| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-09-26 | |||
| 12:16:21 | gibi | bauzas: thanks | |
| 12:16:39 | gibi | mdbooth: I'm on a meeting but opened the review so I will check it | |
| 12:17:02 | efried | ratailor Did you get your questions answered about the osc microversion? | |
| 12:17:24 | ratailor | efried, not yet. could you help ? | |
| 12:18:01 | efried | ratailor Your observation is correct, and by design. As for the reasoning behind it... that's complicated. | |
| 12:18:42 | efried | ratailor But if you want a specific microversion in osc, you can ask for it with an env var, CLI opt, or (I think) conf var. | |
| 12:19:13 | bauzas | efried: if the microversion is not asking for a new attribute :p | |
| 12:19:27 | efried | Yeah, there's that :) | |
| 12:19:37 | mdbooth | efried: OSC will happily send invalid requests because it doesn't understand those microversions, though. | |
| 12:20:02 | manasm | bauzas: running into an issue with resize on an upgraded env from Ocata to Pike, the resize fails at following call in scheduler_utils.setup_instance_group method - | |
| 12:20:04 | mdbooth | At that point, the user might as well just open telnet and type the http themselves :) | |
| 12:20:07 | efried | ratailor A lot of it has to do with the fact, as bauzas and mdbooth are pointing out, that osc has not maintained functional parity with microversions. | |
| 12:20:07 | manasm | request_spec.instance_group.hosts = list(group_info.hosts) | |
| 12:20:49 | manasm | bauzas: this is because the instance_group is set to None in the request_spec. | |
| 12:21:03 | manasm | Is there a known issue around that? | |
| 12:21:04 | bauzas | manasm: do you have an open bug for that? | |
| 12:21:14 | manasm | not yet | |
| 12:21:18 | ratailor | efried, To make it as stable as possible. right ? | |
| 12:21:30 | ratailor | efried, and backward compatible. | |
| 12:21:49 | efried | ratailor The reasoning may not be quite that high-minded :) | |
| 12:21:59 | ratailor | efried, cool | |
| 12:22:07 | mdbooth | This is a data corruptor with a simple fix: https://review.openstack.org/#/c/507202/ | |
| 12:22:20 | bauzas | manasm: AFAIR, we create a new InstanceGroup object when looking at the primitive dicts | |
| 12:22:40 | ratailor | efried, Thanks! | |
| 12:22:56 | efried | ratailor If you want more history/explanation on the osc side, find dtroyer. I don't remember offhand which channel is the accepted one for talking about the CLI, but I know he's in #openstack-dev | |
| 12:22:58 | mdbooth | Affects libvirt live migration with attached volumes | |
| 12:23:25 | efried | ratailor Possibly #openstack-sdks | |
| 12:23:52 | ratailor | efried, sure, Thanks! | |
| 12:24:12 | efried | ratailor yw, good luck. | |
| 12:24:19 | ratailor | efried, actually I was searching him only, but didn't found in core openstack channels. | |
| 12:24:27 | ratailor | efried, Thanks for your help! | |
| 12:24:52 | efried | It may be a tad early for him to be on just yet, but he's at least got bouncers in those two channels. | |
| 12:25:10 | manasm | bauzas: yes, we have the new InstanceGroup object but the we are trying to assign it to request_spec.instance_group which is None. | |
| 12:25:47 | bauzas | efried: ratailor: FWIW, I already provided a point for the Sydney Forum about OSC vs. our CLI in https://etherpad.openstack.org/p/SYD-nova-brainstorming L35 | |
| 12:26:08 | bauzas | I'd love to hear operators about what they use and how | |
| 12:27:29 | ratailor | bauzas, ++1 | |
| 12:28:06 | bauzas | manasm: you said when you resize? | |
| 12:28:16 | bauzas | manasm: so we're getting the ReqSpec from the DB first | |
| 12:28:53 | bauzas | if there was no group yet for that instance, then the ReqSpec.instance_group field could be None | |
| 12:28:55 | manasm | yes, btw we get the object from db using - objects.InstanceGroup.get_by_instance_uuid(context, instance_uuid), and if it is None then we don't hit the issue. | |
| 12:32:27 | bauzas | manasm: wait, I need to look at the code for resize | |
| 12:32:36 | bauzas | do you have a stacktrace to share? | |
| 12:38:52 | kashyap | mdbooth: FWIW, looks good. Compact summary, and a nice fix: https://review.openstack.org/#/c/507202/2 | |
| 12:39:56 | mdbooth | It breaks some tempest tests, though :/ | |
| 12:40:02 | mdbooth | Probably means I didn't think of something. | |
| 12:40:20 | mdbooth | It does look obvious, though. | |
| 12:40:26 | stephenfin | mdbooth: Reviewed both. Tidy job | |
| 12:40:45 | kashyap | mdbooth: Wonder if you want to link to to the virDomainMigrateFlags page: https://libvirt.org/html/libvirt-libvirt-domain.html#virDomainMigrateFlags | |
| 12:40:58 | kashyap | Which talks about WTF actually the flag VIR_MIGRATE_NON_SHARED_INC is | |
| 12:41:07 | kashyap | If you think it's not needed, disregard me | |
| 12:42:30 | kashyap | Maybe not required. Those who want to know about it, a simple look up will fetch it | |
| 12:44:40 | kashyap | BTW, the "tempest-dsvm-full-xenial" succeeded | |
| 12:51:23 | kashyap | Ah, it's the "gate-grenade-dsvm-neutron-ubuntu-xenial" that failed. | |
| 12:52:03 | mdbooth | kashyap: Yep, in the block migration tests :) | |
| 12:52:13 | mdbooth | I suspect the failure there is probably related ;) | |
| 12:52:44 | mdbooth | kashyap: If you have time to investigate those failures, btw, I have to do something else right now. Otherwise I'll get to them later. | |
| 12:52:53 | kashyap | mdbooth: I'm looking at the log - http://logs.openstack.org/02/507202/2/check/gate-tempest-dsvm-py35-ubuntu-xenial/8485b63/console.html | |
| 12:53:10 | kashyap | I'll note on the review if I learn something new from the log | |
| 12:53:56 | openstackgerrit | Yikun Jiang proposed openstack/nova master: Update Instance action's updated_at when action event updated. https://review.openstack.org/507473 | |
| 12:54:51 | kashyap | So the right log is - http://logs.openstack.org/02/507202/2/check/gate-grenade-dsvm-neutron-multinode-live-migration-nv/b210147/console.html | |
| 12:55:33 | openstackgerrit | konstantin proposed openstack/nova master: switch from filesystem to disk for parallels containers https://review.openstack.org/506687 | |
| 12:55:34 | openstackgerrit | konstantin proposed openstack/nova master: don't add device address if there is no any units https://review.openstack.org/506686 | |
| 12:57:18 | openstackgerrit | Yikun Jiang proposed openstack/nova master: Update Instance action's updated_at when action event updated. https://review.openstack.org/507473 | |
| 12:59:33 | openstackgerrit | Eric Fried proposed openstack/nova master: nova.utils.get_ksa_adapter() https://review.openstack.org/488137 | |
| 13:00:37 | openstackgerrit | Eric Fried proposed openstack/nova master: Use ksa adapter for placement conf & requests https://review.openstack.org/492247 | |
| 13:00:43 | openstackgerrit | OpenStack Proposal Bot proposed openstack/os-vif stable/ocata: Updated from global requirements https://review.openstack.org/490256 | |
| 13:02:52 | openstackgerrit | Eric Fried proposed openstack/nova master: Get auth from context for glance endpoint https://review.openstack.org/490057 | |
| 13:02:58 | openstackgerrit | Eric Fried proposed openstack/nova master: Get auth from context for glance endpoint https://review.openstack.org/490057 | |
| 13:06:19 | openstackgerrit | Sean Dague proposed openstack/nova master: Break out BasicTestCase https://review.openstack.org/507253 | |
| 13:06:20 | openstackgerrit | Sean Dague proposed openstack/nova master: Don't use mock.patch.stopall https://review.openstack.org/507527 | |
| 13:06:20 | openstackgerrit | Sean Dague proposed openstack/nova master: Remove REQUIRES_LOCKING as nothing needs process locking in the tests https://review.openstack.org/507528 | |
| 13:06:21 | openstackgerrit | Sean Dague proposed openstack/nova master: WIP: demonstrate no use of external locking https://review.openstack.org/507529 | |
| 13:08:33 | openstackgerrit | Eric Fried proposed openstack/nova master: Don't fix protocol-less glance api_servers anymore https://review.openstack.org/505317 | |
| 13:12:57 | openstackgerrit | Yikun Jiang proposed openstack/nova master: Update Instance action's updated_at when action event updated. https://review.openstack.org/507473 | |
| 13:16:40 | sdague | gibi: so, interesting fact from this morning, I'm pretty convinced we don't need any of the REQUIRES_LOCKING code | |
| 13:20:47 | sdague | gibi: also, if you are able to take a look at the qemu 2.10 support patch, that would be cool - https://review.openstack.org/#/c/505673/ | |
| 13:20:53 | sdague | or bauzas | |
| 13:23:43 | bauzas | ack | |
| 13:36:22 | mdbooth | stephenfin: Eurgh. So it seems that in python2.7 libvirt binding's migrateToURI3() barfs if destination_xml is unicode. | |
| 13:38:55 | mdbooth | stephenfin: think I'm going to leave that etree.tostring() patch always producing unicode, but I'm going to add a 2/3 hack next to the migrateToURI3 which handles the difference, because IMHO the libvirt bindings should accept a unicode string there, and it's just saner. | |
| 13:39:03 | kashyap | Matt, hmm, the assertThat() mismatch seems to be related to ID: | |
| 13:39:04 | kashyap | 2017-09-26 12:26:17.199293 | 2017-09-26 12:26:17.198 | reference = u'ubuntu-xenial-2-node-rax-ord-11137317-923993' | |
| 13:39:07 | kashyap | 2017-09-26 12:26:17.200739 | 2017-09-26 12:26:17.200 | actual = u'ubuntu-xenial-2-node-rax-ord-11137317' | |
| 13:39:19 | mdbooth | kashyap: Yes, that's just asserting that the migration happened | |
| 13:39:28 | mdbooth | i.e. assert id == the destination | |
| 13:40:00 | mdbooth | I hate python3 | |
| 13:40:15 | mdbooth | And python2 | |
| 13:40:28 | mdbooth | But the 2->3 split especially | |
| 13:40:49 | kashyap | Ah, I see. (about the assert ID) | |
| 13:49:02 | openstackgerrit | Ed Leafe proposed openstack/nova master: Add Selection objects https://review.openstack.org/499239 | |
| 13:51:32 | sdague | mriedem: I was diving through thinking more about https://review.openstack.org/#/c/507239/ last night, why do you think that external locking is required there? Because it should blow up if not provided but is needed | |
| 13:52:21 | mriedem | sdague: i thought that also locked those tests to run serially | |
| 13:52:26 | mriedem | maybe i should be using https://github.com/openstack/oslo.concurrency/blob/master/oslo_concurrency/fixture/lockutils.py#L22 | |
| 13:52:32 | mriedem | ^ is actually what i started using | |
| 13:52:36 | sdague | mriedem: serial to what? | |
| 13:52:49 | openstackgerrit | konstantin proposed openstack/nova master: switch from filesystem to disk for parallels containers https://review.openstack.org/506687 | |
| 13:52:49 | openstackgerrit | konstantin proposed openstack/nova master: don't add device address if there is no any units https://review.openstack.org/506686 | |
| 13:53:11 | mriedem | sdague: to other tests | |
| 13:53:26 | mriedem | like, things that would be mocking those nova.rpc globals | |
| 13:53:42 | sdague | all the tests in a worker run serially | |