Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2018-03-13
10:40:21 kashyap mdbooth: That's the redundant Cinder call, is it?
10:40:22 kashyap bdms = objects.BlockDeviceMappingList.get_by_instance_uuid(
10:40:22 kashyap ctxt, instance.uuid)
10:40:43 kashyap In nova/compute/manager.py, _post_live_migration()
10:40:53 mdbooth kashyap: No, the redundant cinder call is the second call to initialize_connection
10:41:09 mdbooth To fetch the connection_info which we had before, but overwrote during pre_live_migration
10:42:06 kashyap Err, the yeah. /me reads closer
10:42:32 mdbooth kashyap: With my patch we're still fetching those BDMs, we just do it earlier, before calling pre_live_migration
10:42:56 kashyap Yeah, re-reading your commit
10:43:08 mdbooth That means we've stored the 'old' values, which is also why I've explicitly named them source_bdms.
10:44:24 kashyap Yeah, noticed the 'source_bdms' parameter, and the accompanying code comment
10:45:22 mdbooth kashyap: I'd be glad of a critical eye on that, btw. Substantive -1s very much appreciated!
10:46:27 kashyap mdbooth: Understood. I don't claim expertise in the BDM area, first educating myself about the code area.
10:46:50 kashyap mdbooth: Also sounds like you want to see it exercise in a functional test by "someone"?
10:47:41 kashyap (Besides the 'mocks' in the unit tests)
10:48:11 mdbooth Well testing is something I'm not sure of. I have tested it manually (well, the equivalent backported to Newton), but ideally it's not doing anything it wasn't doing before, just in a different order.
10:48:35 mdbooth So if you've got an idea for a useful additional test that would also be good.
10:48:42 mdbooth stephenfin: ^^^
10:48:43 kashyap Ah, okay. I see you're more looking for any potential races
10:49:30 mdbooth There shouldn't be an issue with races because the call to pre_live_migration is synchronous, but again please verify.
10:56:42 kashyap `git review -d` for a tiny change is horribly slow for me; wonder if it's just me
10:57:25 kashyap (It's _still_ "downloading", for Pete's sake.)
11:03:23 kashyap Done; /me should've timed it
11:05:44 openstackgerrit Bence Romsics proposed openstack/osc-placement master: Do not depend on jenkins user in devstack gate https://review.openstack.org/552476
11:19:45 openstackgerrit Stephen Finucane proposed openstack/nova master: Transform live_migration.post.dest notifications https://review.openstack.org/469784
11:39:17 openstackgerrit Raoul Hidalgo Charman proposed openstack/nova master: Expose shutdown retry interval as config setting https://review.openstack.org/552483
12:18:46 kashyap mdbooth: Should the comment here in post_live_migration() be updated, in light of your change: https://github.com/openstack/nova/blob/master/nova/virt/libvirt/driver.py#L7716,#L7729
12:18:51 kashyap Tell me if I'm talking nonsense...
12:25:01 mdbooth kashyap: Um... no you're not. I've lost the libvirt driver change in the forward port :/
12:25:28 kashyap mdbooth: Wanted to check here before I added a comment in the review;
12:31:55 cdent edleafe: do you have an opinion on whether we need both 'any' and 'all' on member_of? (I think you answered everything else I mentioned on that review)
12:34:04 openstackgerrit Raoul Hidalgo Charman proposed openstack/nova master: Expose shutdown retry interval as config setting https://review.openstack.org/552483
12:35:42 edleafe cdent: the only use cases ever discussed were for the 'any' case
12:36:13 edleafe cdent: I can't think of an 'all' case, but I think that would be appropriate for an iterative approach if it ever does arise
12:37:25 cdent my brain struggles to articulate the reason for the 'any' case. It's the 'all' case that seems obvious to me but I suspect I'm not understanding the path of things in nova-scheduler
12:38:49 edleafe Right now we get back all RPs. This is simply a way to begin to limit that through aggs
12:39:30 edleafe IIRC, the initial approach was a single agg, but later discussions felt that was a bit too restrictive (don't remember details), so multiple agg candidates were proposed
12:40:13 openstackgerrit Surya Seetharaman proposed openstack/nova master: [WIP] Add disabled field to CellMapping object https://review.openstack.org/550090
12:40:14 openstackgerrit Surya Seetharaman proposed openstack/nova master: [WIP] Add CellMappingList.get_all_enabled() query method https://review.openstack.org/550188
12:40:14 openstackgerrit Surya Seetharaman proposed openstack/nova master: [WIP] Allow scheduling only to enabled cells (Filter Scheduler) https://review.openstack.org/550527
12:40:15 openstackgerrit Surya Seetharaman proposed openstack/nova master: [WIP] Add disabled column to cell_mappings table. https://review.openstack.org/552505
12:43:06 openstackgerrit Surya Seetharaman proposed openstack/nova master: [WIP] Bump versions of objects dependent on CellMapping https://review.openstack.org/550432
12:52:21 openstackgerrit Bence Romsics proposed openstack/osc-placement master: WIP Do not depend on jenkins user in devstack gate https://review.openstack.org/552476
12:57:52 openstackgerrit Surya Seetharaman proposed openstack/nova-specs master: Support disabling a cell https://review.openstack.org/546684
13:10:52 openstackgerrit sahid proposed openstack/nova master: libvirt: slow live-migration to ensure network is ready https://review.openstack.org/497457
13:11:47 HW-Peter hey
13:12:12 HW-Peter is there a way to regenerate the libvirt config file for an active instance using the API, without recreating the instnace?
13:12:41 openstackgerrit Merged openstack/nova master: Fix N358 hacking check https://review.openstack.org/547670
13:12:51 openstackgerrit Merged openstack/nova master: Remove old flavor_access_get_by_flavor_id db api method https://review.openstack.org/539700
13:13:12 openstackgerrit Merged openstack/nova master: Make the InstanceMapping marker UUID-like https://review.openstack.org/539323
13:26:12 edleafe cdent: my gabbi-fu is weak these days. Is there a way to use a UUID defined in the fixture in a URL call? E.g.: GET /allocation_candidates?member_of=$ENVIRON['AGG1_UUID']
13:26:51 cdent yeah, pretty much just like that
13:27:16 cdent the fixture needs to set the environ (it does a few of those already)
13:27:33 edleafe cdent: ok, wasn't sure where the substitutions were allowed
13:27:52 cdent pretty much anywhere on the right hand side
13:27:54 sean-k-mooney cdent: no mather how many times i see it in opnestack code it still amuses me we use environ for things like this
13:28:53 cdent sean-k-mooney: it comes out of the fact that gabbi is capable of being used for running tests from any language (using gabbi-runner, the cli) and the environ is sort of the universal thing
13:29:11 cdent edleafe: you can do left hand subs in response_json_paths
13:29:28 cdent https://gabbi.readthedocs.io/en/latest/format.html#substitution
13:29:49 openstackgerrit Bence Romsics proposed openstack/osc-placement master: WIP Do not depend on jenkins user in devstack gate https://review.openstack.org/552476
13:30:28 HW-Peter if an instance had it's libvirt xml file deleted, is there a way to regenerate it?
13:32:40 openstackgerrit Merged openstack/nova master: Transform servergroup.addmember notification https://review.openstack.org/541101
13:32:48 openstackgerrit Merged openstack/nova master: conf: Remove deprecated 'allow_instance_snapshots' opt https://review.openstack.org/499621
13:36:35 edleafe cdent: thanks; I got that part. I wasn't sure if putting them in the URL was allowed
13:37:15 cdent yeah, from gabbi's perspective (at that stage) it is just a string
13:47:38 tblakes gibi: In the case where a deploy fails with NoValidHost, do you know where the notification is sent out that the deploy failed with that exception? Was that changed at all in the queens release?
13:54:15 openstackgerrit Matthew Booth proposed openstack/nova master: Avoid redundant initialize_connection on source post live migration https://review.openstack.org/551302
13:55:08 openstackgerrit Matthew Booth proposed openstack/nova master: Avoid redundant initialize_connection on source post live migration https://review.openstack.org/551302
13:56:19 openstackgerrit Bence Romsics proposed openstack/osc-placement master: Do not depend on jenkins user in devstack gate https://review.openstack.org/552476
13:58:27 stephenfin gibi: Fancy sending these on their merry way? https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:master+topic:crypto-cleanup
13:58:57 openstackgerrit Surya Seetharaman proposed openstack/nova master: Marker reset option for nova-manage map_instances https://review.openstack.org/539501
13:59:40 gibi stephenfin: I will do that shortly
13:59:56 gibi tblakes: let me dig out
14:00:29 edleafe cdent: no, I guess you can't put an ENVIRON value in the URL. I tried, and the call was rejected as invalid, with this output: u'member_of': u'$ENVIRON[AGGA_UUID]'
14:00:51 edleafe cdent: Any other ideas?
14:01:04 edleafe cdent: I'd really rather not hardcode UUIDs into the tests
14:01:14 cdent you can do it, so there's something else wrong
14:01:23 cdent are you working in a new file, or an existing one?
14:01:41 gibi tblakes: is this about the preemptible instances feature?
14:01:42 edleafe existing
14:01:58 cdent which fixture did you set your eviron in? is it liste in fixtures: at the top of the yaml file?
14:02:18 gibi tblakes: if yes then I summarized what we have today in notifications in the nova rocky etherpad https://etherpad.openstack.org/p/nova-ptg-rocky starting at L470
14:02:40 lennyb Hi, I have nova instance stuck in 'deleting' after trying to deploy it with ironic. nova force-delete is not working as well. I am working with master branches. Any ideas how to solve this?
14:02:42 edleafe yeah - I'm adding tests to the allocation-candidates.yaml GABBIT
14:02:56 edleafe eek! capslock!
14:04:27 cdent edleafe: It's hard for me to guess, if you wanna push or paste I can look
14:04:55 edleafe cdent: ok, gimme a few
14:05:48 tblakes gibi: No this isn't about preemptible instances. I'm just wondering about the basic case where a VM deploy fails with NoValidHost. I haven't been able to find where that failure notification should get sent out.
14:09:24 gibi tblakes: searching the code...
14:14:15 edleafe cdent: ok, flase alarm. A missing quote mark was the culprit. The output made me think that gabbi didn't substitute the value
14:14:22 edleafe false alarm, even
14:14:41 gibi tblakes: I think this is the place https://github.com/openstack/nova/blob/107ad403d46c9d6961c5b15c68df0fa24014a832/nova/conductor/manager.py#L588
14:15:02 gibi tblakes: this emits the legacy compute_task.build_instances notification
14:15:14 gibi tblakes: example of the payload is here http://paste.openstack.org/show/688874/
14:15:46 cdent edleafe: if you see the original string in the error output it can only be two thing: the sub couldn't happen because it wasn't able to be parsed as a sub (your case?) or it was parseable but the info couldn't be retrieved (the environ is not set)
14:16:04 gibi tblakes: transformation of that legacy notification is still ongoing in https://review.openstack.org/#/c/482629/
14:16:12 cdent in the former case we can't error that the sub didn't happen because gabbi doesn't know it was supposed to be a sub
14:16:38 gibi tblakes: I don't know about any change around these code path in Queens
14:17:58 gibi tblakes: at least not for the notification
14:20:06 edleafe cdent: makes sense

Earlier   Later