| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-01-15 | |||
| 15:10:10 | bauzas | dansmith: mdbooth: FYI, just posted a new revision for the VGPU series starting with https://review.openstack.org/#/c/528832/8 that now lookups existing instances when restarting nova-compute | |
| 15:10:38 | bauzas | dansmith: mdbooth: that said, I still need to work around the XML removal when rebooting/restarting/etc. | |
| 15:11:11 | bauzas | wondering if that could be a follow-up patch | |
| 15:11:49 | dansmith | bauzas: yeah I think we just can't merge the first one until that follow-up is ready behind it | |
| 15:12:04 | bauzas | k | |
| 15:12:59 | mdbooth | bauzas: What dansmith said. | |
| 15:19:13 | openstackgerrit | Carlos Goncalves proposed openstack/os-vif master: Git ignore .stestr https://review.openstack.org/533699 | |
| 15:49:03 | lyarwood | stephenfin: https://review.openstack.org/#/c/527046/ - any chance of a +W? Had to rebase over the weekend. | |
| 15:50:55 | dansmith | bauzas: that recreate on boot patch is nice.. and pretty simple.. too simple not to do :) | |
| 15:51:18 | bauzas | dansmith: I'll need to fully rewrite the bottom change | |
| 15:51:38 | stephenfin | lyarwood: Sure, done | |
| 15:51:48 | dansmith | bauzas: because why? | |
| 15:52:07 | bauzas | dansmith: given I'm changing all the interfaces by not passing the allocations to _get_guest_xml() but rather asks to allocate a mdev for spawn() and passing it to _get_guest_xml() | |
| 15:52:14 | dansmith | bauzas: ah okay | |
| 15:52:23 | lyarwood | stephenfin: thanks | |
| 15:52:29 | openstackgerrit | Carlos Goncalves proposed openstack/os-vif master: WIP: vif_plug_ovs: add should_provide_port https://review.openstack.org/533713 | |
| 15:52:34 | bauzas | dansmith: so for hard_reboot(), I'll instead ask for the current mdev for passing it to _get_guest_xml() | |
| 15:53:15 | dansmith | bauzas: aye | |
| 15:58:11 | edleafe | efried: thanks for that huge fix on https://review.openstack.org/#/c/526436/ :) | |
| 15:58:42 | efried | It was an enormous sacrifice of time and mental anguish, but you're worth it edleafe | |
| 15:58:57 | edleafe | Your suffering has not gone unnoticed | |
| 15:59:11 | efried | edleafe Caught up on sched meeting yet? | |
| 15:59:36 | edleafe | no, just poured me a cup of coffee and was getting ready to read back | |
| 16:00:14 | efried | ight. There was some question of how you felt about ProviderTree.snapshot (formerly known as ReadOnlyProvider) | |
| 16:01:11 | bauzas | jaypipes: I think we already discussed about that previously, but someone pinged me about reviewing https://review.openstack.org/#/c/526900/19/nova/db/sqlalchemy/models.py | |
| 16:01:18 | bauzas | jaypipes: would it be a performance issue ? | |
| 16:01:30 | jaypipes | bauzas: no | |
| 16:02:01 | jaypipes | bauzas: shouldn't be a performance issue at all. | |
| 16:02:04 | bauzas | ok, I remembered your point, but I wasn't sure it was for VARCHAR(255) > TEXT | |
| 16:02:23 | bauzas | ok, will review it when I have time then | |
| 16:03:32 | jaypipes | bauzas: that said, doesn't the REST API limit the length of the metadata value? | |
| 16:03:44 | jaypipes | bauzas: in which case a microversion is gonna be necessary | |
| 16:03:52 | bauzas | jaypipes: yup, possibly | |
| 16:04:02 | bauzas | tbh, I haven't reviewed it | |
| 16:04:11 | bauzas | just looking at the main modification | |
| 16:04:39 | bauzas | jaypipes: for your point, that'll depend if we cut the string by the API or not | |
| 16:05:10 | bauzas | but yeah, if it's all about having the API giving more than 255 chars for that, then yes, it's a microversion needed | |
| 16:09:46 | efried | namedtuple. Fabulous. | |
| 16:12:20 | prometheanfire | nova doesn't look like it supports the new os_vif https://review.openstack.org/532987 | |
| 16:12:35 | prometheanfire | will bug the list if you want | |
| 16:12:39 | edleafe | efried: not that fabulous | |
| 16:12:49 | efried | edleafe You don't like? | |
| 16:12:57 | edleafe | efried: see the comment I just posted | |
| 16:14:59 | efried | edleafe You set an attribute on the class, not an instance thereof. | |
| 16:15:31 | edleafe | ugh, was just thinking that through more. Coffee hasn't fully kicked in yet | |
| 16:15:40 | openstackgerrit | Merged openstack/nova master: Do not attempt volume swap when guest is stopped/suspended https://review.openstack.org/389798 | |
| 16:15:55 | openstackgerrit | Merged openstack/nova master: Add aggregates check in allocation candidates https://review.openstack.org/522407 | |
| 16:20:25 | stephenfin | prometheanfire: Oh, looking | |
| 16:20:37 | stephenfin | sean-k-mooney, ralonsoh: Seen https://review.openstack.org/#/c/532987/ ? | |
| 16:21:06 | sean-k-mooney | stephenfin: looking now | |
| 16:21:06 | prometheanfire | stephenfin: cool | |
| 16:21:23 | prometheanfire | hopefully it's just a overly specific test | |
| 16:21:38 | ralonsoh | stephenfin: I'm taking a look now | |
| 16:22:35 | sean-k-mooney | stephenfin: at first glance these see unrelated to os-vif | |
| 16:23:00 | stephenfin | sean-k-mooney: Quite possibly. I haven't looked into it in detail | |
| 16:23:12 | stephenfin | Just saw the multiple rechecks. Usually a sign of something more sinister | |
| 16:23:25 | sean-k-mooney | stephenfin: the network basic opts test were failing due to a an ip confict | |
| 16:24:06 | openstackgerrit | Jackie Truong proposed openstack/nova master: Add trusted_certs object https://review.openstack.org/489408 | |
| 16:25:05 | sean-k-mooney | stephenfin: yes however that fact that tempest.scenario.test_network_basic_ops.TestNetworkBasicOps passes on multinode (one of the failing test) means its likely not a os-vif issue | |
| 16:25:59 | sean-k-mooney | stephenfin: ill dig a little deeper | |
| 16:27:14 | sean-k-mooney | stephenfin: there are a whole bunch of Too many open files errors in n-cpu logs | |
| 16:27:54 | sean-k-mooney | e.g. http://logs.openstack.org/87/532987/1/check/legacy-tempest-dsvm-neutron-full/cb9c0bb/logs/screen-n-cpu.txt.gz?level=ERROR#_Jan_12_14_00_56_222328 | |
| 16:29:18 | openstackgerrit | Eric Fried proposed openstack/nova master: SchedulerReportClient.set_traits_for_provider https://review.openstack.org/532564 | |
| 16:29:18 | openstackgerrit | Eric Fried proposed openstack/nova master: Track associated sharing RPs in report client https://review.openstack.org/526539 | |
| 16:29:19 | openstackgerrit | Eric Fried proposed openstack/nova master: Raise on API errors getting aggregates/traits https://review.openstack.org/526540 | |
| 16:29:19 | openstackgerrit | Eric Fried proposed openstack/nova master: ProviderTree.populate_from_iterable https://review.openstack.org/520756 | |
| 16:29:20 | openstackgerrit | Eric Fried proposed openstack/nova master: Track tree-associated providers in report client https://review.openstack.org/526541 | |
| 16:29:20 | openstackgerrit | Eric Fried proposed openstack/nova master: WIP: Scheduler[Report]Client.get_provider_tree https://review.openstack.org/521098 | |
| 16:29:21 | openstackgerrit | Eric Fried proposed openstack/nova master: WIP: ComputeDriver.update_provider_tree() https://review.openstack.org/521187 | |
| 16:29:21 | openstackgerrit | Eric Fried proposed openstack/nova master: ProviderTree.snapshot => ProviderSnapshot https://review.openstack.org/533244 | |
| 16:29:22 | openstackgerrit | Eric Fried proposed openstack/nova master: WIP: Use update_provider_tree from resource tracker https://review.openstack.org/520246 | |
| 16:29:40 | efried | jaypipes cdent edleafe ^ -- and FYI, the bottom patch is also needed by mgoddard to get ironic traits moving. | |
| 16:29:58 | jaypipes | efried: sure, just finishing up a rebase | |
| 16:30:23 | efried | gibi also ^ -- thanks for the comments on the ProviderTree.snapshot patch. I've now slotted it into the series where it's gonna wind up. | |
| 16:30:36 | edleafe | efried: ack | |
| 16:31:07 | cdent | noted | |
| 16:42:08 | artom | Pci knowledgable folks, if I have a PciDevice (a <hostdev> in the libvirt xml), what would be the best way to find its MAC? | |
| 16:42:32 | artom | sahid, stephenfin ^^ ? | |
| 16:43:49 | sean-k-mooney | stephenfin: ill stack with the latest release of os-vif and see if i can replicate the ci failure on my end | |
| 16:44:08 | sahid | artom: not sure i understand, we do set the mac address when configuring the libvirt <interface>, no? | |
| 16:44:38 | artom | sahid, I mean when passing through an entire PF | |
| 16:44:47 | artom | It's a <hostdev> in the XML, not an <interface> | |
| 16:44:52 | artom | So there's no MAC | |
| 16:44:53 | sahid | oh i see | |
| 16:45:05 | stephenfin | sean-k-mooney: Sounds good. I'll take a look myself tomorrow | |
| 16:45:11 | stephenfin | See if we can figure out what's going on | |
| 16:45:16 | sahid | artom: no idea sorry | |
| 16:45:29 | artom | sahid, no worries :) | |
| 16:46:01 | artom | Who's the most knowledgeable PCI person we have now that vladikr and nikola aren't with us anymore? | |
| 16:47:12 | sean-k-mooney | stephenfin: it could be litally that we are not hitting the openfile limit but it could be something else too. if its ulimit then it something we need to fix in devstack | |
| 16:48:47 | sahid | artom: in such situation it's the guest os which have the device, the host can't really see it | |
| 16:48:58 | sahid | i mean i guess it can't | |
| 16:49:01 | sean-k-mooney | stephenfin: fyi this local.conf should allow you to reproduce http://paste.openstack.org/show/645256/ | |
| 16:49:09 | sahid | so probably your change is qemu-guest-agent | |
| 16:49:27 | stephenfin | sean-k-mooney: Funny how it would trigger now. Surely we were using privsep before? | |
| 16:49:29 | sahid | s/have the device/own the device | |
| 16:50:21 | sean-k-mooney | stephenfin: we were but maybe with the change to useing the python libs we are a using more filedicriptors or b we are leaking them? | |
| 16:50:58 | stephenfin | We could validate that by turning on warnings. Python 3 warns for open fds | |
| 16:54:51 | openstackgerrit | Lee Yarwood proposed openstack/nova stable/pike: Do not attempt volume swap when guest is stopped/suspended https://review.openstack.org/533733 | |
| 16:55:23 | sean-k-mooney | stephenfin: it could possibly be a provider difference? the test that were failing on rackspace were passing on inap when run as part of the grenade multinode job. anyway ill see if i can reproduce locally | |