Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-02-21
18:55:57 mriedem i assume it's setting OS_TENANT_ID
18:56:44 diana_clarke mriedem: Okie dokie. I'll look into it then. I just assumed someone else probably got to it already, but perhaps not.
19:10:53 dansmith alaski: just thought you might want to know.. I just started the first ever instance in a second cell(v2)
19:11:30 gmann gibi: bauzas I am ok to add that tests in tempest but few inline comment and yes we can reduced the testing bits if that really saving time and resource - https://review.openstack.org/#/c/374887/12
19:11:31 melwitt dansmith: did you have to do anything more than the cell-aware compute node get?
19:12:01 dansmith melwitt: yes, lots
19:12:06 melwitt oh
19:12:25 melwitt (I meant as the last hurdle)
19:12:53 dansmith compute node and service get per cell, and then (super-)conductor was creating the instance action in the wrong cell
19:12:54 dansmith i.e. not targeted
19:12:55 dansmith and failing
19:13:06 dansmith which means that's a bug we would not have caught if we didn't go with the full three-way deal right way
19:13:09 dansmith so, glad we did that
19:13:41 melwitt ah
19:16:04 bauzas gmann: FWIW, I don't like that because it just consumes a lot of infra resources for some very small coverage
19:16:53 bauzas gmann: instead, I'm just writing a functional test in the nova tree that would setup a fake env using fixtures and would verify the output without requiring further space
19:18:56 openstackgerrit Anusha Unnam proposed openstack/nova master: Remove mox from nova/tests/unit/virt/vmwareapi/test_vif.py https://review.openstack.org/436596
19:19:12 gmann bauzas: yea functional tests on nova is nice idea. But on tempest side also i think this is something nice things to tests. yea infra resource is issue but we can sort out that where to run those tests. Making Tempest more useful for production env testing also :)
19:19:43 gmann bauzas: i mean yea infra resource and gate is 1 think but it will be for production cloud testing also
19:21:06 alaski dansmith: \o/ awesome
19:25:26 gmann bauzas: like we have same host and different host filter tests on tempest
19:26:25 bauzas aaaaand I don't like that :)
19:26:45 bauzas we also have ServerGroup testing in the nova functional tree y'know
19:27:05 bauzas so, in between both ways, I prefer the less consuming one
19:27:08 gmann bauzas: I can check if we can make hotse filter and aggregate filter tests in single class and can be run on single multinode setup
19:27:33 gmann so that infra resource things can be shared
19:27:41 bauzas again, I don't see the added value of running it in a multinode env if we can just spin a couple of services using fixtures for doing that
19:28:01 openstackgerrit Matt Riedemann proposed openstack/python-novaclient master: Add functional test for "nova boot --image-with" https://review.openstack.org/436608
19:28:05 mriedem sdague: ^ check out how fun that is
19:28:36 bauzas gmann: and that would require to deploy all the filters for testing each of those, meaning no isolation, good luck with that
19:28:37 openstackgerrit Matt Riedemann proposed openstack/python-novaclient master: Remove deprecated image commands/API bindings https://review.openstack.org/436494
19:31:17 gmann bauzas: you mean no isolation? even test will add different tenants on aggregate and test will not boot server with any other scheduler hint
19:32:28 gmann i mean let's devstack enabled those filters on nova side(it does currently and we can add aggregate tenant isolation also) and specific tests for aggregate filter will just add tenant metadata on aggregate and boot server
19:36:01 bauzas gmann: say you want to test two filters
19:36:16 bauzas gmann: and you don't want to spin a devstack for every case
19:37:01 bauzas gmann: what you need then is, spin my devstack, pimp my scheduler config, restart my scheduler config, do my POST, verify my result, and loop back for the second filter
19:37:21 openstackgerrit Roman Podoliaka proposed openstack/nova-specs master: Ironic: add support for instance RAID configuration https://review.openstack.org/408151
19:37:25 bauzas if you wanna make sure that a 2nd filter result wouldn't pollute
19:38:25 bauzas gmann: because spin devstack, pimp my sched config by adding two filters, restart, do POST, verify against a specific filter means that you could get NoValidHost because of the 2nd filter you don't explicitely wanna test
19:39:19 gibi bauzas: only if the two filters interfere.
19:39:42 bauzas gmann: either way, I leave you convince the tempest folks, on my side, I'm writing the in-tree functional test for that
19:39:47 gibi bauzas: and here the tenant isolation filter only filter out hosts if the aggregate has special metadata
19:40:15 gmann bauzas: I understand. currently we have those are filter enabled from devstack side - https://github.com/openstack-dev/devstack/blob/master/lib/nova#L94
19:41:02 gmann so as gibi mentioned if those filters are not conflicting it should be fine right?
19:44:11 gibi bauzas: and I'm not against having the functional test but it has a benefit having a tempest test against a deployed system to ensure that depolyment does the isolation
19:44:54 openstackgerrit Dinesh Bhor proposed openstack/nova-specs master: Policy support for specific metadata keys https://review.openstack.org/436493
19:45:07 bauzas gibi: if that's not running in the gate, then maybe
19:45:29 mriedem ildikov: upstream training email sent
19:45:31 bauzas but I'm seriously concerned by that being a thing for the gate
19:45:40 mriedem ildikov: and it's on the etherpad for nova for tomorrow morning
19:45:54 ildikov mriedem: thanks much!
19:46:00 bauzas upstream training ?
19:46:08 bauzas need for volunteers ?
19:46:12 mriedem bauzas: yes
19:47:08 bauzas ildikov: well, I did run upstream trainings in the past (Vancouver and Paris) so I could help but I don't know yet if I could be budgeted for Boston
19:47:09 ildikov bauzas: I'm looking for a liaison from the team who would help us with the training and on boarding new contributors in general
19:47:41 ildikov bauzas: great, thank you!
19:48:09 bauzas ildikov: I'm not signing off anything yet :)
19:48:14 gibi bauzas, gmann: what is the policy now, can we have tempest test merged in tempest repo while those test are not running in any gate job?
19:48:21 ildikov bauzas: we can keep in touch with the Boston travel, I hope it will work out
19:48:51 ildikov bauzas: noted :)
19:48:53 bauzas ildikov: do you need teachers or liaisons ?
19:49:21 gmann gibi: no, it should be running somewhere in some job
19:49:55 openstackgerrit Roman Podoliaka proposed openstack/nova master: Revert "fix usage of opportunistic test cases with enginefacade" https://review.openstack.org/434843
19:50:01 ildikov bauzas: as we are continuously evolving the training I would like to have someone who is a liaison as well
19:50:30 gmann bauzas: gibi tempest has multinode tests and so infra do setup multinode env on gate so testing 2 test cases or 5 should not matter much
19:50:35 ildikov bauzas: I don't intend to overload you with tasks
19:50:53 ildikov bauzas: if you're interested we can chat when we have a break
19:52:00 openstackgerrit Sean Dague proposed openstack/nova master: Verify project id for flavor access calls https://review.openstack.org/435432
19:52:00 openstackgerrit Sean Dague proposed openstack/nova master: Verify project_id when quotas are checked https://review.openstack.org/435010
19:52:00 ildikov bauzas: also I sent out a mail reminder to the Dev ML that we will meet with people who volunteered so far tomorrow at lunch
19:52:11 gibi gmann, bauzas: I feel a deadlock here as bauzas doesn't want to run this specific tempest test on the gate
19:52:59 gmann gibi: but bauzas concern is infra resource consumption which is not affected by adding that tests :)
19:53:26 bauzas gmann: I'm concerned by the infra resource consumption we could get if we add more tests
19:53:26 gmann if tempest not having any multinode tests then yes we should think twice to add any multinode tests
19:53:59 gmann but in current situation, it hardly make any difference as multinode set is there
19:55:11 gmann bauzas: ah, i see. yea we should take care those for sure but filter things looks like good to tests will full stack without fixture things
19:57:17 gmann but its all upto you, where you want to add more filter tests I am ok if you totally against testing more filter things deeply on tempest and think nova tests with fixture is enough.
19:58:37 openstackgerrit Roman Podoliaka proposed openstack/nova-specs master: Ironic: add support for instance RAID configuration https://review.openstack.org/408151
19:58:40 gibi unfortunately I still have a use case to run this test against a deployed cloud so either it is in the tempest repo or I have to keep it in an internal tempesttest-plugin repo
19:59:19 gibi as I cannot run nova functional test against a deployed cloud
19:59:23 gmann gibi: i see, from QA team point of view we encourage to upstream those tests
19:59:49 mriedem how long does the test actually take?
19:59:57 mriedem if it's 20 seconds, then it's not a big deal right?
20:00:20 mriedem we'd also have to enable the filter in one of the jobs, i was suggesting the nova-next job
20:00:27 mriedem gmann: gibi: ^
20:00:47 mriedem we could enable it globally in devstack i guess, i don't care much either way
20:02:05 gibi mriedem: I think we can enable in devstack globally as it only filter out hosts if there is specific aggregate metadata is present
20:02:47 gibi mriedem: the proposed tests boot 3 VMs so It should not take much time
20:05:01 gmann need to check when we add aggregate filter enabled on devstack side
20:06:06 gmann mriedem: gibi but yea should not be time consuming as it just boot 2-3 VM on different hosts
20:06:07 gmann mriedem: i can try to enable filter things on devstack and check the time
20:06:50 gibi mriedem, gmann: I'm checking our internal test logs to see the times there
20:07:09 gmann gibi: thanks
20:13:34 openstackgerrit Matt Riedemann proposed openstack/python-novaclient master: Remove deprecated image commands/API bindings https://review.openstack.org/436494
20:13:35 openstackgerrit Matt Riedemann proposed openstack/python-novaclient master: Add functional test for "nova boot --image-with" https://review.openstack.org/436608
20:18:26 openstackgerrit Dan Smith proposed openstack/nova master: Make scheduler get hosts from all cells https://review.openstack.org/436648
20:18:27 openstackgerrit Dan Smith proposed openstack/nova master: Make conductor create InstanceAction in the proper cell https://review.openstack.org/436649
20:18:52 openstackgerrit Matt Riedemann proposed openstack/python-novaclient master: Remove deprecated image commands/API bindings https://review.openstack.org/436494
20:18:52 openstackgerrit Matt Riedemann proposed openstack/python-novaclient master: Add functional test for "nova boot --image-with" https://review.openstack.org/436608

Earlier   Later