| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2019-02-08 | |||
| 19:16:50 | openstackgerrit | Corey Bryant proposed openstack/nova master: xenapi/agent: Change openssl error handling https://review.openstack.org/635533 | |
| 19:22:25 | cfriesen | If I were to request a release of os-traits (in order to make use of the new vTPM traits) would I ask for release 0.10.1 or 0.11.0? | |
| 19:23:03 | melwitt | new traits = "new features" so a minor version bump | |
| 19:23:35 | aspiers | melwitt: speaking of traits, do you have any thoughts on the above? I can submit a review easily | |
| 19:25:13 | leakypipes | aspiers: no reason other than legacy I think... I'd welcome a cleanup patch. | |
| 19:25:28 | melwitt | ok, good. I was about to say, I don't know and I'd ask jay | |
| 19:25:36 | aspiers | :) | |
| 19:25:50 | aspiers | leakypipes, melwitt: coming in about 15 seconds ... | |
| 19:25:51 | cfriesen | aspiers: I see you just added some stuff to os-traits...were you planning on requesting a new release? | |
| 19:26:13 | openstackgerrit | Adam Spiers proposed openstack/nova master: Convert CPU_TRAITS_MAPPING to use os_traits https://review.openstack.org/635919 | |
| 19:26:16 | aspiers | cfriesen: don't need one urgently yet - still working on the code to consume the new trait :) | |
| 19:26:39 | cfriesen | aspiers: I'm in the same boat, just don't want to get caught waiting too long | |
| 19:26:52 | aspiers | cfriesen: fair enough | |
| 19:27:33 | aspiers | leakypipes, melwitt: added you as reviewers | |
| 19:28:01 | mriedem | dansmith: yeah good point, that should move | |
| 19:28:10 | dansmith | I shall convert | |
| 19:28:27 | dansmith | I shall wait until you drop your +2 | |
| 19:28:31 | melwitt | aspiers: ack | |
| 19:28:44 | mriedem | dansmith: already done | |
| 19:28:55 | dansmith | mriedem: I shall be already done as well | |
| 19:28:56 | mriedem | dansmith: are you just going to move it to https://review.openstack.org/#/c/591657/ ? | |
| 19:29:28 | openstackgerrit | Elod Illes proposed openstack/nova stable/pike: Handle IndexError in _populate_neutron_binding_profile https://review.openstack.org/635921 | |
| 19:30:41 | dansmith | mriedem: am I? | |
| 19:31:06 | aspiers | leakypipes: just checking IIUC: is update_provider_tree() the right place to add a call to new code which tests for SEV support and then calls provider_tree.add_traits() if support is found? | |
| 19:31:51 | aspiers | leakypipes: I guess _get_cpu_traits() might be the other candidate, but that seems dependent on the guest CPU mode, which I don't think we want for SEV | |
| 19:32:32 | mriedem | dansmith: i'm not sure what 'convert' meant | |
| 19:32:36 | mriedem | oh -1? | |
| 19:32:42 | dansmith | mriedem: heh yeah, convert my vote | |
| 19:32:48 | mriedem | ah i see | |
| 19:32:55 | mriedem | saved my sweet sweet stats sheet | |
| 19:33:01 | mriedem | a true gentleman | |
| 19:33:21 | mriedem | aspiers: yes upt | |
| 19:33:28 | mriedem | to set inventory/aggregates/traits/etc | |
| 19:33:31 | aspiers | mriedem: thanks! | |
| 19:33:38 | mriedem | aspiers: fried_rice is your captain for upt | |
| 19:33:44 | aspiers | got it :) | |
| 19:33:55 | aspiers | hmm, that nick is making me hungry | |
| 19:53:07 | aspiers | leakypipes, fried_rice, mriedem: I'm wondering if it makes sense to detect SEV support in get_available_resource(), possibly under data["cpu_info"] | |
| 19:53:39 | mriedem | for what reason? | |
| 19:54:10 | aspiers | for one, so it's visible via the API | |
| 19:54:37 | mriedem | it is visible the API, | |
| 19:54:38 | mriedem | the placement API | |
| 19:54:42 | mriedem | as a trait on the compute node resource provider | |
| 19:55:09 | aspiers | right, but not via "openstack hypervisor show" IIUC? | |
| 19:55:11 | mriedem | we're not trying to game all new traits into the old hypervisor stats API | |
| 19:55:15 | mriedem | correct | |
| 19:55:16 | mriedem | and that's ok | |
| 19:55:28 | mriedem | in the long long future the hypervisor stats/show stuff could be proxied to placement | |
| 19:55:30 | aspiers | ah OK, so the idea is to move away from the old API? | |
| 19:55:32 | aspiers | got it | |
| 19:56:34 | aspiers | that thought arose because the get_domain_capabilities() I've added sits next to get_capabilities() https://review.openstack.org/#/c/633855/1/nova/virt/libvirt/host.py | |
| 19:56:50 | aspiers | which is currently called from _get_cpu_info() | |
| 19:57:34 | mriedem | but you can also get that via update_provider_tree yes? | |
| 19:57:40 | mriedem | self.host.get_domain_capabilities() | |
| 19:57:43 | aspiers | yup | |
| 19:57:46 | mriedem | if 'sev' in caps: add_my_trait() | |
| 19:57:53 | aspiers | makes sense, thanks! | |
| 19:57:59 | mriedem | the cpu feature flag stuff sort of works like that | |
| 19:58:54 | fried_rice | aspiers: Give me a bit, I'm on a call, but - yes, at a glance, it makes sense to me that SEV support would be discovered and the trait injected in update_provider_tree. | |
| 19:59:14 | aspiers | fried_rice: cool, thanks! | |
| 19:59:28 | aspiers | I need to grab some dinner | |
| 19:59:31 | aspiers | back in a bit | |
| 19:59:35 | fried_rice | aspiers: Or, during init_host, add it to the virt driver's capabilities dict. | |
| 20:00:18 | fried_rice | but that relies on a) an in-flight patch from mriedem that converts those to traits on the compute RP, and b) making the capabilities dict an instance var on the relevant virt drivers, like PowerVM has. | |
| 20:02:05 | aspiers | interesting, I'll check that out | |
| 20:02:50 | fried_rice | aspiers: a) is https://review.openstack.org/#/c/538498/ | |
| 20:03:10 | mriedem | "in-flight" | |
| 20:03:13 | mriedem | more like crash landed | |
| 20:03:23 | aspiers | LOL :) | |
| 20:03:58 | cfriesen | melwitt: fried_rice: if we're using the "cycle-with-intermediary" library pattern for os-traits, shouldn't there have been a release already after stein-2? | |
| 20:04:46 | fried_rice | aspiers: b) is visible in the out-of-tree PowerVM driver where we set a capability depending on which disk driver we're using: https://github.com/openstack/nova-powervm/blob/master/nova_powervm/virt/powervm/driver.py#L175-L181 | |
| 20:05:04 | melwitt | cfriesen: yes. I had thought the release team was going to propose them but it looked like they didn't | |
| 20:05:07 | fried_rice | cfriesen: I wouldn't expect os-traits releases to be a big deal wrt milestones, but whatevs. | |
| 20:05:49 | cfriesen | fried_rice: not that it's a big deal, but according to http://lists.openstack.org/pipermail/openstack-dev/2018-October/135689.html it was going to be automatic | |
| 20:05:56 | fried_rice | gotcha | |
| 20:05:57 | melwitt | they did a release of os-traits at s-1 but not for s-2 | |
| 20:06:18 | aspiers | fried_rice: in case you're interested, this is the first step towards detection of SEV support which will be consumed by the code which provides the new trait | |
| 20:06:21 | aspiers | https://review.openstack.org/#/c/633855/ | |
| 20:06:25 | fried_rice | were there changes between m-1 and m-2? | |
| 20:06:35 | melwitt | yeah, I wondered if maybe there weren't | |
| 20:06:50 | fried_rice | aspiers: ack, I've had that spec on my list for yonks. hope to get around to reading it eventually ... | |
| 20:06:55 | aspiers | fried_rice: thanks a lot for all the info, I'll take a look | |
| 20:07:04 | aspiers | hehe | |
| 20:07:39 | aspiers | OK, I really REALLY need dinner now. bbl ;-) | |
| 20:07:40 | cfriesen | melwitt: what was the exact cutoff? the vTPM stuff went in on Jan 8, and there was a minor change on Jan 5 | |
| 20:08:05 | cfriesen | anyways, I'll propose a release | |
| 20:08:53 | melwitt | not sure. smcginnis, do you know how the auto release works? we noticed there wasn't an automatic release proposed for os-traits at s-2 | |
| 20:16:55 | melwitt | yeah, I don't find any auto-proposed releases for s-2 like this one from s-1 https://review.openstack.org/612843 | |
| 20:58:40 | smcginnis | melwitt, cfriesen: Were there more than 1 or 2 non-zuul, functional changes? It's possible it was accidentally skipped if it didn't look like there were really changes to get out. | |
| 20:59:29 | cfriesen | smcginnis: one actual change went in on jan 8, so maybe that was it. | |
| 21:00:30 | smcginnis | Actually, looks like that didn't merge until the 11th, so it was most likely after we checked. | |
| 21:00:33 | smcginnis | https://review.openstack.org/#/c/629299/ | |
| 21:01:27 | melwitt | smcginnis: there was one functional change, I think. I tried a gerrit search for project:openstack/releases message:"stein milestone-2" and didn't find any other auto-generated releases | |
| 21:01:51 | melwitt | (unless the commit message format changed) | |
| 21:02:16 | smcginnis | I can't remember if we used "stein milestone-2" in the commit messages for those. | |
| 21:05:19 | melwitt | ok, I was curious if any other project got an auto-proposed release for s-2 but didn't know how to find | |
| 21:06:29 | melwitt | it was nice having it for s-1 anyway :) | |
| 21:10:25 | melwitt | mriedem: can you sanity check me, is this backward incompatible, the 'only_contrib' removal? it looks to me like it's removed but not backward incompat https://github.com/openstack/python-novaclient/compare/11.1.0...master | |
| 21:13:58 | mriedem | no it's not, so a minor version bump at least | |
| 21:14:06 | mriedem | that warning has been there since pike i think | |