Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2019-04-17
11:24:45 NewBruce Not that im aware of, it pre-dates my time here - but i think it was pretty ad-hoc
11:25:30 sean-k-mooney so its a manual install of RDO without a tool?
11:25:39 NewBruce but going back to basics and what we know; we can localise it to post_migrate and an attempt to update the port profiles, which it shouldnt be doing. right?
11:25:51 sean-k-mooney im wondering if we can use a gate job to repoduce
11:26:23 openstackgerrit Jiri Suchomel proposed openstack/nova master: WIP: Initial prototype for simple RPC health checks https://review.openstack.org/653406
11:27:07 sean-k-mooney mnaser: it should be fairly simple to create an all in one OSA right, im wondering how hard it would be to also add a singel node rdo computej too maybe via packstack to repoducs NewBruce's issue
11:27:10 NewBruce so, peppering the api.py with enough debug statements we’ve been able to localise where thats done - and we should be able to walk backwards right?
11:28:11 NewBruce sean-k-mooney: did you see the full log traces i sent?
11:28:44 sean-k-mooney yes it is ending in post migrate so we should be able to walk backward.
11:29:12 NewBruce grabbed your email from you launchpad page - will drop you a mail shortly
11:29:13 sean-k-mooney NewBruce: no not yet. bit of a late start today so still catching up on things
11:33:06 NewBruce quick question to the assembled masses - does anyone know if Enable cold migration with target host - Queen (https://specs.openstack.org/openstack/nova-specs/specs/queens/implemented/cold-migration-with-target-queens.html was ever implemented? ive not had any luck getting it to work
11:34:06 sean-k-mooney well its in the implemented section so it should have been yes
11:35:08 sean-k-mooney you cant do it via osc
11:35:53 NewBruce only apl?
11:36:00 sean-k-mooney nova migrate [--host <host>] [--poll] <server>
11:36:10 NewBruce Aha - gotcha; cheers
11:36:26 sean-k-mooney you have to add --host and the order is cahnged
11:36:34 sean-k-mooney we should proablly add that to osc too
13:24:54 mriedem thoughts on a potentially less invasive fix (see inline comments) on this stein regression https://review.openstack.org/#/c/653273/ are appreciated
13:25:10 mriedem i wrote the fix late last night but aren't super happy with how grossly tightly coupled all of that code is
13:32:17 sean-k-mooney mriedem: i was talking to NewBruce about the bug they are hitting and i tink i might know how to fix it but still not clear on why its happening. im starting on a patch now. should i add you to review list when i submit
13:42:39 sean-k-mooney lyarwood: mriedem when ye have time can one of ye +2 +W this https://review.openstack.org/#/c/644035/ so that we use the correct upper_constratins
13:49:09 mriedem sean-k-mooney: done and yes
13:51:55 sean-k-mooney thanks and cool will do
13:54:26 mriedem bauzas: want to hit this backport https://review.openstack.org/#/c/647911/ ?
13:54:39 bauzas sure
13:54:48 mriedem all of them actually since it's a gate race issue https://review.openstack.org/#/q/If966eb1161c842ff49aa530e4482dbca87b61a3e+status:open
13:54:50 bauzas looking
13:54:53 mriedem http://status.openstack.org/elastic-recheck/index.html#1820337
14:08:24 kashyap aspiers: Hiya...seems like I need the getDomainCapabilities() API call for Secure Boot
14:08:31 kashyap aspiers: Wondering if you're planning to split it out: https://review.openstack.org/#/c/633855/11/nova/virt/libvirt/host.py@680
14:09:01 aspiers I wasn't, as per previous conversation
14:09:13 aspiers but maybe that can be revisited
14:09:58 kashyap aspiers: Yeah, let's revisit that conversation, plese
14:10:14 aspiers Can you do me a favour and find the link to it? ;-)
14:10:17 kashyap aspiers: Because, IMHO, it is not needed to tie that to SEV at all
14:10:18 aspiers Maybe in that review
14:10:40 aspiers I remember explicitly explaining the dependency somewhere
14:11:02 aspiers Oh, that *is* the link
14:11:08 kashyap Yeah :-)
14:11:17 aspiers Yeah, my comment on March 19th
14:11:39 kashyap aspiers: Oh, I'm blind. Will read it
14:11:45 aspiers :)
14:11:59 kashyap aspiers: The reason I ask is ... just now a libvirt dev was suggesting:
14:12:19 kashyap "I suggest Nova queries domcapabilities first, to see if 'efi' with Secure Boot is available"
14:12:33 kashyap (When implementing SB support, i.e.)
14:21:55 mriedem so we have a RESIZE_CONFIRMING task_state enum but never use it while confirming a resize https://github.com/openstack/nova/blob/e25d59078e61fe9f925dbef53dfe88e575d34dab/nova/compute/api.py#L3535
14:21:59 mriedem that seems...not good?
14:22:28 mriedem unlike when reverting a resize we set the task_state in the API before casting to compute https://github.com/openstack/nova/blob/e25d59078e61fe9f925dbef53dfe88e575d34dab/nova/compute/api.py#L3515
14:26:43 openstackgerrit Vladyslav Drok proposed openstack/nova master: libvirt: Do not reraise DiskNotFound exceptions during resize https://review.openstack.org/571410
14:28:09 mriedem vdrok: mdbooth: lyarwood: more comments in https://review.openstack.org/#/c/571410/9
14:29:10 mriedem i'm inclined to take a more conservative incremental approach to fixing the VERIFY_RESIZE steady state issue
14:29:20 mriedem which we could also backport more safely i think
14:31:31 vdrok meaning somewhat like it was done in ps7 https://review.openstack.org/#/c/571410/7/nova/virt/libvirt/driver.py
14:31:53 mriedem exactly
14:33:05 mriedem i'm not entirely sold on the completely ignore case like in bug 1824974 where the disk on the host is garbage and we just put our heads in the sand
14:33:06 openstack bug 1824974 in OpenStack Compute (nova) "Nova Compute Manager (Resource update) fails if a disk is missing" [Low,In progress] https://launchpad.net/bugs/1824974 - Assigned to Vladyslav Drok (vdrok)
14:39:35 mnaser as far as I know RDO ships upstream code
14:40:13 mnaser NewBruce: I wonder if it would be a good exercise to change one of the compute nodes to log it's current highest negotiated rpc version
14:40:26 mnaser and then see if for some reason it's negotiating a bad number
14:41:28 artom mnaser, I believe you're right, RDO is straight up upstream
14:41:35 artom Any backports we do is OSP, not RSO
14:41:37 artom *RDO
14:45:52 mnaser yeah, so RDO is upstream, OSA is upstream
14:47:12 openstackgerrit Dan Smith proposed openstack/nova master: Make libvirt expose supported image types https://review.openstack.org/653454
14:48:47 sean-k-mooney mnaser: im currently updating unit test but i think http://paste.openstack.org/show/749431/ will fix NewBruce issue
14:51:05 efried mriedem: There's no blueprint or anything for sdk-in-nova yet because up to this point we've been experimenting to see how it would work and what we can get done. I do have it on the ptg etherpads in some form. But will also try to file a bp soon.
14:51:35 sean-k-mooney efried: well you can file a blueprint just as a tracker right
14:51:49 efried yeah, could do.
14:51:51 openstackgerrit Vladyslav Drok proposed openstack/nova master: libvirt: Do not reraise DiskNotFound exceptions during resize https://review.openstack.org/571410
14:52:02 mnaser sean-k-mooney: that's fair, I do feel like that might be a bit of a workaround I guess :-p
14:52:03 efried dustinc: How do you feel about leading that?
14:52:08 sean-k-mooney efried: it does not need to have a solution atached just "sdk is a thing we should use it"
14:52:51 mriedem "x exists we should use it" isn't sufficient justification imo
14:52:53 sean-k-mooney mnaser: well i have not found the root casue so yes but i think we likely should make this chagne on master and backport and then see if we can create a reporducer
14:53:09 efried sean-k-mooney: Part of what we've been working on is the scope too. "Plumb SDK" is one aspect; "stop using python-${service}client" is another.
14:53:28 sean-k-mooney mriedem: maybe not to approve the blueprint but it enough to file it
14:53:49 sean-k-mooney efried: ah ok
14:53:52 mriedem sure, blueprints are free
14:54:23 mriedem fixing the bugs they introduce is not
14:54:40 sean-k-mooney was "stop using python-*-client" in services approved as a comunity goal by the way
14:55:06 mriedem https://governance.openstack.org/tc/goals/
14:55:40 efried sean-k-mooney, mriedem: The reasoning would go something like: "We want to stop using python-${service}client => We could use ksa adapter stuff, which we've already set up, but then we would have to write do-a-thing wrappers around get/put/etc primitives => so let's use the sdk, which already has those wrappers"
14:55:53 sean-k-mooney actully i guess that is a dumb question that will be decided at the fourm/ptg in 2 weeks
14:55:55 openstackgerrit Adam Spiers proposed openstack/nova master: Add ids to sections of flavors guide to allow deep-linking https://review.openstack.org/653458
14:56:46 mriedem sean-k-mooney: is that even a proposed goal for train or are you just making it up?
14:56:56 sean-k-mooney efried: part of the scoping would be ensuring the sdk has the actual wrapper we need
14:57:13 efried mriedem: there was a proposed goal to stop using python-${service}client for CLIs.
14:57:25 mriedem yeah, that's quite a bit different
14:57:28 efried which is distantly related to this, but not really overlapping.
14:57:28 efried yeah.
14:57:44 sean-k-mooney ah that is proably what im mixing it up with
14:58:19 sean-k-mooney it was one of the tread i lost track of on the mailing list
14:58:21 efried mriedem: anyway, the ironic team (as an example) is strongly behind the idea of ripping ironicclient out of nova
14:58:40 efried so I poked at a couple of ways of doing that
14:59:27 efried and it's about time we did *something* to break our pattern of "set up a whole KSA adapter just to grab the endpoint out of it and send that into *client which uses it to create a whole nother Adapter to do its work"
14:59:38 efried which we do (in different ways!) for glance, ironic, neutron...
15:00:45 mriedem the ksa stuff in nova still feels new to me, but i'm slow and old
15:00:53 mriedem and was a big improvement over the per-service config and all that

Earlier   Later