| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2019-09-17 | |||
| 11:58:20 | openstackgerrit | Kashyap Chamarthy proposed openstack/nova master: libvirt: Methods to handle request for Secure Boot & non-Q35 machine types https://review.opendev.org/682628 | |
| 11:58:20 | openstackgerrit | Kashyap Chamarthy proposed openstack/nova master: libvirt: Add methods to detect firmware auto-selection and SB https://review.opendev.org/682627 | |
| 12:01:38 | gibi | mriedem, bauzas: comments fixed i https://review.opendev.org/#/c/666857 | |
| 12:05:41 | stephenfin | efried: Are we ever going to be able to merge a patch again? :D | |
| 12:10:46 | kashyap | (And remove needless compat code; we should have much less of it, though, as we've already cleaned up a good chunk of it a few months.) | |
| 12:17:25 | efried | stephenfin: appears this time it was mirror problems. But yeah, the gods really seem to hate that top pmem patch. | |
| 12:17:52 | efried | stephenfin: though mriedem and I fixed a regression from the forbidden aggs series which was causing a bunch of our failures. | |
| 12:18:11 | stephenfin | o rly? link? | |
| 12:18:12 | efried | now if we could just kill the innodb bug | |
| 12:18:28 | efried | stephenfin: https://review.opendev.org/#/c/682475/ | |
| 12:18:57 | efried | also a couple WIPs to poison multiple sched services from starting. | |
| 12:19:23 | sean-k-mooney | kashyap: didnt we already bump to a minium of 3.0 earlier in the cycle | |
| 12:19:28 | efried | https://review.opendev.org/#/c/682485/ and https://review.opendev.org/#/c/682486/ | |
| 12:20:00 | kashyap | sean-k-mooney: We picked for Train, but we didn't pick anything for U release and gave a heads-up to the operators. | |
| 12:20:23 | stephenfin | kashyap: It's not too late to do that though, right? | |
| 12:20:29 | stephenfin | given that Train hasn't gone out | |
| 12:21:02 | kashyap | sean-k-mooney: No, not late :-) I just wanted to write it out loud, so I feel compelled to do it :D | |
| 12:21:45 | kashyap | I mean, stephenfin ^ | |
| 12:22:43 | sean-k-mooney | kashyap: we dont need to actully pick before we do the relaease. we can do it in early U | |
| 12:23:41 | kashyap | stephenfin: sean-k-mooney: Right. But the thing is - we didn't bump the version we said we will bump for Train | |
| 12:24:14 | kashyap | +NEXT_MIN_QEMU_VERSION = (2, 11, 0) | |
| 12:24:14 | kashyap | +NEXT_MIN_LIBVIRT_VERSION = (4, 0, 0) | |
| 12:24:14 | kashyap | We gave heads-up to bump Train to these: | |
| 12:24:28 | kashyap | ... in Sep 2018 (commit: 489b5f762e2c828354362667cd647086dc4eed60) | |
| 12:25:01 | stephenfin | kashyap: Sounds like something you might want to bring up at the team meeting | |
| 12:25:04 | kashyap | stephenfin: Think we're too late to do _that_, I suppose -- as people don't like the bump at the end of the release | |
| 12:25:31 | kashyap | stephenfin: Yeah, I take full accountability of not doing it in time earlier in the release :-( Just dropped through the cracks | |
| 12:26:03 | shilpasd | Eric: mriedem: thanks for setUp cleanup https://review.opendev.org/#/c/682475/ | |
| 12:26:30 | kashyap | stephenfin: Ah, the meeting is today | |
| 12:26:52 | sean-k-mooney | kashyap: i pesonally dont like that how we do the bumps in general. | |
| 12:26:56 | kashyap | stephenfin: My guess: Matt(R) will (correctly) dislike it... | |
| 12:27:16 | sean-k-mooney | but as i said we can bump early in U | |
| 12:27:29 | kashyap | sean-k-mooney: Like how we announce up ahead what we'll bump in the next release, and actually bump in the said release? | |
| 12:27:37 | kashyap | Any alternatives welcome :-) | |
| 12:27:49 | kashyap | sean-k-mooney: Yes, early U is fine. We've done such a thing before | |
| 12:27:54 | efried | shilpasd: yw, that would have been a tough one to spot | |
| 12:32:44 | sean-k-mooney | anyway there have been several realses in the past were we did not bump | |
| 12:32:52 | kashyap | sean-k-mooney: While I have you here, can I borrow you brain for a bit more on an unrelated thing... | |
| 12:32:56 | sean-k-mooney | in fact i woudl say it more common not too looking at the table | |
| 12:33:16 | kashyap | sean-k-mooney: Yep, indeed... | |
| 12:33:26 | sean-k-mooney | sure | |
| 12:33:59 | kashyap | sean-k-mooney: It's the Secure Boot-related scaffolding code - I have some sketch here (and the next patch): https://review.opendev.org/682627 | |
| 12:34:02 | kashyap | (No tests yet, mind you) | |
| 12:34:14 | kashyap | sean-k-mooney: Just want a different pair of eyes; as I'm feeling weary | |
| 12:35:24 | kashyap | sean-k-mooney: The design refresher for detecting firmware auto-selection and SB-enabled OVMF is in the above commit message | |
| 12:35:30 | sean-k-mooney | no test oh easy -1 the :) | |
| 12:35:31 | kashyap | sean-k-mooney: Might want to read it, when you have a block of time | |
| 12:35:34 | kashyap | :D | |
| 12:35:44 | kashyap | sean-k-mooney: It's Workflow -1 already | |
| 12:36:42 | kashyap | Actually, I used that commit message as "documentation" (as some of that should - I will - be added in documentation) | |
| 12:36:49 | sean-k-mooney | so we need to eventually talk about the inter for use get_domain_caps | |
| 12:36:57 | sean-k-mooney | domain_caps = self.get_domain_capabilities() | |
| 12:36:59 | sean-k-mooney | for arch in domain_caps: | |
| 12:37:01 | sean-k-mooney | for machine_type in domain_caps[arch]: | |
| 12:37:07 | sean-k-mooney | i have see that way to much | |
| 12:37:09 | kashyap | sean-k-mooney: Also, see the "Firmware auto-detection in action" part in the message | |
| 12:37:18 | kashyap | Hmm | |
| 12:37:25 | sean-k-mooney | every use of it i has see so far does the nested looping | |
| 12:37:39 | kashyap | I might get actually get rid of one level of it at least | |
| 12:38:02 | sean-k-mooney | and i mentioned to aspires that i want too change the to take parmaters for arch and machine type optionally | |
| 12:38:42 | kashyap | sean-k-mooney: Yeah, that's something I want consider | |
| 12:38:51 | kashyap | I already began by taking 'arch' as the param | |
| 12:39:06 | sean-k-mooney | yes so i want to change how the fuction works | |
| 12:39:16 | sean-k-mooney | if you pass nothinbg it will work as it currently does | |
| 12:39:42 | sean-k-mooney | if you pass parmater it woul check if it in the cache and if not call libivrt with those paramater set and look it up | |
| 12:39:47 | sean-k-mooney | then cache it | |
| 12:40:12 | sean-k-mooney | the current way we use it does not actully work in all cases | |
| 12:41:43 | sean-k-mooney | kashyap: by the way im not sure the fireware changes per machinetype | |
| 12:41:45 | kashyap | Yeah, I was considering the caching approach; will keep fiddling | |
| 12:41:58 | kashyap | sean-k-mooney: Sorry, what do you mean? | |
| 12:42:22 | sean-k-mooney | well at least on x86 you use the same uefi firmware image for both pc and q35 | |
| 12:42:38 | kashyap | sean-k-mooney: You mean the firmware values returned ("uefi" / "bios") won't change per machine type? | |
| 12:42:44 | sean-k-mooney | yes | |
| 12:42:58 | sean-k-mooney | they might | |
| 12:43:18 | sean-k-mooney | but then dont on x86 anyway as far as i know | |
| 12:43:19 | kashyap | For UEFI boot, q35 is mandatory; see the next patch, please, where we handle it | |
| 12:43:32 | sean-k-mooney | no its not | |
| 12:43:42 | sean-k-mooney | at least libvirt does not require that | |
| 12:44:05 | kashyap | Hmm, /me tries :-) | |
| 12:50:30 | kashyap | Modifying existing SB-enabled VMs to 'pc' is not a good idea; as the entire PCIe topo needs to change, etc. Will test on a diff. box later today | |
| 12:51:00 | sean-k-mooney | secure boot wont work for pc | |
| 12:51:06 | sean-k-mooney | but uefi should | |
| 12:51:11 | sean-k-mooney | imy trying it locally too | |
| 12:51:27 | sean-k-mooney | but im pretty sure i have done this in the past | |
| 12:51:31 | kashyap | Right, please try it. (Yeah, that's for sure: SB won't of course work for 'pc') | |
| 12:51:56 | kashyap | But I'm also 99.9999981% sure, UEFI with OVMF won't work with 'pc' | |
| 12:53:28 | sean-k-mooney | it boots with a black scree and 100% cpu usage so it does not look like its that happy | |
| 12:53:55 | sean-k-mooney | that said i am pretty sure i did use uefi with pc machine type in the past | |
| 12:53:56 | kashyap | Yep. OVMF is 'q35' only | |
| 12:54:05 | sean-k-mooney | maybe it was a different firmware image | |
| 12:54:06 | kashyap | I bet euros-to-croissants | |
| 12:56:00 | kashyap | sean-k-mooney: Thanks for the eyes so far. | |
| 13:01:32 | openstackgerrit | Sundar Nadathur proposed openstack/nova-specs master: Nova Cyborg interaction specification. https://review.opendev.org/682637 | |
| 13:05:26 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: Test heal port allocations in nova-next https://review.opendev.org/669879 | |
| 13:17:47 | gibi | mriedem: fixed your nits in ^^ | |
| 13:18:06 | gibi | mriedem: and also in https://review.opendev.org/#/c/666857 | |
| 13:18:45 | gibi | ahh you already hit the heal port patch, thanks for that | |
| 13:20:52 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Only allow one scheduler service in tests https://review.opendev.org/682486 | |
| 13:21:46 | mriedem | oh no a full rebase on 666857 | |
| 13:21:50 | mriedem | means i can't diff to PS7 | |