| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-11-06 | |||
| 15:15:49 | mriedem | meaning if it's a new compute node, clear_provider_cache() so that when we hit the report client later, it will see it's empty and force the POST of the resource provider for the new compute node record? | |
| 15:18:57 | efried | mriedem: It won't force the POST necessarily. It'll do the GET /rps?in_tree - which we would otherwise skip if it was already in the cache - and iff that comes up empty, it'll do the POST to create. | |
| 15:19:32 | mriedem | yeah i left comments, i think i understand | |
| 15:19:50 | mriedem | also, nice coincidental use of iff https://review.openstack.org/#/c/614037/2/nova/scheduler/host_manager.py@679 | |
| 15:31:20 | mriedem | i just came in hot and late and dumped some comments on a few changes in that series | |
| 15:36:23 | gibi | mriedem: now I managed to read back the diffs in the two bandwidth patch you fixed up and approved. Thank you for making those changes. They look good to me in retrospect | |
| 15:36:45 | mriedem | cool, yw | |
| 15:36:57 | gibi | mriedem: I think I this week finally I will have time to pick up that patch series again. | |
| 15:44:56 | efried | mriedem: https://review.openstack.org/#/c/615646/ see response please. I may be misunderstanding some side effect of making the conf option mutable. | |
| 15:47:43 | mriedem | hmm yeah i guess in addition to changing the refresh cache interval, we also need to clear the in-memory "cache" in the compute | |
| 15:48:30 | mriedem | dropped the -1 | |
| 15:48:37 | efried | correct. I agree making the conf option mutable (in the bottom patch) is a good idea. | |
| 15:48:41 | mriedem | we should make the refresh interval option mutable somewhere in this series shouldn't we? | |
| 15:48:45 | mriedem | ok | |
| 15:48:53 | efried | mriedem: unless you think that should be separated. | |
| 15:49:21 | mriedem | i'm fine with it either going in the bottom change that adds min=0 or the sighup change to clear the cache | |
| 15:49:50 | mriedem | it might make more logical sense to be in the sighup change itself | |
| 15:49:53 | mriedem | but idk, not a big deal probably | |
| 15:51:25 | efried | It was tempting to do both of those things at the same time, since (rats, I forgot to do this...) I'm going to update the conf option help and the reno text in the SIGHUP patch anyway. | |
| 16:02:39 | openstackgerrit | Jay Pipes proposed openstack/nova master: prevent common kwargs from glance client failure https://review.openstack.org/614351 | |
| 16:05:09 | jaypipes | efried: ^^ took your advice and re-worked that patch. thanks again for the great idea. | |
| 16:05:23 | efried | jaypipes: cool mon, will look. | |
| 16:05:49 | efried | jaypipes: Have you seen my -1KLOC rt patch yet? | |
| 16:06:09 | efried | (that's a minus :) | |
| 16:09:23 | jaypipes | efried: it's in the queue... | |
| 16:09:41 | efried | jaypipes: It's not quite finished yet, just thought it would bring a smile to your face. | |
| 16:15:20 | mriedem | melwitt: is there any reason we don't call _clean_instance_console_tokens when we shelve offload from a host? | |
| 16:15:34 | mriedem | i guess we don't *have* to because the _cleanup_expired_console_auth_tokens periodic should clean them up? | |
| 16:16:07 | openstackgerrit | Merged openstack/nova master: Add bandwidth related standard resource classes https://review.openstack.org/570847 | |
| 16:26:05 | melwitt | mriedem: I think the only reason is because when we moved from consoleauth => database backend we didn't introduce any behavior changes (didn't add any additional token removal sites other than the periodic). there were some sites during the review where we said, "there should probably be token removal here" | |
| 16:32:59 | efried | stephenfin: https://review.openstack.org/#/c/614351/ is back atcha, when you get a chance. | |
| 16:41:17 | efried | mriedem, jaypipes: If we're gonna make the default "no refreshing", why not just remove the conf option and all unprompted periodic refreshing entirely? | |
| 16:43:40 | mriedem | because i continue to think there was originally some desire to allow external services to frequently adjust things that might affect what we report, like inventory and traits, i.e. powervm and vcenter things | |
| 16:43:55 | mriedem | that are trying to sync information back from those "hypervisors" to nova | |
| 16:44:51 | mriedem | if there aren't, and the refresh was just purely a bridge b/c that's how the RT always worked before placement, then i'm a big meh on keeping the refresh stuff | |
| 16:44:57 | efried | mriedem: Well, anything that happens in upt will get synced back right away. | |
| 16:45:13 | efried | and we already allow tailoring the interval that runs upt | |
| 16:45:42 | mriedem | so external changes from let's say vcenter for inventory/traits would just get synced back via the driver -> RT -> reportclient -> placement | |
| 16:46:02 | efried | assuming the driver is set up to detect them and push them into the provider tree, yeah. | |
| 16:46:31 | efried | but actually | |
| 16:47:02 | cdent | from the vcenter side there some slim chance that what might be considered weird resource providers will be being created within the virt driver during upt. If that's cool and upt is happening regularly, then goorvy | |
| 16:47:05 | efried | if they're truly oob, then the next time upt thinks something *else* has changed, its update will bounce 409 (due to the oob change) and we'll clear the cache and go again on the next periodic. | |
| 16:47:30 | efried | but only if upt thinks something else has changed, and that something else happens to be on the same provider that changed oob. So that's not a reliable model. | |
| 16:47:50 | efried | No, we have to count on the virt driver being the sole source of truth OR the oob changes being followed with SIGHUP. Unless we keep the refresh. | |
| 17:30:59 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add Destination.cross_cell_move field https://review.openstack.org/614035 | |
| 17:30:59 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add Migration.cross_cell_move and get_by_uuid https://review.openstack.org/614012 | |
| 17:31:00 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Change HostManager to allow scheduling to other cells https://review.openstack.org/614037 | |
| 17:31:00 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add InstanceAction/Event create() method https://review.openstack.org/614036 | |
| 17:31:01 | openstackgerrit | Matt Riedemann proposed openstack/nova master: WIP: Cross-cell resize https://review.openstack.org/603930 | |
| 17:31:01 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add CrossCellWeigher https://review.openstack.org/614353 | |
| 17:36:28 | bauzas | mriedem: mmm, I'm working on a customer issue in Newton, and AFAICS we don't call the late check for anty-affinity in the compute service if it's a live-migration :( | |
| 17:36:44 | bauzas | mriedem: looks like we don't do this even now | |
| 18:02:29 | efried | mriedem: my brain finally stumbled across that powervm case you've been hinting at | |
| 18:03:02 | efried | I've kept saying we don't need to know about changes made outside of the virt driver, unless made by hand in which case SIGHUP. | |
| 18:03:24 | efried | But that doesn't account for OOB changes made *in* the virt driver, on a *different* compute host. | |
| 18:03:33 | efried | Like the way I've been saying we were going to do shared storage providers. | |
| 18:04:52 | efried | So yeah, still going to wait to burn that bridge when I cross it, but I think the operator's process would have to be something like: Add the SSP, then SIGHUP all the computes to make them recognize it. | |
| 18:19:40 | mriedem | bauzas: correct, we've never had a late affinity check for live migratoin | |
| 18:19:52 | mriedem | or resize or unshelve for that matter | |
| 18:20:04 | mriedem | only initial create and evacuate (since i think queens/rocky) | |
| 19:10:46 | brandor5 | hello everyone: I'm trying to deploy some new baremetal instances and have them land on specific pieces of hardware... I've accomplished this in the past by using scheduler_hints for hypervisor_hostname... however this isn't working currently and I'm not sure why... I'm not seeing anything in the scheduler logs that doesn't look normal... where else should I look? | |
| 19:14:48 | mriedem | by scheduler hint you mean the --availability-zone ZONE:HOST:NODE trick? | |
| 19:14:58 | mriedem | which release? | |
| 19:15:21 | brandor5 | mriedem: newton (rhosp10) | |
| 19:15:33 | brandor5 | and I'm using this: query='[\"=\",\"$hypervisor_hostname\",\"59c8dcfa-57d5-49ee-a366-664b11f39620\"]'" | |
| 19:16:18 | mriedem | so the JsonFilter | |
| 19:16:31 | mriedem | don't really know then, it's not tested very well outside of unit tests as far as i know | |
| 19:16:32 | brandor5 | 59c8dcfa... being a baremetal node in ironci | |
| 19:16:34 | mriedem | plus newton is eol | |
| 19:16:54 | mriedem | this isn't really the channel for that type of support question. having said all that, the simpler way to do that is just the az trick i mentioned above | |
| 19:17:07 | mriedem | --availability-zone nova::59c8dcfa-57d5-49ee-a366-664b11f39620 | |
| 19:17:24 | mriedem | https://docs.openstack.org/nova/latest/admin/availability-zones.html | |
| 19:17:32 | brandor5 | when was that introduced? | |
| 19:17:53 | mriedem | forever ago | |
| 19:17:58 | mriedem | it'd be in newton | |
| 19:18:05 | brandor5 | k, will try that to see what happens, thanks | |
| 19:18:09 | mriedem | yw | |
| 19:20:45 | efried | melwitt: In case you're not watching #openstack-oslo: https://review.openstack.org/615978 | |
| 21:14:56 | openstackgerrit | sean mooney proposed openstack/os-vif master: always create ovs port during plug https://review.openstack.org/602384 | |
| 21:14:57 | openstackgerrit | sean mooney proposed openstack/os-vif master: add isolate_vif config option https://review.openstack.org/612534 | |
| 21:23:02 | openstackgerrit | Merged openstack/nova master: Trivial: add reminder to update Tempest's scheduler_enabled_filters https://review.openstack.org/615436 | |
| 21:23:27 | openstackgerrit | Merged openstack/nova master: prevent common kwargs from glance client failure https://review.openstack.org/614351 | |
| 21:59:04 | openstackgerrit | Adam Spiers proposed openstack/nova-specs master: Add spec for libvirt driver launching AMD SEV-encrypted instances https://review.openstack.org/609779 | |
| 22:43:16 | tomtom001 | lyarwood: hello I was wondering you could say how the encryption on nova works with nova libvirt and ceph... we've noticed really slow performance when using luks encryption... I've turned off xts,and only use aes but it doesn't seem to change anything, what specifically needs to be in place to get a fast encryption for disks? | |
| 22:51:10 | openstackgerrit | Eric Fried proposed openstack/nova master: Remove LazyLoad of Scheduler Clients https://review.openstack.org/615641 | |
| 22:51:10 | openstackgerrit | Eric Fried proposed openstack/nova master: Allow resource_provider_association_refresh=0 https://review.openstack.org/615606 | |
| 22:51:11 | openstackgerrit | Eric Fried proposed openstack/nova master: WIP: Reduce calls to placement from _ensure https://review.openstack.org/615677 | |
| 22:51:11 | openstackgerrit | Eric Fried proposed openstack/nova master: SIGHUP n-cpu to refresh provider tree cache https://review.openstack.org/615646 | |
| 22:51:12 | openstackgerrit | Eric Fried proposed openstack/nova master: WIP: Commonize _update code path https://review.openstack.org/615705 | |
| 22:51:12 | openstackgerrit | Eric Fried proposed openstack/nova master: WIP: Remove redundant inventory refresh https://review.openstack.org/615695 | |
| 22:59:31 | openstackgerrit | sean mooney proposed openstack/os-vif stable/rocky: always create ovs port during plug https://review.openstack.org/609850 | |
| 23:14:54 | openstackgerrit | Sundar Nadathur proposed openstack/nova-specs master: Nova Cyborg interaction specification. https://review.openstack.org/603955 | |
| 23:19:59 | openstackgerrit | Eric Fried proposed openstack/nova master: Consolidate inventory refresh https://review.openstack.org/615695 | |
| 23:20:00 | openstackgerrit | Eric Fried proposed openstack/nova master: WIP: Commonize _update code path https://review.openstack.org/615705 | |
| 23:33:39 | openstackgerrit | Eric Fried proposed openstack/nova master: WIP: Reduce calls to placement from _ensure https://review.openstack.org/615677 | |
| 23:33:40 | openstackgerrit | Eric Fried proposed openstack/nova master: WIP: Commonize _update code path https://review.openstack.org/615705 | |
| 23:33:40 | openstackgerrit | Eric Fried proposed openstack/nova master: Consolidate inventory refresh https://review.openstack.org/615695 | |
| 23:46:53 | mgagne | question: How can I determine in which cell a service is running? And also, duplicated ID? and my hostname are nova01.example.org and nova01.c1.example.org https://gist.github.com/mgagne/322f8cb5b9f9a3130006b2f46b52c4d6 | |
| 23:49:25 | openstackgerrit | Eric Fried proposed openstack/nova master: WIP: Turn off rp association refresh in nova-next https://review.openstack.org/616033 | |
| 23:49:41 | efried | mriedem: Does this look right? ^ | |