Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2018-04-17
20:30:53 arvindn05 mriedem: thanks. so i need to load compute node using the get_by_host_and_nodename method and get the uuid that way?
20:31:29 mriedem this is to get the compute node uuid to find the resource provider for rebuild right?
20:31:36 arvindn05 yup
20:32:00 mriedem arvindn05: yeah then likely just use this https://github.com/openstack/nova/blob/master/nova/objects/compute_node.py#L262
20:32:20 arvindn05 i have both the host and node names...so it looks appropriate
20:32:25 mriedem that's a bit heavy since we don't need the full compute node object, but that should work for starters
20:33:26 arvindn05 mriedem: thanks. is there any other lightweight mapping other than going to the db?
20:33:58 arvindn05 i saw somewhere that the host id is generated from a combination of something....let me look it up
20:34:59 arvindn05 https://ask.openstack.org/en/question/29085/which-api-to-be-used-to-get-uuid-of-compute-host/?answer=29091#post-id-29091
20:35:21 mriedem no that's not it
20:35:25 cfriesen arvindn05: going to the DB should be relatively cheap
20:35:26 mriedem that's the hashed instance.hostId
20:35:58 mriedem the computenode.host is the hostname of the compute service
20:36:01 mriedem the physical hostname
20:36:11 mriedem the nodename is the same as the host for all virt drivers except ironic
20:36:12 arvindn05 mriedem: cfriesen: thanks...will go to the db then. if there are more effiecint methods, hopefully they will be pointed out in the CR
20:36:22 mriedem in the case of ironic, the computenode.hypervisor_hostname (nodename) is the ironic node uuid
20:36:41 mriedem there are not more efficient methods without writing your own new DB API query method via the ComputeNode object
20:36:48 mriedem like, get_uuid_by_host_and_nodename
20:37:17 arvindn05 ^ was hoping someone already wrote that method :)
20:37:48 mriedem i don't think it's a huge issue to lookup the compute node for the rebuild case, it's not like scheduling for that should be very expensive since all we're doing is validating the new image
20:38:38 arvindn05 yup...its not for all rebuild cases either...only rebuild cases where they have traits expressed in image....so its a minor subset
20:42:11 openstackgerrit Matt Riedemann proposed openstack/nova master: WIP: Add functional recreate test for bug 1764556 https://review.openstack.org/562041
20:42:11 openstack bug 1764556 in OpenStack Compute (nova) ""nova list" fails with exception.ServiceNotFound if service is deleted and has no UUID" [Undecided,New] https://launchpad.net/bugs/1764556
20:42:12 mriedem cfriesen: getting further but still not failing when listing instances ^
20:44:21 efried imacdonn: Quick delta there and I'm happy
20:45:28 imacdonn efried: Ahh, good catch
20:45:43 efried imacdonn: I checked and that's the only place it was used.
20:45:58 imacdonn yep, I just 'grep -r'ed too ;)
20:46:33 imacdonn ooooh
20:46:39 efried ...which is probably less reliable than grep -r
20:46:43 imacdonn heh
20:46:52 efried but way more efficient.
20:47:07 efried You know, as long as you discount all the freaking processing power the IDE had to do up front to indext the whole freaking project.
20:47:17 efried s/indext/index/
20:47:22 imacdonn yeah
20:47:30 mriedem cfriesen: if i revert this https://review.openstack.org/#/c/543970/ then the test fails with ServiceNotFound
20:48:31 openstackgerrit Matt Riedemann proposed openstack/nova master: DNM: Revert "Lazy-load instance attributes with read_deleted=yes" https://review.openstack.org/562051
20:48:39 mriedem but, that basically means it's already fixed
20:49:11 cfriesen mriedem: does host1.stop() delete the entry in the DB? or do we need host1.kill()?
20:49:24 cfriesen I definitely tested it with current stable/pike in devstack
20:49:33 mriedem .stop() doesn't delete the service
20:50:16 openstackgerrit Merged openstack/nova master: Parse forbidden in extra_specs https://review.openstack.org/561677
20:50:20 cfriesen ah, I see. you're stopping it then deleting it
20:50:23 mriedem besides, yeah
20:50:49 mriedem i stop the service, delete it, and then have to null out the service uuid because as noted, deleting the service via the rest api looks it up and will migrate the uuid if not set
20:50:59 cfriesen let me just double-check my devstack
20:52:13 mriedem cfriesen: did you have https://review.openstack.org/#/c/545988/ in your devstack env?
20:52:30 openstackgerrit Merged openstack/nova stable/pike: libvirt: Block swap volume attempts with encrypted volumes prior to Queens https://review.openstack.org/543569
20:52:32 mriedem oh you said current stable/pike
20:53:42 openstackgerrit iain MacDonnell proposed openstack/nova master: Remove placement config check https://review.openstack.org/558089
20:56:01 mriedem well, if anything, i guess i can still leave this test in place for bug 1745977 so we have coverage of the scenario
20:56:01 openstack bug 1745977 in OpenStack Compute (nova) queens "When source compute service up, will not destroy and clean up those instances which be evacuated then be deleted." [Medium,Fix committed] https://launchpad.net/bugs/1745977 - Assigned to Matt Riedemann (mriedem)
20:57:08 cfriesen mriedem: when you list the servers, are you doing it as an admin user?
20:57:21 cfriesen mriedem: regular user doesn't get the host info
20:57:57 cfriesen mriedem: and so doesn't trigger the bug
20:58:41 cfriesen I just reproduced in devstack after restarting all the nova services, so should be valid
20:59:39 openstackgerrit melanie witt proposed openstack/nova master: Add periodic task to clean expired console tokens https://review.openstack.org/325381
20:59:40 openstackgerrit melanie witt proposed openstack/nova master: Use ConsoleAuthToken object to generate authorizations https://review.openstack.org/325414
20:59:41 openstackgerrit melanie witt proposed openstack/nova master: Convert websocketproxy to use db for token validation https://review.openstack.org/333990
20:59:42 openstackgerrit melanie witt proposed openstack/nova master: Update layout docs for running console proxies https://review.openstack.org/557489
20:59:44 cfriesen I wonder if you just replaced self.api.get_servers() at the bottom with self.admin_api.get_servers()...
20:59:45 mriedem cfriesen: no, not admin
21:00:09 mriedem i can try that
21:00:50 efried jaypipes: Quick +A on previously +2'd https://review.openstack.org/#/c/558089/ ?
21:04:26 mriedem there will be an obvious error in the n-cpu logs when placement isn't configured without that?
21:04:37 mriedem or will we just get the safe_connect warnings?
21:05:10 mriedem cfriesen: that doesn't change anything, listing with admin
21:05:47 efried mriedem: We should see MissingAuthPlugin
21:05:53 efried mriedem: But IMO safe_connect is still broken.
21:06:15 mriedem so we'll get https://github.com/openstack/nova/blob/master/nova/scheduler/client/report.py#L77
21:06:21 mriedem and nova-compute will be running and not reporting into placement
21:06:57 dansmith we shouldn't warn_limit that I think
21:07:04 efried mriedem: Yup, I believe that's right. You'll get that nag the first ten times (or whatever) and then it'll quit.
21:07:28 efried It's more broken than warn_limit can fix.
21:07:51 efried What's broken about it is that we don't raise an exception when it happens. We just let the decorated code continue.
21:08:08 dansmith yeah we should explode I think
21:08:14 mriedem well, for nova-compute sure
21:08:20 mriedem but for nova-api in rocky no
21:08:38 efried If safe_connect was used extremely carefully - only in the outermost scope of whatever is calling into placement - then it would be less broken.
21:08:55 mriedem i.e. https://review.openstack.org/#/c/560706/2/releasenotes/notes/bug-1679750-local-delete-allocations-cb7bfbcb6c36b6a2.yaml
21:09:09 efried But as it stands, it is theoretically possible for it to reconnect and then proceed, having skipped the first step of whatever that outer process was.
21:09:34 efried For this patch, it makes no difference - *nothing* will get done to placement. But for random midstream disconnects, it could potentially be disastrous.
21:11:07 efried I guess ^ only applies for the EndpointNotFound branch.
21:11:17 efried cause that's the only place we try to reconnect.
21:12:28 dansmith safe_connect was to make moving from no-placement to placement very smooth
21:12:36 dansmith it probably needs to change quite a bit in today's world
21:12:46 dansmith logging every single time we fail to talk to placement at least,
21:13:03 dansmith and probably raising a lot more so we don't fail to make changes we think we're making
21:14:31 mriedem so before rocky we never said that nova-api required config to talk to placement, are we now saying it does and if you don't have that config in place things are going to blow up?
21:14:51 mriedem specifically thinking about https://review.openstack.org/#/q/topic:bug/1756179+(status:open+OR+status:merged) and the mirror aggregates stuff
21:15:03 dansmith personally that doesn't bother me, but api-initiated code can be graceful about it
21:15:05 mriedem the former i care about more b/c i thought we would backport those
21:15:09 dansmith without needing safe_connect hiding everything from us
21:16:05 dansmith IMHO, ultimately the api aggregate mirroring stuff should be strict about not making local changes if it can't tell placement, so it needs good exception handling anyway
21:16:22 dansmith making it graceful now and strict later is easy if it's all up there
21:16:34 dansmith well, not all, but .. the handling of this not-configured case at least
21:16:34 mriedem so basically move safe_connect to api
21:16:43 openstackgerrit Merged openstack/nova master: libvirt: Report the allocated size of preallocated file based disks https://review.openstack.org/561704
21:16:45 mriedem and drop it in stein

Earlier   Later