| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-05-16 | |||
| 12:21:40 | mriedem | mdbooth: ok, dropped it from the runway slot; time to have a meeting about how you're in too many meetings. | |
| 12:22:18 | mriedem | alex_xu: a bit worried that https://review.openstack.org/#/c/560317/ isn't getting attention from the author since the last -1 was from may 7, over a week and a half ago | |
| 12:23:58 | mriedem | naichuans: is https://blueprints.launchpad.net/nova/+spec/vgpu-rocky ready to go into a runway slot for the xenapi changes? | |
| 12:24:06 | mriedem | starting at https://review.openstack.org/#/c/520313/ | |
| 12:25:54 | alex_xu | mriedem: yea, I will reach him soon | |
| 12:26:02 | alex_xu | very soon by wechat :) | |
| 12:26:24 | alex_xu | mriedem: I think his patch has a dependence | |
| 12:26:27 | mriedem | alex_xu: ok i guess let me know if i should put that into a runway slot today, otherwise i'll put the xenapi vgpu n-r-p series in there | |
| 12:27:00 | mriedem | alex_xu: efried commented that the code should be using update_provider_tree https://review.openstack.org/#/c/560317/2/nova/virt/driver.py@1684 | |
| 12:27:06 | mriedem | but i thought that was ready | |
| 12:28:35 | efried | mriedem: Sorry, refreshing my memory. | |
| 12:28:54 | alex_xu | mriedem:, yes, one of patch for the update_provider_tree, but trying to figure out which one is the dependence | |
| 12:28:57 | efried | mriedem: IIRC, the spec doesn't say anything about a new driver method, so it's still reasonably okay without a delta. Let me confirm that. | |
| 12:29:35 | mriedem | https://specs.openstack.org/openstack/nova-specs/specs/rocky/approved/report-cpu-features-as-traits.html#work-items | |
| 12:29:41 | mriedem | "Implement an interface which returns a list of traits for the CPU features in each virt driver." | |
| 12:29:45 | mriedem | that should be update_provider_tree | |
| 12:30:01 | mriedem | which is what the 2nd work item says :) | |
| 12:30:49 | efried | mriedem: Yeah, and L114. | |
| 12:30:58 | efried | mriedem: So yeah, my -1 on the patch stands firm | |
| 12:32:21 | alex_xu | efried: it should be calle dby the update_provider_tree, is there anyone work on the update_provider_tree for libvirt driver? | |
| 12:33:09 | efried | alex_xu: I started this series, which bhagyashris is working on finishing up: https://review.openstack.org/#/c/560444/ | |
| 12:33:36 | efried | alex_xu: bhagyashris is having trouble with the functional testing. I've got it on my backlog to help out there, but haven't gotten to it yet. | |
| 12:34:06 | mriedem | ok so i'll skip the cpu features bp this time for a runway slot since it's not ready | |
| 12:34:21 | alex_xu | efried: got it, thanks | |
| 12:34:32 | alex_xu | mriedem: yea, thanks, let us skip this time | |
| 12:35:30 | mriedem | naichuans: i've put https://blueprints.launchpad.net/nova/+spec/vgpu-rocky into a runway slot for the xenapi driver changes | |
| 12:35:34 | mriedem | https://etherpad.openstack.org/p/nova-runways-rocky | |
| 12:35:46 | mriedem | dansmith: we have a few new runway entries so should probably update the channel topic | |
| 12:35:46 | efried | alex_xu: If you wanted to help with that, it would be welcomed. Here's the test case bhagyashris has so far: http://paste.openstack.org/raw/720752/ (test_servers.py) | |
| 12:37:46 | bhagyashris | efried: Hi, i will push the patch soon | |
| 12:38:04 | efried | bhagyashris: Okay. Sorry I haven't gotten to that yet. Have you made progress? | |
| 12:46:28 | alex_xu | bhagyashris: let me know if you still have question | |
| 12:50:06 | bhagyashris | efried: no issue yes | |
| 12:50:18 | bhagyashris | alex_xu: thank you :) | |
| 12:50:18 | efried | great | |
| 12:54:36 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Implement granular policy rules for placement https://review.openstack.org/524425 | |
| 12:54:37 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add granular policy rules for /resource_classes* https://review.openstack.org/565578 | |
| 12:54:38 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add granular policy rules for resource providers inventories https://review.openstack.org/568666 | |
| 12:54:39 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add granular policy rules for usages https://review.openstack.org/568706 | |
| 13:01:29 | openstackgerrit | Eric Fried proposed openstack/nova master: Use placement.inventory.inuse in report client https://review.openstack.org/568639 | |
| 13:05:29 | openstackgerrit | Eric Fried proposed openstack/nova master: Use GET.get instead of GET.getall in alloc-cands https://review.openstack.org/568713 | |
| 13:06:06 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: Add requested_resources field to RequestSpec https://review.openstack.org/567267 | |
| 13:06:07 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: Transfer port.resource_request to the scheduler https://review.openstack.org/567268 | |
| 13:06:08 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: Make placement.lib.RequestGroup a versioned object https://review.openstack.org/568840 | |
| 13:18:49 | openstackgerrit | Vladyslav Drok proposed openstack/nova-specs master: Allow having placement inventories with reserved value equal to total https://review.openstack.org/568613 | |
| 13:23:16 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Deduplicate config/policy reference docs from main index https://review.openstack.org/564846 | |
| 13:23:17 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Implement granular policy rules for placement https://review.openstack.org/524425 | |
| 13:23:18 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add granular policy rules for /resource_classes* https://review.openstack.org/565578 | |
| 13:23:19 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add granular policy rules for resource providers inventories https://review.openstack.org/568666 | |
| 13:23:20 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add granular policy rules for usages https://review.openstack.org/568706 | |
| 13:24:21 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Cleanup ugly stub in TestLocalDeleteAllocations https://review.openstack.org/561997 | |
| 13:27:30 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: Make placement.lib.RequestGroup a versioned object https://review.openstack.org/568840 | |
| 13:27:31 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: Add requested_resources field to RequestSpec https://review.openstack.org/567267 | |
| 13:27:32 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: Transfer port.resource_request to the scheduler https://review.openstack.org/567268 | |
| 13:44:42 | dansmith | jaypipes: on this: https://review.openstack.org/#/c/568613/5 | |
| 13:45:00 | jaypipes | dansmith: yes? | |
| 13:45:01 | dansmith | jaypipes: isn't not allowing reserved=total just a bug? I feel like maybe we don't need the whole microversion dance | |
| 13:45:28 | dansmith | like someone did < instead of <= in some code somewhere? | |
| 13:45:30 | jaypipes | dansmith: I feel the same way. | |
| 13:45:50 | dansmith | jaypipes: does anyone not feel that way? | |
| 13:45:53 | mriedem | lyarwood: you +2ed the ocata backport but not the pike one :) https://review.openstack.org/#/c/568369/ | |
| 13:46:15 | dansmith | my feeling is that there is no client that would depend on total=reserved returning a 400 | |
| 13:46:17 | mriedem | dansmith: i brought it up when it was being discussed, | |
| 13:46:23 | mriedem | because no one knew why the capacity check was that way | |
| 13:47:21 | mriedem | and its a behavior change in the api, so you kind of need to know when you can make that kind of request and expect it to be ok | |
| 13:47:22 | lyarwood | mriedem: appears I closed the tab before gerrit actually updated the review, noice! | |
| 13:48:01 | dansmith | mriedem: but microversions aren't for providing bug-for-bug compatibility right? that's why we don't do one to fix a 500? | |
| 13:48:22 | mriedem | this isn't a 500 | |
| 13:48:31 | dansmith | I know | |
| 13:48:51 | dansmith | but really, if we do it and get a 400, we can fall back to the same behavior as today | |
| 13:48:56 | dansmith | I dunno, this just seems overkill to me | |
| 13:49:13 | mriedem | i don't think i ever said i was going to block this when we talked about it in -placement weeks ago, | |
| 13:49:21 | mriedem | i -2ed the change this week because there is a spec up, | |
| 13:49:29 | mriedem | if people don't care about the api behavior change, then whatever | |
| 13:50:22 | dansmith | I would think that if this is a bug, with no microversion, that we'd backport the fix as .. a fix | |
| 13:50:54 | mriedem | meh - this doesn't even fix any real problem right now does it? | |
| 13:51:09 | dansmith | just MHO, I'm not strongly against a microversion I guess, I just .. I dunno. | |
| 13:51:09 | mriedem | the actual bug was already fixed some other way is my understanding | |
| 13:51:40 | dansmith | it fixes some logic that prevents nova from setting total=reserved when that should be totally legit | |
| 13:52:01 | dansmith | I dunno, I don't want to argue about it, but I definitely think total=reserved should be legit | |
| 13:52:04 | dansmith | and should have always been legit | |
| 13:52:04 | mriedem | so instead of doing ^ it deletes the node inventory right? | |
| 13:52:05 | mriedem | during cleaning | |
| 13:52:15 | dansmith | it doesn't now, | |
| 13:52:20 | dansmith | it just punts | |
| 13:52:21 | dansmith | well, wait, | |
| 13:52:30 | dansmith | it doesn't if there are no nodes, that's the failure/startup case I guess | |
| 13:52:37 | dansmith | so maybe it's still deleting inventory currently I guess | |
| 13:52:38 | mriedem | my understanding is we worked around the problem by deleting the inventory during cleaning | |
| 13:53:04 | dansmith | yeah, probably, I'm confusing the startup zero-day or ironic-is-down case, but anyway | |
| 13:53:31 | mriedem | i tend to think if this isn't needed to actually fix anything, and it's a behavior change, then it's a microversion | |
| 13:53:43 | dansmith | from the placement point of view, | |
| 13:53:54 | dansmith | I feel like it's a fix to a piece of logic that was wrong | |
| 13:54:11 | dansmith | sure, nova could continue to do the thing it does today to work around it, but | |
| 13:54:32 | dansmith | anyway, whatever, I'd rather just get it done regardless | |
| 13:55:59 | mriedem | do what y'all want, i will step aside on this one | |
| 13:57:33 | dansmith | no, it's not that big of a deal, we've got a spec, let's just roll | |
| 13:57:37 | dansmith | getting it working is the important thing | |
| 14:00:34 | openstackgerrit | Vladyslav Drok proposed openstack/nova-specs master: Allow having placement inventories with reserved value equal to total https://review.openstack.org/568613 | |
| 14:01:30 | BobBall | mriedem / naichuans: Naichuan is on PRC time but yes, the XenAPI bits of the vgpu-rocky series are definitely ready for a runway slot. | |