| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-08-07 | |||
| 07:46:35 | gibi | good morning | |
| 09:12:42 | openstackgerrit | Vladyslav Drok proposed openstack/nova master: [placement] Add api-ref for allocation_candidates https://review.openstack.org/481112 | |
| 09:12:43 | openstackgerrit | Vladyslav Drok proposed openstack/nova master: [placement] Add api-ref for usages https://review.openstack.org/480563 | |
| 09:12:43 | openstackgerrit | Vladyslav Drok proposed openstack/nova master: [placement] Make placement_api_docs.py failing https://review.openstack.org/480924 | |
| 09:12:44 | openstackgerrit | Vladyslav Drok proposed openstack/nova master: [placement] Add api-ref for RP usages https://review.openstack.org/450105 | |
| 09:12:44 | openstackgerrit | Vladyslav Drok proposed openstack/nova master: [placement] Add api-ref for RP allocations https://review.openstack.org/450550 | |
| 09:16:38 | openstackgerrit | Danfly proposed openstack/nova master: update availablity_zone in the instances table after live-migration https://review.openstack.org/491226 | |
| 09:30:23 | cdent | gibi: I decided to take a page from your testing and do random stuff and find bugs. Seen these: https://bugs.launchpad.net/nova/+bug/1708961 https://bugs.launchpad.net/nova/+bug/1708958 | |
| 09:30:25 | openstack | Launchpad bug 1708961 in OpenStack Compute (nova) "migration of single instance from multi-instance request spec fails with IndexError" [Undecided,New] | |
| 09:30:26 | openstack | Launchpad bug 1708958 in OpenStack Compute (nova) "disabling a compute service does not disable the resource provider" [Low,New] | |
| 09:37:34 | bauzas | cdent: gibi: that doesn't help but I'm pretty concerned by how we just missed https://bugs.launchpad.net/nova/+bug/1708961 | |
| 09:37:35 | openstack | Launchpad bug 1708961 in OpenStack Compute (nova) "migration of single instance from multi-instance request spec fails with IndexError" [Undecided,New] | |
| 09:38:06 | bauzas | cdent: gibi: that is a serious blocker to me | |
| 09:38:18 | cdent | bauzas: yeah, I think we’ve got some gaps in testing. | |
| 09:38:43 | bauzas | it's more than that IMHO | |
| 09:39:19 | bauzas | I remember being super cautious during reviews about how we tried to make sure that num_instances was equal or not to the list of instances... | |
| 09:39:21 | gibi | bauzas: nice findings | |
| 09:39:50 | bauzas | but we totally forgot to take care of move operations | |
| 09:40:24 | cdent | careful review is important, but it shouldn’t be a substitute for real tests | |
| 09:40:33 | bauzas | gibi: cdent: about https://bugs.launchpad.net/nova/+bug/1708958 we have ComputeFilter for that | |
| 09:40:34 | openstack | Launchpad bug 1708958 in OpenStack Compute (nova) "disabling a compute service does not disable the resource provider" [Low,Confirmed] | |
| 09:40:37 | cdent | if we have to rely on careful review, then we need better tests | |
| 09:41:00 | bauzas | cdent: I specifically asked for conservative testing about that | |
| 09:41:13 | bauzas | anyway, it's done now | |
| 09:41:22 | cdent | bauzas: yeah, 1708958, as I say on the bug report, isn’t a functionality bug, more a “in what way do we want this data to be correct" | |
| 09:45:24 | bauzas | cdent: just made a comment in the bug report to clarify what I want | |
| 09:45:40 | bauzas | cdent: I don't want placement to have nova-specifisms | |
| 09:45:53 | gibi | I mean, cdent, nice findings | |
| 09:45:54 | cdent | yes | |
| 09:45:57 | bauzas | cdent: but I'd love placement to consider inventory freshness | |
| 09:45:59 | gibi | I'm sooo slow today | |
| 09:47:13 | cdent | bauzas: I think the issue is simpler than that. The choice is really: should disabling a hypervisor cause (something in nova) to delete the resource provider | |
| 09:47:23 | cdent | I think it should | |
| 09:47:34 | cdent | since if a service is diabled it is not providing resources | |
| 09:48:21 | bauzas | cdent: that would be one option, I agree | |
| 09:48:26 | gibi | cdent: it is not that simple. As disabling a host doesn't mean that the VMs running on that host freed up all the resources | |
| 09:48:43 | gibi | cdent: so that host still providing resources to existing VMs | |
| 09:48:54 | bauzas | gibi: from a placement perspective, we don't care about the space that RP has | |
| 09:49:14 | cdent | bauzas: actaully gibi’s right: we want to use allocations to indicate quota use | |
| 09:49:19 | bauzas | gibi: unless *cough cough* we have shared resources from that honst :) | |
| 09:49:30 | cdent | and we can’t delete an rp (or inventory) if there are allocations in use | |
| 09:49:46 | bauzas | yet another can of worms spotted IMHO | |
| 09:49:54 | cdent | sigh, what a twisted web we weave | |
| 09:50:04 | bauzas | anyway, should only be optimization, since we have ComputeFilter | |
| 09:50:58 | bauzas | that's just we're providing a whole list that can be long of potential candidates that are eventually not selected because of the computefilter | |
| 09:51:07 | bauzas | again, low | |
| 09:51:16 | bauzas | the other bug is more concerning me | |
| 09:51:38 | cdent | I _think_ the fix is relatively straightforward on that one | |
| 09:51:39 | bauzas | I'll try to look at a proper resolution once I'm done with writing the prelude section explaining how great we are :) | |
| 09:52:11 | cdent | but given how many gears within gears we have, who knows what the cascades will be | |
| 09:52:39 | gibi | cdent: you mean this is an interesting challenge? :) | |
| 09:53:10 | cdent | gibi: heh, that’s very positive of you for a monday morning | |
| 09:53:20 | bauzas | cdent: I really wanted to help placement bugfixing, permission to work on the critical one ? :) | |
| 09:53:31 | bauzas | the resize ones were just killing my brain | |
| 09:54:18 | cdent | bauzas: I don’t think you need to ask my permission, if you want that one, take it. I haven’t started on it. I need to catch up on a bunch of reviews, and then later in the day I’ll go back to breaking stuff and reporting bugs | |
| 09:54:29 | bauzas | roger. | |
| 09:54:38 | cdent | I set up a 6 node devstack over the weekend so I can do some real experimentation. I fear what I might find. | |
| 09:54:44 | gibi | cdent: I just want to motivate you to dig into it :) | |
| 09:54:49 | bauzas | permission is more for making sure I'm not duplicating efforts | |
| 09:55:08 | bauzas | like, /me taking mutex on it | |
| 09:55:09 | cdent | bauzas: godspeed | |
| 09:55:12 | bauzas | voilà. | |
| 09:56:28 | bauzas | FWIW, I'm in a perfect place for chasing bugs | |
| 09:56:38 | bauzas | litterally 200m away from the meditarean sea | |
| 10:09:43 | ratailor | Can anybody help me in debugging this code block. | |
| 10:09:44 | ratailor | https://github.com/openstack/nova/blob/master/nova/objects/host_mapping.py#L90-L93 | |
| 10:10:19 | ratailor | How can I check value of api_models.HostMapping.host in above code block. | |
| 10:10:22 | stephenfin | sdague, bauzas, cdent: Is live migration an operator or user-level operation, i.e. 'doc/source/admin' or 'doc/source/user'? | |
| 10:12:15 | stephenfin | I ask because I think I might need to move this document https://review.openstack.org/#/c/490952/1/doc/source/admin/migration.rst | |
| 10:12:29 | stephenfin | and this, for evacuate https://review.openstack.org/#/c/490952/1/doc/source/admin/evacuate.rst | |
| 10:13:15 | bauzas | stephenfin: live migrations are only possible for users that are having admin keystone rights | |
| 10:13:29 | stephenfin | bauzas: and evacuations? | |
| 10:14:12 | bauzas | stephenfin: AFAIR, evacuations are only for admin-only too since it requires a knowledge of your host health | |
| 10:14:19 | bauzas | stephenfin: but I can check the policy | |
| 10:14:44 | bauzas | stephenfin: the thing is, we show those operations by the CLI | |
| 10:14:45 | stephenfin | bauzas: So they both warrant staying in the admin guide. I'll keep them there so. Thanks! :) | |
| 10:14:53 | stephenfin | what do you mean? | |
| 10:14:59 | stephenfin | "by the CLI"? | |
| 10:15:07 | bauzas | stephenfin: I mean a single user can see those operations | |
| 10:15:14 | openstackgerrit | Sylvain Bauza proposed openstack/nova master: Add a prelude section for Pike https://review.openstack.org/491424 | |
| 10:15:34 | stephenfin | bauzas: You mean regardless of whether they're an admin or not? | |
| 10:15:40 | bauzas | stephenfin: yup | |
| 10:15:52 | bauzas | it will just give'em a 401 if they try | |
| 10:16:43 | bauzas | the fact is, I don't remember which specific move operation is either admin-only or regular user, depending if people provided a target or not | |
| 10:17:08 | bauzas | but anyway, those two tools (live-mig and evac) are aimed for admins anyway | |
| 10:18:21 | stephenfin | bauzas: Right, they can stay where they are now so | |
| 10:19:08 | bauzas | it'll take a while tho, under 1Mbps b/w | |
| 10:24:37 | bauzas | stephenfin: okay, confirmed, live-migrations and evacuations are admin-only, and cold migrations depend on whether you specify a target or not | |
| 10:30:07 | stephenfin | bauzas: Awesome. It stays in the admin guide so | |
| 10:35:48 | sdague | stephenfin: right, with default policy they are both admin only | |
| 10:36:09 | sdague | bauzas: I'm confused about your single user statement | |
| 10:42:58 | vks1 | sdague: hi | |
| 10:44:56 | sdague | vks1: morning | |
| 10:45:14 | vks1 | sdague: good morning | |
| 10:46:49 | vks1 | sdague: for REST API we use iso format for timestamp, which oslo API is used to generate that ? | |
| 10:50:17 | bauzas | sdague: sorry, was disconnected, which statement in particular? | |
| 10:56:43 | openstackgerrit | Stephen Finucane proposed openstack/nova master: doc: Add additional content to admin guide https://review.openstack.org/490952 | |
| 10:59:29 | sdague | vks1: I'm not sure it's coming from oslo | |
| 11:00:54 | sdague | vks1: I think they are going through this - https://github.com/openstack/nova/blob/6e77e75993c5b431c68c963bb06494ae5d36b944/nova/utils.py#L1344-L1355 | |
| 11:12:08 | openstackgerrit | Sylvain Bauza proposed openstack/nova master: Fix migrate single instance when it was created concurrently https://review.openstack.org/491439 | |