| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2019-09-06 | |||
| 09:31:16 | openstackgerrit | Merged openstack/nova master: Add delete_on_termination to volume-attach API https://review.opendev.org/673133 | |
| 09:44:27 | dtantsur | morning folks, we seem to have broken ironic with the recent nova patches | |
| 09:45:11 | cdent | dtantsur: sounds fun | |
| 09:46:52 | dtantsur | as always! | |
| 09:50:08 | cdent | when did it stop working? | |
| 09:51:24 | dtantsur | cdent: it's because of https://review.opendev.org/#/c/656027/, a fix is already posted | |
| 09:51:41 | dtantsur | efried: and I also wonder why our regular job did not detect ^^ | |
| 09:51:41 | cdent | ah | |
| 09:51:58 | dtantsur | maybe list_instances is not normally hit or is not critical? | |
| 09:55:49 | cdent | i would guess that most of the time individual instances are retrieved | |
| 09:55:52 | cdent | but that's totally a guess | |
| 10:00:33 | openstackgerrit | hulina proposed openstack/nova master: Nova raise exceptions when extending volume fails https://review.opendev.org/680648 | |
| 10:09:52 | artom | stephenfin, sean-k-mooney, can I get your thoughts on a thing? | |
| 10:10:01 | artom | You know this bit: https://github.com/openstack/nova/blob/master/nova/virt/libvirt/driver.py#L4525-L4528 | |
| 10:10:42 | artom | Normally on a host not intended for NUMA instances, https://github.com/openstack/nova/blob/master/nova/virt/libvirt/driver.py#L4549 will return None because vcpu_pin_set will not be set | |
| 10:11:51 | stephenfin | looking | |
| 10:12:21 | stephenfin | artom: Yeah, correct | |
| 10:12:33 | artom | If, however, an instance without a NUMA topology lands on a host with vcpu_pin_set not None, guest.cpuset being not None ends up triggering https://github.com/openstack/nova/blob/master/nova/virt/libvirt/config.py#L2587-L2590 and we end up pinning guest CPUs that way | |
| 10:13:05 | artom | (via https://libvirt.org/formatdomain.html#elementsCPUAllocation, not our usual vcpupin XML bits) | |
| 10:14:19 | artom | Do we care about *that* form of CPU pinning (the one via <vcpu cpuset=blah>) for live migration? | |
| 10:14:22 | stephenfin | artom: um, I think that's correct, yeah | |
| 10:14:27 | artom | Theoretically we should be recalculating it | |
| 10:14:36 | bauzas | now I'm a bit afraid of my application be rejected because of some coffee traces on my passport :/ | |
| 10:14:39 | stephenfin | yup | |
| 10:14:43 | stephenfin | we should | |
| 10:14:51 | artom | But... in practice, does it actually happen? | |
| 10:14:56 | stephenfin | ohh | |
| 10:15:01 | stephenfin | hmm, I'm not actually sure | |
| 10:15:05 | artom | We tell everyone to split NUMA instances into host aggregates | |
| 10:15:17 | bauzas | stephenfin: I also think we should correctly document the weird case you mention in https://review.opendev.org/#/c/680107/3 | |
| 10:15:19 | stephenfin | I had this written off as being fixed by your series | |
| 10:15:29 | artom | So a non-NUMA instances landing on a host with vcpu_pin_set is... unlikely? | |
| 10:15:38 | artom | stephenfin, hah, I've just noticed that it's a thing | |
| 10:15:42 | bauzas | stephenfin: because the behaviour will change for those not RTFM'ing | |
| 10:15:53 | stephenfin | bauzas: Nope, it won't change | |
| 10:15:56 | stephenfin | *Break | |
| 10:15:58 | stephenfin | bauzas: see here | |
| 10:16:24 | stephenfin | bauzas: https://review.opendev.org/#/c/680107/3/nova/compute/manager.py@861 | |
| 10:16:28 | stephenfin | I'm only warning | |
| 10:16:32 | stephenfin | Not raising | |
| 10:17:01 | stephenfin | bauzas: For those other two conditionals, not that I'm only checking the _new_ options (cpu_dedicated_set, cpu_shared_set) | |
| 10:17:03 | bauzas | see my comment | |
| 10:17:05 | stephenfin | *note | |
| 10:17:18 | stephenfin | Yup. The conditional is here https://review.opendev.org/#/c/680107/3/nova/compute/manager.py@827 | |
| 10:17:29 | openstackgerrit | Luyao Zhong proposed openstack/nova master: db: Add resources column in instance_extra table https://review.opendev.org/678447 | |
| 10:17:30 | openstackgerrit | Luyao Zhong proposed openstack/nova master: object: Introduce Resource and ResourceList objs https://review.opendev.org/678448 | |
| 10:17:31 | openstackgerrit | Luyao Zhong proposed openstack/nova master: Retrieve the allocations early https://review.opendev.org/678450 | |
| 10:17:31 | openstackgerrit | Luyao Zhong proposed openstack/nova master: Add resources dict into _Provider https://review.opendev.org/678449 | |
| 10:17:32 | openstackgerrit | Luyao Zhong proposed openstack/nova master: libvirt: report VPMEM resources by provider tree https://review.opendev.org/678454 | |
| 10:17:32 | openstackgerrit | Luyao Zhong proposed openstack/nova master: libvirt: Enable driver discovering PMEM namespaces https://review.opendev.org/678453 | |
| 10:17:32 | openstackgerrit | Luyao Zhong proposed openstack/nova master: Claim resources in resource tracker https://review.opendev.org/678452 | |
| 10:17:33 | openstackgerrit | Luyao Zhong proposed openstack/nova master: Parse vpmem related flavor extra spec https://review.opendev.org/678456 | |
| 10:17:33 | openstackgerrit | Luyao Zhong proposed openstack/nova master: libvirt: Support VM creation with vpmems and vpmems cleanup https://review.opendev.org/678455 | |
| 10:17:34 | openstackgerrit | Luyao Zhong proposed openstack/nova master: libvirt: Enable driver configuring PMEM namespaces https://review.opendev.org/679640 | |
| 10:17:35 | openstackgerrit | Luyao Zhong proposed openstack/nova master: doc: attaching virtual persistent memory to guests https://review.opendev.org/680300 | |
| 10:17:35 | openstackgerrit | Luyao Zhong proposed openstack/nova master: Add functional tests for virtual persistent memory https://review.opendev.org/678470 | |
| 10:17:37 | stephenfin | and I early return here https://review.opendev.org/#/c/680107/3/nova/compute/manager.py@839 | |
| 10:18:08 | stephenfin | so lines 841 etc. will only be triggered if the instance has a NUMA topology and that topology has pinning enabled | |
| 10:18:23 | bauzas | oh shit, that's what happen when you review in a crowded Starbucks place | |
| 10:18:31 | bauzas | you're totally right | |
| 10:18:46 | stephenfin | bauzas: but at least you've mediocre coffee on hand :P | |
| 10:18:57 | stephenfin | If it's not clear, I can rework to use an if-else | |
| 10:20:59 | openstackgerrit | Merged openstack/nova master: update allocation in binding profile during migrate https://review.opendev.org/656422 | |
| 10:21:22 | bauzas | stephenfin: nah nah, it's cool | |
| 10:21:42 | bauzas | stephenfin: we could nitpick on getting an extra comment, but meh | |
| 10:22:28 | bauzas | and yeah Starbucks is mediocre, but that's the only place where you can guaranttee both decent wifi coverage and working outlets | |
| 10:22:53 | stephenfin | In the code or the commit message? I can do the latter in a follow-up | |
| 10:23:08 | stephenfin | Yeah, agreed. Costa just isn't the same (and it's UK only) | |
| 10:23:15 | bauzas | I wish I could afford a Regus usage :) | |
| 10:23:46 | bauzas | stephenfin: well, I was thinking of just a comment line in the code itself but that's gross | |
| 10:23:49 | bauzas | nevermind it | |
| 10:23:58 | bauzas | it's just me who can't read coee | |
| 10:24:01 | bauzas | code* | |
| 10:24:29 | openstackgerrit | Brin Zhang proposed openstack/python-novaclient master: Microversion 2.79: Add delete_on_termination to volume-attach API https://review.opendev.org/673485 | |
| 10:24:47 | stephenfin | 'sall good :D | |
| 10:36:35 | stephenfin | artom: how do you decide to recalculate the '<cpuset>' values on a destination host in your series? | |
| 10:37:01 | artom | stephenfin, I currently don't, at all | |
| 10:37:21 | sean-k-mooney | you just always recaluate it right | |
| 10:37:25 | sean-k-mooney | as part of the claim | |
| 10:37:53 | artom | sean-k-mooney, no, currently it's not handled at all | |
| 10:38:02 | sean-k-mooney | which cpuset | |
| 10:38:05 | artom | But in theory it should happen when https://github.com/openstack/nova/blob/master/nova/virt/libvirt/driver.py#L4525-L4528 | |
| 10:38:15 | artom | sean-k-mooney, the one in https://libvirt.org/formatdomain.html#elementsCPUAllocation | |
| 10:38:32 | sean-k-mooney | oh that should be the union of hte cores used for the vcpus | |
| 10:38:39 | sean-k-mooney | and should be updated | |
| 10:38:42 | artom | It only matters when an instance without a NUMA topology lands on a host with vcpu_pin | |
| 10:38:46 | stephenfin | alex_xu: You around today? I'm planning to work through the VPMEM series again, but is there any chance you could look at https://review.opendev.org/#/c/671800 ? | |
| 10:38:53 | sean-k-mooney | artom: yes | |
| 10:39:03 | artom | sean-k-mooney, right, so is that... a real thing? | |
| 10:39:10 | sean-k-mooney | yes | |
| 10:39:10 | artom | We tell people to split them by aggregates | |
| 10:39:17 | stephenfin | artom: wait for it | |
| 10:39:23 | artom | They don't? | |
| 10:39:29 | sean-k-mooney | you shoudl set vcpu_pin_set on all host ideally | |
| 10:39:37 | stephenfin | (sean-k-mooney is about to tell you that vcpu_pin_set has nothing to do with pinning and should always be set) | |
| 10:39:39 | sean-k-mooney | and it will be need for the new stuff stephenfin is working on | |
| 10:39:53 | sean-k-mooney | stephenfin: :) | |
| 10:39:53 | artom | Haha | |
| 10:40:18 | openstackgerrit | Gorka Eguileor proposed openstack/nova master: Use os-brick locking for volume attach and detach https://review.opendev.org/614190 | |
| 10:40:19 | stephenfin | but yeah, vcpu_pin_set != pinning (or NUMA, for that matter) | |