep, that's mostly the plan to keep a window, also until the service-to-service interaction becomes available, this all will be a placeholder and won't be functionally active i.e. old path will be used
16:42:05 bauzas whoami-rajat: how do you plan to identify all the nova calls that modify the image location ?
16:51:32 whoami-rajat bauzas, we've the glance code in nova that calls the glanceclient, eg: for add location https://github.com/openstack/nova/blob/93a65f06df67ce39d65827692150c78013c7f6d5/nova/image/glance.py#L558
16:51:41 whoami-rajat s/code/file
16:52:24 whoami-rajat all the places calling glance should be calling this file but not sure if that's what you mean
17:20:17 gibi sean-k-mooney: one more info for the todays pci-tracking work PciDeviceStats is not good for me as the pool count is 0 if the devices are consumed, but I still need to create inventory for consumed devices too. So I need to build the inventory based on the PciDevice objects alone and match them to DeviceSpec one by one to get the metadata (resource_class, and traits from the config)
17:34:21 sean-k-mooney ok so you can either get the rows directly from the tracker or build them i guess
17:34:39 sean-k-mooney i tought the stats also had the capastity but i guess not
17:36:55 sean-k-mooney ah yes
17:37:05 sean-k-mooney https://github.com/openstack/nova/blob/master/nova/objects/pci_device_pool.py#L55=
17:38:40 gibi I can access count via stats.pools directly but that is only meaningful if there is no pci allocation at all on the compute
17:38:50 gibi and yes I saw that the count is not visible via the object interface
17:42:00 sean-k-mooney right i tought we also had a total in the stats
17:42:11 sean-k-mooney that did not change in addtion to the count of free devices
17:42:13 sean-k-mooney we do not
17:50:13 gibi sean-k-mooney: quick question, does nova alway have a PciDevice object for parent devices? E.g. if only the VF is whitelisted nova still create a PciDevice for the parent is it so?
17:52:56 gibi never mind I will track it down tomoroow
17:52:59 gibi o/
17:58:27 sean-k-mooney no i dont think we do
17:58:43 sean-k-mooney we could build oen but i dont know if we do by default
17:58:50 sean-k-mooney in the pci tracker today
17:59:11 sean-k-mooney if you whitelist a vf we do not create a row for the vf in the db
18:24:06 melwitt elodilles: I wanted to get your opinion on backporting a refactor that skips a release https://review.opendev.org/c/openstack/nova/+/844750
18:28:47 elodilles melwitt: hmm. is there any reason to skip ussuri? O.o
18:29:51 sean-k-mooney im not aware of one
18:30:08 melwitt elodilles: not that I know of. we also don't know why the patch helps or why it's needed to make CI pass. have you ever heard of something like that?
18:30:39 sean-k-mooney unless it merged in ussuri when it was master
18:30:46 melwitt I looked at it and nothing rang a bell why it would help ¯\_(ツ)_/¯
18:31:07 sean-k-mooney it merged in victoria
18:31:18 sean-k-mooney https://github.com/openstack/nova/commit/980711f3d3f2db9dfd792a8c6fd15cb45a85f615
18:31:48 elodilles i guess it was added in train to avoid some kind of conflict maybe? (and i guess it was not needed in ussuri)
18:32:08 sean-k-mooney i would guess so too
18:32:27 sean-k-mooney perhaps it was merged in in ussurit into one of the other commtis
18:32:45 sean-k-mooney or artom just forgot
18:33:12 melwitt the commit message says CI was failing with "No such file or directory: '/home/zuul/src/opendev.org/openstack/nova/instances/tmpif84g_jd'" i.e. looking for a tempdir and it doesn't exist. but it's weird it wasn't needed in ussuri
18:33:39 elodilles anyway, i'd rather see that either removed from the patch series, or backported to ussuri as well (if that is really needed)
18:34:27 melwitt ++
18:35:37 melwitt thanks
18:36:03 sean-k-mooney is it because of python 2?
#openstack-nova - 2022-06-16
02:08:39 opendevreview Steve Baker proposed openstack/nova master: Align ironic driver secure boot with other drivers https://review.opendev.org/c/openstack/nova/+/844243

Earlier   Later