| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-12-22 | |||
| 16:47:31 | openstackgerrit | Merged openstack/nova master: Make conductor pass and use host_lists https://review.openstack.org/511358 | |
| 16:47:36 | mriedem | woot ^ | |
| 16:48:04 | mriedem | #success nova merged alternate hosts support for server build | |
| 16:48:06 | openstackstatus | mriedem: Added success to Success page | |
| 16:48:11 | artom | mdbooth, I'd see the point if we let them both run to completion in different processes | |
| 16:48:27 | artom | mdbooth, but in the end, all your test does it call _create_uuid twice in successions | |
| 16:48:29 | mdbooth | artom: They do both run to completion! | |
| 16:48:35 | mdbooth | In effectively different processes. | |
| 16:48:52 | artom | ... | |
| 16:49:07 | artom | mdbooth, but when the second one is called via the mock, the first's flow is interreupted | |
| 16:49:11 | artom | Until the second one returns | |
| 16:49:12 | artom | No? | |
| 16:49:13 | mdbooth | By mocking _create_uuid specifically, it triggers an execution order which causes a race | |
| 16:49:30 | figleaf | mriedem: It's a xmas miracle! | |
| 16:49:38 | openstackgerrit | Merged openstack/nova master: doc: update supported drivers for cpu topology https://review.openstack.org/529294 | |
| 16:49:54 | openstackgerrit | Matt Riedemann proposed openstack/python-novaclient master: Add support for the 2.57 microversion https://review.openstack.org/528128 | |
| 16:50:14 | mdbooth | artom: Yes. That's an execution order which would cause incorrect behaviour if we didn't handle it. | |
| 16:50:34 | openstackgerrit | Matthew Booth proposed openstack/nova master: Make BlockDeviceMapping object support uuid https://review.openstack.org/242603 | |
| 16:50:35 | openstackgerrit | Matthew Booth proposed openstack/nova master: DriverBlockDevice: make subclasses inherit _proxy_as_attr https://review.openstack.org/524167 | |
| 16:50:35 | openstackgerrit | Matthew Booth proposed openstack/nova master: Add an online migration for BDM.uuid https://review.openstack.org/525599 | |
| 16:50:36 | openstackgerrit | Matthew Booth proposed openstack/nova master: Expose BDM uuid to drivers https://review.openstack.org/529037 | |
| 16:50:42 | mdbooth | artom: However, it's gone ^^^ ;) | |
| 16:50:51 | artom | mdbooth, dammit | |
| 16:51:03 | artom | mdbooth, I still want to understand you | |
| 16:51:17 | mdbooth | artom: It's a good test, but a totally agree that it's more complicated that it needs to be | |
| 16:52:00 | mdbooth | Anyway, I'm going to run for the hills! | |
| 16:52:13 | mdbooth | Merry Christmas, all! See you in the New Year. | |
| 16:55:06 | openstackgerrit | Sylvain Bauza proposed openstack/nova master: libvirt: create vGPU for instance https://review.openstack.org/528832 | |
| 16:57:49 | artom | mdbooth, enjoy your ugly sweaters ;) | |
| 17:12:48 | mriedem | figleaf: heh my bash fu always fails the first few times | |
| 17:14:25 | figleaf | mriedem: don't feel bad - I felt uneasy +1'ing that because my bash-fu is so weak | |
| 17:14:37 | figleaf | it looked logical, thougn | |
| 17:14:39 | figleaf | though | |
| 17:15:29 | mriedem | it does do the correct thing for nova_cell1.conf | |
| 17:15:30 | mriedem | so that's good | |
| 17:19:22 | mriedem | someone was mentioning issues with the NumInstances filter yesterday, and i got thinking about that one, if you ran multiple filter scheduler processes, that filter would be totally racey | |
| 17:19:38 | mriedem | depending on which process gets a request, or if both are processing requests at the same time for the same host | |
| 17:20:26 | mriedem | if the limit is 10 and you've got 9 on a host already, and both workers get a request at the same time, they'd both think there is room for one more and could choose it | |
| 17:20:32 | mriedem | placement doesn't fix that | |
| 17:20:37 | mriedem | only reschedules would fix that | |
| 17:21:30 | mriedem | seems like a totally useless filter anyway assuming you have any non-trivial number of flavors | |
| 17:21:48 | kashyap | mriedem: Hey, when you have a moment, you know why this is failing? (I know it's coming from my change) -- http://logs.openstack.org/11/529611/3/check/openstack-tox-py27/f611b46/job-output.txt.gz#_2017-12-21_17_29_32_011876 | |
| 17:21:54 | kashyap | ( https://review.openstack.org/#/c/529611/ ) | |
| 17:22:09 | kashyap | It's just doing something super simple; adding a config class & test that XML element | |
| 17:22:22 | kashyap | Ah, I see it | |
| 17:22:28 | kashyap | My test is trying to do it as: | |
| 17:22:29 | kashyap | 2017-12-21 17:29:32.012231 | ubuntu-xenial | </smm> | |
| 17:22:29 | kashyap | 2017-12-21 17:29:32.012215 | ubuntu-xenial | <smm state="on"/> | |
| 17:22:29 | kashyap | 2017-12-21 17:29:32.012197 | ubuntu-xenial | <smm> | |
| 17:22:34 | kashyap | However, it should be: | |
| 17:23:26 | kashyap | <features> | |
| 17:23:26 | kashyap | ... | |
| 17:23:27 | kashyap | <smm state='on'/> | |
| 17:23:27 | kashyap | <apic/> | |
| 17:23:27 | kashyap | <acpi/> | |
| 17:23:29 | kashyap | </features> | |
| 17:23:31 | kashyap | ... | |
| 17:23:40 | kashyap | mriedem: Disregard me. I'll rework the tests | |
| 17:24:03 | mriedem | looks like it doesn't like <smm state="on"/> | |
| 17:24:13 | mriedem | http://logs.openstack.org/11/529611/3/check/openstack-tox-py27/f611b46/job-output.txt.gz#_2017-12-21_17_29_32_012165 | |
| 17:24:17 | kashyap | mriedem: But if you see, the test is trying to do: | |
| 17:24:21 | kashyap | <smm> | |
| 17:24:32 | kashyap | <smm state="on"> | |
| 17:24:34 | kashyap | </smm> | |
| 17:24:42 | kashyap | But there's no top-level 'smm' attribute | |
| 17:24:55 | kashyap | It is _one_ of the sub elements of the 'features' attribute | |
| 17:28:09 | mriedem | commented in the patch | |
| 17:29:31 | kashyap | That's fast. /me clicks | |
| 17:30:55 | kashyap | mriedem: Thanks. You could actually -1 it, don't be so polite. | |
| 17:30:57 | kashyap | :P | |
| 17:31:29 | mriedem | there you go | |
| 17:34:24 | kashyap | Heh, you (-)one-upped it :-) | |
| 17:39:37 | openstackgerrit | Matt Riedemann proposed openstack/nova master: WIP: Sanity check for NumInstancesFilter https://review.openstack.org/529866 | |
| 17:40:11 | bauwser | folks, bailing out for vacations | |
| 17:40:18 | bauwser | see you in 2018 \o | |
| 17:40:54 | bauzas | haha | |
| 17:40:57 | bauzas | i'm back! | |
| 17:41:18 | kashyap | mriedem: Hmm, I actually modelled it just like other features, the class -- LibvirtConfigGuestFeatureAPIC() or LibvirtConfigGuestFeaturePAE() | |
| 17:41:38 | bauzas | donne moi du fromage savoureux | |
| 17:41:41 | kashyap | More appropriate one is - LibvirtConfigGuestFeatureKvmHidden(). Because it also as sub-element | |
| 17:44:25 | kashyap | No, I'm wrong, the KvmHidden() thing has its own element. | |
| 17:47:26 | mriedem | kashyap: more inline | |
| 17:47:32 | mriedem | for your subelement woes | |
| 17:47:52 | mriedem | your root element is the smm element, | |
| 17:48:00 | mriedem | you are adding a sub-element to that, which is wrong, | |
| 17:48:05 | mriedem | you need to set an attribute on the smm element | |
| 17:48:08 | mriedem | which is your root | |
| 17:48:26 | mriedem | xml is f'ing awesome | |
| 17:48:34 | kashyap | Yeah, I should model it something like LibvirtConfigGuestSMBIOS() | |
| 17:48:41 | kashyap | I know, XML... | |
| 17:49:02 | kashyap | From what I recall about why libvirt chose XML was... | |
| 17:50:02 | kashyap | The initial author of libvirt Daniel Veillard was on W3C standards, and 'libxml' maintainer... | |
| 17:50:11 | kashyap | Not sure JSON was there at that time yet | |
| 17:50:41 | kashyap | So now they're (we) stuck with it... | |
| 17:51:27 | mriedem | i worked on a project for 6 years dealing with https://www.oasis-open.org/committees/sdd/ | |
| 17:51:30 | mriedem | so i'm familiar | |
| 17:52:06 | mriedem | you'd think i'd actually know something about tosca and heat and nfv stuff, but... | |
| 17:52:09 | kashyap | Wauw. Yeah, I knnow OSAIS standards; the 'virtio' documentation also is hosted there | |
| 17:52:38 | kashyap | But at the recent KVM Forum in Prague, I saw users of virtio, who want to implement devices complaining of the super dense documentation / standards. | |
| 17:52:55 | mriedem | maybe they would prefer CIM | |