| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-03-01 | |||
| 19:36:14 | FL1SK | what would be different with Nova KVM? Does anyone have any clues | |
| 19:55:44 | openstackgerrit | melanie witt proposed openstack/nova-specs master: Proposal to count resources to check quota in API for cells https://review.openstack.org/431749 | |
| 20:01:03 | mriedem | mtreinish: i think i see the issue with that fuzzy ip filtering | |
| 20:02:31 | openstackgerrit | Timofey Durakov proposed openstack/python-novaclient master: Release note for cell_name deprecation https://review.openstack.org/439822 | |
| 20:03:12 | mtreinish | mriedem: cool, so I was right it was a nova bug? | |
| 20:03:21 | openstackgerrit | Timofey Durakov proposed openstack/python-novaclient master: Release note for cell_name deprecation https://review.openstack.org/439822 | |
| 20:04:04 | tdurakov | mriedem: hi, release note for nova-client^ | |
| 20:05:50 | mriedem | tdurakov: thanks, some nits inline | |
| 20:06:23 | mriedem | mtreinish: here https://github.com/openstack/nova/blob/3a2a181796664fea7f6069d49f9cd367b8cff825/nova/compute/api.py#L2496 | |
| 20:06:31 | mriedem | we don't filter the ip in the db query actually | |
| 20:06:58 | mriedem | on the first match, we return https://github.com/openstack/nova/blob/3a2a181796664fea7f6069d49f9cd367b8cff825/nova/compute/api.py#L2508 | |
| 20:07:20 | openstackgerrit | Timofey Durakov proposed openstack/python-novaclient master: Release note for cell_name deprecation https://review.openstack.org/439822 | |
| 20:07:28 | tdurakov | mriedem: done | |
| 20:09:03 | mriedem | http://paste.openstack.org/show/600998/ | |
| 20:09:05 | mriedem | mtreinish: ^ | |
| 20:09:07 | mriedem | that's the bug | |
| 20:10:56 | mriedem | mtreinish: so i'm thinking we build the list of matches, and if more than one, we have to filter the matches from most specific to least specific | |
| 20:11:10 | mtreinish | mriedem: oh, so that's much easier. It would have been a bit more involved if it was in the db layer | |
| 20:11:38 | mriedem | i'm not entirely sure how to do that, first thought is you throw ^<address>$ on there | |
| 20:11:45 | mtreinish | mriedem: yeah that makes sense. Could do a first pass on exact match (without a regex) or would that be too big a departure from the existing api semantics | |
| 20:12:07 | mriedem | that's actually easier | |
| 20:12:21 | mriedem | exact match first, if no hits, use regex | |
| 20:12:31 | mriedem | if multiple hits after regex... shrug? | |
| 20:12:40 | openstackgerrit | Timofey Durakov proposed openstack/python-novaclient master: Release note for cell_name deprecation https://review.openstack.org/439822 | |
| 20:12:51 | mtreinish | mriedem: return all? | |
| 20:13:07 | mtreinish | or return error | |
| 20:13:11 | mriedem | return all | |
| 20:13:14 | mriedem | this is a LIST operation | |
| 20:13:25 | mriedem | let the client sort it out | |
| 20:13:41 | mriedem | let me see how latent this is | |
| 20:15:18 | mriedem | i also wonder if it behaves this way because of https://github.com/openstack/nova/blob/3a2a181796664fea7f6069d49f9cd367b8cff825/nova/compute/api.py#L2362 | |
| 20:15:35 | mriedem | if you filter on fixed_ip it uses an exact regex match | |
| 20:16:09 | mriedem | the tempest test doesn't use that though | |
| 20:18:28 | openstackgerrit | Matt Riedemann proposed openstack/python-novaclient master: Release note for cell_name deprecation https://review.openstack.org/439822 | |
| 20:19:12 | mriedem | mtreinish: i put a -1 on the tempest patch and added nova to the bug report | |
| 20:22:05 | mriedem | mtreinish: so it looks like when vish added that long ago it returned all matches https://github.com/openstack/nova/commit/ae781ee97947c33d6d43e4c21df4f338c875bf1c | |
| 20:22:15 | mriedem | https://github.com/openstack/nova/commit/ae781ee97947c33d6d43e4c21df4f338c875bf1c#diff-516904cc81cade24a9122ecf96707bf0R2029 | |
| 20:22:44 | mriedem | this made it short circuit https://github.com/openstack/nova/commit/8fc7eee58ebd929dda75e467f93c381091da7ace | |
| 20:22:53 | erhudy | does anyone have any top tips for taking a particular SQL query and walking back to the SQLAlchemy statements that are responsible for it? perhaps particular settings in default_log_levels would print something useful out? | |
| 20:23:20 | mriedem | erhudy: i think there are trace options in the oslo.db section | |
| 20:24:25 | mtreinish | mriedem: heh, it always comes back to vish | |
| 20:25:56 | mriedem | heh, it's actually kaufer's regression | |
| 20:26:09 | mriedem | don't worry, i'm harranging him internally | |
| 20:26:18 | mriedem | for someone he did 2 years ago and doesn't remember at all | |
| 20:26:21 | mriedem | *something | |
| 20:32:13 | sdague | erhudy: the generated sql is emitted by sqla at debug level, but we turn that off because its *so* noisy. If you play with some of the oslo.log default configs you can tweak it from there | |
| 20:32:37 | erhudy | yeah, i just tried connection_trace and that blew up in my face with a stack trace, connection_debug is pretty noisy but getting closer | |
| 20:32:43 | Gorian | mriedem: sorry, ran away to lunch | |
| 20:32:50 | Gorian | I write python, so I 'm okay with Zen :P | |
| 20:32:57 | erhudy | somebody is running some ansible playbook against a cluster with 25 forks and it is causing the database to go nuclear | |
| 20:33:23 | erhudy | so i am on a spelunking expedition to understand why | |
| 20:33:26 | mriedem | erhudy: ok so you found those, sorry i was looking them up but got distracted | |
| 20:33:33 | mriedem | erhudy: zzzeek might have pointers | |
| 20:34:04 | sdague | erhudy: https://github.com/openstack/oslo.log/blob/5d1a93bc0f597768e27bd116de9d7a66b4cd1f24/oslo_log/_options.py#L20 | |
| 20:34:30 | erhudy | thanks | |
| 20:35:11 | mriedem | sdague: so i guess https://review.openstack.org/#/c/324720/ is happyish | |
| 20:35:17 | mriedem | not sure the xen or vmware failures are related | |
| 20:35:38 | mriedem | oh the vmware one is from november | |
| 20:35:39 | sdague | erhudy: I think that exposes as default_log_levels back in any of the projects, so replace it with the list you like, then debug logs for sqla should come back | |
| 20:35:39 | mriedem | hah | |
| 20:35:59 | sdague | mriedem: I guess it didn't like the recheck | |
| 20:36:36 | sdague | mriedem: yeh, just comment what you would like in a reno (if so), and I'll update the commit message for merge | |
| 20:36:47 | mriedem | the vmware one isn't from a recheck | |
| 20:36:47 | sdague | it pulls out future sec exposures as well | |
| 20:36:58 | sdague | right, but when I update the patch, they should revote | |
| 20:36:59 | mriedem | speaking of vmware....bueller? | |
| 20:37:02 | mriedem | they won't | |
| 20:37:05 | sdague | oh | |
| 20:37:14 | sdague | well, it's a libvirt feature | |
| 20:37:27 | sdague | I'm pretty sure you can't get there from anywhere else | |
| 20:37:29 | mriedem | is it? | |
| 20:39:01 | sdague | git grep inject_data | |
| 20:39:13 | mriedem | the xen ci failure was a stack failure | |
| 20:39:14 | sdague | I see libvirt and xen | |
| 20:39:22 | sdague | but xen is a different function | |
| 20:40:22 | sdague | yeh, xen half end runs around this with their own localfs call | |
| 20:40:29 | sdague | https://github.com/openstack/nova/blob/37cd9a961b065a07352b49ee72394cb210d8838b/nova/virt/xenapi/vm_utils.py#L2479 | |
| 20:41:34 | sdague | this would take it out of the libvirt path | |
| 20:41:45 | mriedem | ok, i assumed you'd be removing VFSLocalFS altogether | |
| 20:41:46 | sdague | the xen path has a fallback here when the xenagent isn't a thing | |
| 20:42:45 | sdague | well this patch was a DNM exploring a security bug | |
| 20:43:18 | sdague | a xen person would have to get on board about what the xen path looks like here, the libvirt one is pretty straight forward | |
| 20:47:25 | openstackgerrit | Maciej Szankin proposed openstack/nova master: conf: remove *_topic config opts https://review.openstack.org/433924 | |
| 21:03:10 | cfriesen | bauzas: it looks like we set request_spec.ignore_hosts in compute.api.API.resize(), so we can't overwrite it later when we call reset_forced_destinations() from nova.conductor.tasks.migrate.MigrationTask.execute() | |
| 21:20:05 | dansmith | dtp: still around? | |
| 21:20:09 | dtp | yessir | |
| 21:20:30 | dansmith | dtp: so I think we're failing to get a list of tests that fail because we're just timing out the job on my patch | |
| 21:20:34 | dansmith | too many fails means too many delays | |
| 21:20:43 | dansmith | I ran the whole deal locally and got a good report | |
| 21:20:57 | dansmith | looking at it, the hypervisors list api is probably a self-contained thing you could work on if you want | |
| 21:21:04 | dansmith | tempest.api.compute.admin.test_hypervisor.HypervisorAdminTestJSON | |
| 21:21:22 | dansmith | if you run that module locally you'll get five tests that fail and then you can go from there | |
| 21:21:38 | dansmith | can probably start with just poking at hypervisors with the client and then run the test once you think it's working | |
| 21:21:41 | dansmith | interested? | |
| 21:22:43 | dtp | sure | |
| 21:22:50 | openstackgerrit | Kaitlin Farr proposed openstack/nova master: Remove deprecated keymgr code https://review.openstack.org/439855 | |
| 21:23:21 | dansmith | cool | |
| 21:23:54 | dtp | i've not done any tempest stuff before. that's external to nova right? | |
| 21:24:23 | dansmith | yeah, you only need to run it though no mods required | |
| 21:24:34 | dansmith | and only after you poke with nova hypervisor-list and think you have it fixed | |