Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2018-04-18
14:59:50 arvindn05 bauzas: yup...it does muddy the waters...but still not seeing how it would effect my changes though. how does instance.save() call effect my flow?
15:01:10 bauzas arvindn05: your flow is about adding an extra check to the scheduler which eventually calls placement for traits query, right?
15:02:37 openstackgerrit sahid proposed openstack/nova master: compute: introduce cpu_shared_set option https://review.openstack.org/561912
15:02:38 openstackgerrit sahid proposed openstack/nova master: libvirt: always pass emulator threads policy https://review.openstack.org/561913
15:02:39 openstackgerrit sahid proposed openstack/nova master: libvirt: place emulator threads on CONF.cpu_shared_set https://review.openstack.org/510897
15:02:48 arvindn05 bauzas: kindof. The extra check calls the GET /resource_providers/{hostUUID}/traits to get traits and match them against the image
15:03:51 arvindn05 so we dont call placemnt allocation candiadates...but make a query to find the traits of the current host and match them against the image traits
15:04:39 arvindn05 bauzas: essentially its this check set(image_traits_required).issubset(set(current_host.traits))
15:07:07 bauzas arvindn05: sure, but that restricts the subset of hosts right?
15:07:31 bauzas arvindn05: so a host that'd be accepted in the past wouldn't be now accepted, leading to possibly more NoValidHosts
15:08:52 arvindn05 bauzas: we are not checking agaisnt a subset of hosts though...only the current host.
15:09:24 arvindn05 we would get novalid hosts only if the current host the instance is residing upon does not have the traits
15:10:12 arvindn05 the hostuuid in the RP call basically comes from the resource spec `spec_obj.force_hosts[0], spec_obj.force_nodes[0]`
15:10:25 arvindn05 *request spec
15:10:58 bauzas arvindn05: okay here is the thign
15:11:01 bauzas I need to disappear
15:11:07 bauzas but I'll consider your spec
15:11:09 cfriesen this may be a weird question, but why does nova have entries for other services in nova.conf rather than using the keystone endpoint list?
15:11:12 bauzas and comment it
15:11:24 arvindn05 bauzas: no problem...thanks.
15:15:31 melwitt gibi: hey, I'm going to be around early tomorrow, so I'll be able to run the meeting. thanks for the offer :)
15:23:04 efried cfriesen: What "entries"?
15:23:19 efried cfriesen: The auth stuff is necessary on a per-service basis, obviously.
15:24:23 efried cfriesen: To identify endpoints, we should be able to use the service catalog for everything other than cinder at this point. I don't think we've fixed up devstack to do that for everything, but I know we've at least hit glance.
15:24:48 gibi melwitt: OK, cool.
15:25:16 fghaas lyarwood: I'll have to drop off shortly for family obligations — I've requested a review from you on https://review.openstack.org/#/c/562145/ which kashyap mentioned earlier; hope that's OK with you. No rush on this at all of course; we do have a workaround in place.
15:25:28 efried Where's ol mriedem today?
15:25:29 openstackgerrit Kashyap Chamarthy proposed openstack/nova master: libvirt: Bump MIN_{LIBVIRT,QEMU}_VERSION for "Rocky" https://review.openstack.org/558783
15:25:30 openstackgerrit Kashyap Chamarthy proposed openstack/nova master: Drop MIN_LIBVIRT_SET_ADMIN_PASSWD https://review.openstack.org/562312
15:25:31 openstackgerrit Kashyap Chamarthy proposed openstack/nova master: libvirt: Drop MIN_LIBVIRT_PARALLELS_SET_ADMIN_PASSWD https://review.openstack.org/562313
15:25:35 gibi melwitt, efried: I've got the approval today to travel to the Vancouver summit so I can help with the onboarding session
15:25:43 efried cool
15:26:14 melwitt gibi: woot, coolness
15:26:40 gibi melwitt, efried: If you have any material about past onboarding sessions then link it to me so I can try thinkig about what I will talk about there
15:29:31 efried dansmith: Should this return True for the 404 case? https://github.com/openstack/nova/blob/master/nova/scheduler/client/report.py#L1849-L1855
15:30:34 efried dansmith: I noticed the return value is only checked in this call: https://github.com/openstack/nova/blob/master/nova/compute/manager.py#L3799-L3808
15:30:40 gibi melwitt: do we plan to have a nova-neutron cross project discussion on the summit?
15:31:00 efried dansmith: So I guess my question is, can/should we still short-circuit there if the instance is already gone?
15:31:10 melwitt gibi: yes, we have this http://forumtopics.openstack.org/cfp/details/124
15:31:45 gibi melwitt: cool, it even contains the bandwidth topic I wanted to add
15:32:48 melwitt gibi: as for past onboarding session material, I'm not sure anyone has anything saved. last time's session was ad hoc, and the time before I wasn't helping with (I was in the room for only the end of it). dansmith, is there any linkable material from the boston onboarding session that gibi could look at?
15:32:55 dansmith efried: well, does 404 there always refer to the allocation? or could it refer to the RP?
15:33:27 dansmith melwitt: we posted an annotated log of an instance boot in devstack
15:33:41 efried dansmith: It refers to the consumer being gone.
15:33:47 dansmith melwitt: that might still exist, but it's probably worth re-generating it at this point, if you're going to do that same schtick
15:34:00 dansmith efried: you mean the allocation yeah/
15:34:35 gibi dansmith, melwitt: thanks
15:34:36 melwitt gibi: this time I'm going to try to prepare some slides in case we need them. apparently this is the sort of thing the foundation recommends we do or at least have available if the room is total newcomers https://www.slideshare.net/JayBryant2/openstack-cinder-onboarding-education-boston-summit-2017
15:35:16 efried dansmith: Looking at what compute manager does after that, it'll try to get that allocation again, and short-circuit out at *that* point.
15:35:41 dansmith efried: tbh, I would still raise a NotFound there instead of returning a boolean, but it should return False, since the code is using that to determine if the other side of the migration was new or not.. "it's already gone" is not what it's checking for
15:35:43 efried dansmith: So the harm is only one redundant placement call (and an info log).
15:36:20 dansmith efried: see above.. this is not just trying to nuke the allocation, it needs to know whether it existed or not
15:36:26 cfriesen efried: I'm thinking of the [glance]/api_servers or [neutron]/url
15:36:55 efried dansmith: Okay. So the report client method is doing the right thing; it's just that that thing is a bit confusing.
15:37:18 dansmith efried: well, like I said, if it raised NotFound there and we were catching it, I think it might be a little more obvious
15:37:29 efried dansmith: The docstring would say:
15:37:29 efried :return: True if the instance's allocations were deleted; False if they were not (which is also the case if the allocations weren't there to begin with)
15:37:31 cfriesen efried: although, I guess those are now optional rather than required
15:37:38 dansmith efried: instead of "did we delete it?" we'd be saying "We deleted it" or "it didn't exist" which is what we're really after
15:37:59 efried cfriesen: Correct.
15:38:24 efried cfriesen: deprecated, in the latter case. In the former case, only use api_servers if you actually *need* multiple endpoints.
15:38:47 efried cfriesen: There was work over the past couple of releases to make that happen. I can find the bp if you like...
15:39:18 efried cfriesen: https://review.openstack.org/#/q/topic:bp/use-ksa-adapter-for-endpoints
15:39:52 efried cfriesen: Note the one that's still open. I would welcome someone jumping in and finishing that one...
15:39:53 cfriesen efried: thanks
15:41:37 openstackgerrit Merged openstack/nova master: Fix docs for confirmResize action https://review.openstack.org/562062
15:42:32 gibi melwitt: OK. If you need help perparing or reviewing such slideset then I'm happy to help
15:43:06 melwitt thanks
15:43:07 openstackgerrit Nguyen Hai proposed openstack/nova-specs master: Follow the new PTI for document build https://review.openstack.org/551802
15:49:36 arvindn05 jaypipes: wanted to get your thoughts on role of placement during rebuild
15:50:07 openstackgerrit Nguyen Hai proposed openstack/nova-specs master: Follow the new PTI for document build https://review.openstack.org/551802
15:50:59 openstackgerrit Nguyen Hai proposed openstack/nova-specs master: Follow the new PTI for document build https://review.openstack.org/551802
15:56:31 jaypipes arvindn05: currently deep in the middle of teasing apart the @safe_connect decorator from the placement report client... may be some time until I can get to that :(
15:59:29 arvindn05 ahh..ok...will bounce the idea off matt for now :)
16:17:42 openstackgerrit Zhenyu Zheng proposed openstack/nova master: Deduplicate notification samples Rocky - 7 https://review.openstack.org/560748
16:42:02 openstackgerrit Jay Pipes proposed openstack/nova master: mock utils.execute() in qemu-img unit test https://review.openstack.org/562339
16:44:27 cfriesen melwitt: had a thought on bug 1764556. If we deleted the compute node, recreated it, and migrated the instance back prior to the upgrade to Pike I think we could end up with the DB in the problem state. On the other hand, that seems like a pretty unlikely scenario.
16:44:27 openstack bug 1764556 in OpenStack Compute (nova) ""nova list" fails with exception.ServiceNotFound if service is deleted and has no UUID" [Undecided,Fix released] https://launchpad.net/bugs/1764556
16:47:50 melwitt cfriesen: what would be the full sequence there? sorry, I forgot what the original sequence was regarding deleting of the service, migration, re-create service, migrate etc
16:48:24 cfriesen melwitt: boot instance, migrate instance, delete compute node, create new compute node with same name, migrate instance to new compute node, upgrade to pike
16:50:49 melwitt okay. does seem unlikely but I wouldn't be surprised if someone someday would happen to do it
16:54:04 cfriesen melwitt: our local solution is to modify the online data migration in pike to put uuids on deleted services
16:54:30 openstackgerrit Merged openstack/os-vif master: Add lower-constraints job https://review.openstack.org/556056
16:55:32 melwitt ack
16:56:24 phuongnh hi all, I would like to add one more line in file https://review.openstack.org/#/c/462759/3/os_traits/hw/cpu/x86.py for CPU_FPGA
16:56:44 phuongnh the link to this cpu type is: https://en.wikipedia.org/wiki/Field-programmable_gate_array
16:56:51 bmace does anyone know or have any pointers to why stable/queens global-requirements is pinned to version 9.1 of python-novaclient when 10.1 seems to be the queens version of python-novaclient from what I can see?
16:56:56 phuongnh Do I need to propose a blueprint in launchpad?
16:57:45 melwitt jaypipes: what's the procedure for proposing a new cpu type to os-traits? ^
16:58:28 jaypipes phuongnh: that's not a CPU type.
16:58:48 jaypipes melwitt: the procedure to add any trait is simply to propose a patch to os-traits.
16:58:57 jaypipes melwitt: the issue here is that CPU_FPGA isn't a trait.
16:59:15 melwitt thanks
16:59:27 phuongnh thanks jaypipes
17:01:28 melwitt bmace: I think there was a problem with 10.1 not working with queens openstackclient. I have to look to see where/how that was noted (probably in some review somewhere)
17:02:46 bmace melwitt: if you have more info or it was just certain projects that would be very helpful. i think without it there are some issues getting multi-attach to work correctly through heat.
17:04:40 melwitt bmace: okay, let me look into it and get back to you. I think mriedem might be able to point to some details about it but he's not around at the moment
17:05:14 bmace melwitt: sure thing, thanks very much for your time looking into it.
17:24:20 openstackgerrit sahid proposed openstack/nova master: compute: introduce cpu_shared_set option https://review.openstack.org/561912
17:48:29 cfriesen melwitt: bmace: I think this might be it: https://review.openstack.org/#/c/538070/

Earlier   Later