| openstackgerrit | Matt Riedemann proposed openstack/nova-specs master: Spec for cross-cell resize https://review.openstack.org/616037 | 01:24 |
|---|---|---|
| *** mriedem has quit IRC | 01:25 | |
| *** ttsiouts has quit IRC | 01:40 | |
| *** efried has joined #openstack-placement | 01:47 | |
| *** efried has quit IRC | 01:50 | |
| *** ttsiouts has joined #openstack-placement | 02:18 | |
| *** ttsiouts has quit IRC | 02:23 | |
| openstackgerrit | Yongli He proposed openstack/nova-specs master: add spec "show-server-numa-topology" https://review.openstack.org/612256 | 03:08 |
| *** ttsiouts has joined #openstack-placement | 03:17 | |
| *** ttsiouts has quit IRC | 03:22 | |
| openstackgerrit | Yongli He proposed openstack/nova-specs master: add spec "show-server-numa-topology" https://review.openstack.org/612256 | 03:22 |
| *** tetsuro has joined #openstack-placement | 04:13 | |
| *** tetsuro has quit IRC | 04:16 | |
| *** tetsuro has joined #openstack-placement | 04:33 | |
| *** tetsuro has quit IRC | 05:17 | |
| *** tetsuro has joined #openstack-placement | 05:17 | |
| *** ttsiouts has joined #openstack-placement | 05:18 | |
| *** ttsiouts has quit IRC | 05:23 | |
| *** ttsiouts has joined #openstack-placement | 07:19 | |
| *** ttsiouts has quit IRC | 07:24 | |
| openstackgerrit | Rui Zang proposed openstack/nova-specs master: support virtual persistent memory https://review.openstack.org/601596 | 08:28 |
| *** tetsuro has quit IRC | 08:36 | |
| *** tetsuro has joined #openstack-placement | 08:45 | |
| *** dtantsur|afk is now known as dtantsur | 08:59 | |
| *** ttsiouts has joined #openstack-placement | 09:20 | |
| *** ttsiouts has quit IRC | 09:24 | |
| *** ttsiouts has joined #openstack-placement | 09:30 | |
| *** e0ne has joined #openstack-placement | 09:42 | |
| *** ttsiouts has quit IRC | 10:33 | |
| *** tetsuro has quit IRC | 10:34 | |
| *** cdent has joined #openstack-placement | 11:41 | |
| *** ttsiouts has joined #openstack-placement | 12:49 | |
| *** mriedem has joined #openstack-placement | 14:14 | |
| *** mriedem has quit IRC | 14:19 | |
| *** mriedem has joined #openstack-placement | 14:21 | |
| *** ttsiouts has quit IRC | 15:18 | |
| *** efried has joined #openstack-placement | 15:25 | |
| *** e0ne has quit IRC | 15:28 | |
| openstackgerrit | Merged openstack/nova-specs master: Add spec for libvirt driver launching AMD SEV-encrypted instances https://review.openstack.org/609779 | 16:07 |
| openstackgerrit | Chris Dent proposed openstack/placement master: WIP: Placement install documentation https://review.openstack.org/628220 | 16:46 |
| openstackgerrit | Matt Riedemann proposed openstack/nova-specs master: add spec "show-server-numa-topology" https://review.openstack.org/612256 | 17:09 |
| cdent | mriedem, jaypipes, efried : can one or some of you have a look at https://review.openstack.org/#/c/623556/ and make it go if possible? In talks with smcginnis it seems wise to do an initial release of placement to pypi so that something working is there, rather than the current broken thing, and having the os-resource-classes stuff in would be good from a dependencies standpoint | 17:43 |
| efried | cdent: looking... | 17:45 |
| efried | cdent: So os-resource-classes is fully a thing now? | 17:45 |
| cdent | yes, released to pypi and everything | 17:45 |
| efried | is there any packaging nonsense that needs to happen for this to work irl? | 17:45 |
| efried | I guess that happens at the end of the release. | 17:45 |
| * cdent nods | 17:46 | |
| efried | cdent: Where did the resource class sync code come from, save me looking around | 17:47 |
| cdent | it was originally in https://review.openstack.org/#/c/584084/ | 17:47 |
| efried | cdent: so it isn't in production anywhere yet? | 17:48 |
| cdent | which sync code do you mean? the stuff that makes the database have the right entries? No, it's not in production yet, because the code you are reviewing has to be merged in order for it to be in production... | 17:49 |
| efried | point is, we weren't doing that (ensure_resource_classes_sync) when resource classes were hardcoded in the (nova/placement) project? | 17:49 |
| cdent | correct | 17:50 |
| efried | It's just, I'm seeing some stuffs, and need to know what level of alarm to raise. | 17:50 |
| cdent | the cache code would first check standards and then go to db | 17:50 |
| cdent | now the cache code always goes to the db | 17:50 |
| *** ttsiouts has joined #openstack-placement | 18:04 | |
| openstackgerrit | Adam Spiers proposed openstack/nova-specs master: AMD SEV: address final feedback received prior to merge https://review.openstack.org/628252 | 18:21 |
| efried | cdent: Sorry for the delay, lots of things going on at once. Left comments on that review. No vote yet. | 18:21 |
| cdent | thanks | 18:22 |
| openstackgerrit | Adam Spiers proposed openstack/nova-specs master: AMD SEV: address final feedback received prior to merge https://review.openstack.org/628252 | 18:23 |
| *** dtantsur is now known as dtantsur|afk | 18:43 | |
| *** mriedem has quit IRC | 19:22 | |
| *** mriedem has joined #openstack-placement | 19:24 | |
| *** sean-k-mooney1 has joined #openstack-placement | 19:25 | |
| openstackgerrit | sean mooney proposed openstack/nova-specs master: Add spec for sriov live migration https://review.openstack.org/605116 | 19:35 |
| efried | cdent, jaypipes: I'm going to be MIA M-W next week for onboarding. Can one of you please take the sched mtg? | 19:43 |
| cdent | aye aye | 19:43 |
| efried | thanks | 19:43 |
| efried | cdent: Is create_incomplete_consumers something that should be done from the placement repo at this point? | 19:56 |
| cdent | I think we decided that by now (stein) it should be done | 19:59 |
| cdent | efried: If you're reviewing https://review.openstack.org/618215 you might give that a -2 to prevent it being merge before everyone agrees its time | 20:02 |
| efried | I am, and will do. | 20:03 |
| efried | cdent: what will constitute "it's time"? | 20:03 |
| cdent | efried: I thought it was time two years ago, so you probably want to refer to a different authority on that one. But if pressed, the agreements on governance are a good proxy | 20:04 |
| cdent | however, it would be good for the deletion to happen sooner than later because the canonical packagers are not able to package placement until the placement code is deleted from nova (for reasons unclear) | 20:05 |
| cdent | mriedem may have thoughts, so let's see what he says | 20:05 |
| efried | cdent: Could the placement-in-nova code even work anymore at this point? Wouldn't we have to unwind a bunch of other stuff in various other projects? | 20:05 |
| cdent | yes, it would still work, but CI has already changed over (for devstack, grenade, tempest) | 20:06 |
| efried | cdent: "governance are a good proxy" <= not sure I follow. We don't need to tie this action to placement being under independent governance surely | 20:08 |
| mriedem | re create_incomplete_consumers - if we drop that from nova and dont have it in placement, | 20:08 |
| mriedem | then we should at least have some check for it before people upgrade to stein, | 20:09 |
| cdent | efried, no, we don't, but it is currently the only list of things we have: thus "proxy". | 20:09 |
| efried | I would think "it's time" means "extracted placement repo works from everywhere we care about." | 20:09 |
| mriedem | which could be a blocker migration or at least a placement-status upgrade check | 20:09 |
| efried | mriedem: would those things be in the nova repo or the placement repo or both? | 20:09 |
| mriedem | create_incomplete_consumers code relies on hitting the consumers objects directly, so that hits the nova_api/placement db | 20:09 |
| mriedem | so a blocker migration for that wouldn't be in nova | 20:10 |
| mriedem | i think it really only makes sense that placement would be checking for those | 20:10 |
| efried | hits nova_api db? how so? | 20:10 |
| mriedem | it's not an api call, | 20:10 |
| *** ttsiouts has quit IRC | 20:10 | |
| mriedem | there is no consumers endpoint | 20:10 |
| efried | afaict it's just making sure consumer IDs in the allocations table have records in the consumer table. | 20:10 |
| mriedem | right | 20:11 |
| mriedem | with placement split out, that's a different db | 20:11 |
| mriedem | which nova shouldn't have direct access to | 20:11 |
| *** ttsiouts has joined #openstack-placement | 20:11 | |
| efried | yes, the call itself should definitely not be in the nova project anywhere. No doubt in my mind about that. | 20:11 |
| efried | question is, is it safe to have no equivalent in the placement repo? | 20:11 |
| mriedem | equivalent to the create_incomplete_consumers CLI for the data migration? | 20:12 |
| efried | right | 20:12 |
| efried | um, not sure if this is the right question, but: is it possible for an independent placement db, operated by extracted placement package, to have incomplete consumers? | 20:13 |
| mriedem | safe? maybe. if you're upgrading from rocky to stein and have deployed extracted placement, and run an upgrade check which fails, i'm assuming you haven't upgraded nova to stein yet | 20:13 |
| mriedem | yes | 20:13 |
| efried | then we probably need. | 20:13 |
| mriedem | if you ran the db data copy script to copy the nova_api db tables to the placement db, | 20:13 |
| mriedem | you could have incomplete consumers in there if you didn't finish the online data migrations for rocky in nova yet | 20:14 |
| mriedem | like i said, the alternative to having create_incomplete_consumers in placement is a blocker migration or at least an upgrade check | 20:14 |
| mriedem | though the latter is not widely used at this point | 20:14 |
| efried | blocker migration <= in placement | 20:15 |
| mriedem | by blocker migration, i mean, placement-manage db sync falis | 20:15 |
| mriedem | *fails | 20:15 |
| efried | And the remedy for that would be... | 20:15 |
| efried | running create_incomplete_consumers somehow | 20:15 |
| mriedem | right | 20:15 |
| efried | i.e. we need a placement manage create_incomplete_consumers. | 20:15 |
| mriedem | but if you've already copied over the database... | 20:15 |
| efried | that would still work, no? | 20:15 |
| *** ttsiouts has quit IRC | 20:16 | |
| cdent | if we were willing to do data migrations, not just schema migrations, then we could just add fixing consumers as a migration and be done: it would correct or not correct without need for thought | 20:16 |
| cdent | an additional command line is overkill | 20:16 |
| cdent | we should both check and fix in one go | 20:16 |
| *** ttsiouts has joined #openstack-placement | 20:17 | |
| efried | okay, fine, but *something* needs to be done. | 20:17 |
| efried | We can't just forget about it. | 20:17 |
| cdent | we're not, we're having this conversation | 20:17 |
| cdent | just because _I_ haven't done it doesn't mean it isn't going to happen | 20:17 |
| cdent | I hope people aren't expecting me to remember everything | 20:17 |
| mriedem | doing the data migration during db sync is a problem if that data migration could take a long time, which i'm not sure how bad that data migration would be if you're churning through several thousand RPs/allocations/consumers table joins | 20:18 |
| mriedem | in the CERN case is what i'm thinking of | 20:18 |
| efried | cdent: yahyah, point of this conversation is, if we decide something is needed (which it sounds like we have), then we need to make some kind of an artifact (a WIP patch, a line in an etherpad, whatever) that is less forgettable than this chat. | 20:18 |
| mriedem | i guess it's the rps table isn't involved here | 20:19 |
| cdent | we've got this: could add something there: https://etherpad.openstack.org/p/placement-extract-stein-4 | 20:19 |
| efried | but of note, this issue needn't block merging the delete-placement-from-nova patch, because the ship (the extracted placement repo itself) has already sailed. | 20:20 |
| mriedem | cern could have 100k+ allocations though... | 20:20 |
| cdent | efried: well yeah we've got two threads running right now, which are orthogonal: | 20:21 |
| *** e0ne has joined #openstack-placement | 20:21 | |
| cdent | a) what are the controls on when we can delete the code | 20:21 |
| cdent | b) how should we deal with allocations without consumers | 20:22 |
| mriedem | i don't know the definitive go/no-go decision on deleting placement for nova and relying on extracted placement for people upgrading to stein, i thought it was at least one deployment project having tested their upgrade support to make sure it's all working - grenade isn't really good enough | 20:22 |
| mriedem | i said a couple of weeks ago that i think we need to have a pow wow in january when people are back to figure that out because this can't go much longer than jan imo | 20:22 |
| mriedem | i'm out next week btw | 20:22 |
| cdent | s/without consumers/withou consumer entries in consumers table/ | 20:22 |
| efried | cdent: Added b) to etherpad L35-6. We can stop talking about that now. Thanks for the convo. | 20:22 |
| efried | cdent: Is "database migration" the right heading, or should it move somewhere else? | 20:23 |
| cdent | thanks efried | 20:23 |
| cdent | no, that's good | 20:23 |
| efried | ight | 20:23 |
| efried | wrt a), probably need to figure out why packagers think they can't move forward | 20:23 |
| mriedem | yeah, lyarwood should know the story for red hat, | 20:24 |
| cdent | efried: as far as I know people aren't blocked, it's more a matter of being enqueued | 20:24 |
| mriedem | i don't know who to ask for suse | 20:24 |
| mriedem | coreycb should know for canonical | 20:24 |
| mriedem | as a result, there might be a catch 22 here where OSA can't test upgrades until canonical packages extracted placement? | 20:25 |
| mriedem | unless they install placement from source? | 20:25 |
| mriedem | not sure how OSA works | 20:25 |
| cdent | they have a "from-tarballs" ode | 20:25 |
| cdent | mdoe | 20:25 |
| cdent | damn | 20:25 |
| cdent | mode | 20:25 |
| efried | cdent: So do you want that -2 on there? (Keeping in mind that I won't be here M-W to remove it.) | 20:25 |
| mriedem | i really don't think that is going to merge next week | 20:26 |
| * cdent looks at mriedem | 20:26 | |
| mriedem | like i said, need to have a meeting | 20:26 |
| mriedem | need melwitt involved | 20:26 |
| mriedem | and i'm out next week | 20:26 |
| mriedem | dansmith: when do you head north? | 20:26 |
| dansmith | mriedem: east | 20:26 |
| cdent | is dansmith santa claus? | 20:26 |
| dansmith | mriedem: and, I'm out the entire week of the 21st | 20:27 |
| mriedem | isn't alaska north of portland? | 20:27 |
| mriedem | or was that a "head east" the band joke? | 20:27 |
| dansmith | alaska is north and west | 20:27 |
| dansmith | the race is not in alaska though :) | 20:28 |
| mriedem | oh i thought it was the iditarod | 20:28 |
| dansmith | no, it's a qualifier for the iditarod | 20:28 |
| efried | cdent: I put the -2. Otherwise I'm +2 on that bottom patch. | 20:28 |
| cdent | dansmith: are you participating in a dog race? | 20:28 |
| mriedem | efried: cdent: melwitt: dansmith: jaypipes: ok then i think plan for a go/no-go extracted placement meeting the week of the 14th | 20:28 |
| dansmith | cdent: I do every year | 20:28 |
| cdent | that's awesome | 20:29 |
| dansmith | not as a musher, in case it's not clear | 20:29 |
| mriedem | mushee | 20:29 |
| cdent | still awesome | 20:29 |
| dansmith | I think the mushees are the dogs, no? | 20:29 |
| efried | that's what mriedem meant. | 20:29 |
| mriedem | heh | 20:29 |
| jaypipes | heh | 20:29 |
| mriedem | like, in a kinky way | 20:29 |
| jaypipes | mriedem: works for me. | 20:29 |
| efried | I missed you guys | 20:30 |
| mriedem | efried: btw, onboarding | 20:30 |
| mriedem | go | 20:30 |
| efried | okay. | 20:30 |
| dansmith | mriedem: I told you not to talk about my leather strap fetish | 20:30 |
| mriedem | efried: i mean, explain | 20:30 |
| cdent | I was in alaska at a b&b once upon a time that trained dogs for iditarod and was asked to over-winter to help run the dogs (the dogs liked me) but I didn't and I've regretted it ever since | 20:30 |
| efried | Tomorrow is my last day at IBM. As of Monday, I work for... a chip company that rhymes with Schmintel. | 20:30 |
| dansmith | efried: AMD? | 20:31 |
| mriedem | efried: so they found their cyborg integrator huh | 20:31 |
| efried | just so | 20:31 |
| efried | Changing companies was my best shot at being able to keep my job. | 20:31 |
| efried | by which I mean, keep doing what i'm doing. | 20:31 |
| mriedem | aye | 20:31 |
| efried | my actual employment was not at risk | 20:31 |
| cdent | can we call the meeting something other than go/no-go. unwinding at this point would be sad and super painful. how about go-faster | 20:31 |
| mriedem | same when i was at ibm | 20:31 |
| efried | mriedem: nod, I recall. | 20:32 |
| mriedem | cdent: "aggressive checkpoint" | 20:32 |
| cdent | wfm | 20:32 |
| *** takashin has joined #openstack-placement | 20:37 | |
| *** e0ne has quit IRC | 20:37 | |
| cdent | efried: this stack will probably help on the consumer migration: https://review.openstack.org/#/c/624942/ | 20:37 |
| efried | ack | 20:38 |
| cdent | i've left a pointer to the above chatter | 20:39 |
| melwitt | mriedem: ack on meeting the week of jan 14 to discuss | 20:48 |
| *** e0ne has joined #openstack-placement | 20:54 | |
| sean-k-mooney1 | efried: best of luck. let me know when the internal plotics start lol | 20:56 |
| sean-k-mooney1 | efried: are you joining ssg/otc? | 20:57 |
| efried | sean-k-mooney1: Thanks. I'll have to pick your brain at some point about all of that. Yes, OTC. On Sundar's team. | 20:57 |
| efried | sean-k-mooney1: But I'm no stranger to big-company internal politics and other gorp. | 20:58 |
| sean-k-mooney1 | efried: haha anytime. well at least you will be in the groupt that pocitcally is ment to drive opensouce contributions so you will be able to bypass several of the intnal issues | 20:58 |
| *** e0ne has quit IRC | 21:50 | |
| *** e0ne has joined #openstack-placement | 21:52 | |
| *** cdent has quit IRC | 21:56 | |
| *** ttsiouts has quit IRC | 21:56 | |
| *** e0ne has quit IRC | 21:57 | |
| *** efried has quit IRC | 22:35 | |
| *** efried has joined #openstack-placement | 22:35 | |
| *** efried has quit IRC | 22:39 | |
| openstackgerrit | Merged openstack/placement master: Use os-resource-classes in placement https://review.openstack.org/623556 | 23:09 |
| openstackgerrit | Matt Riedemann proposed openstack/nova-specs master: Spec for cross-cell resize https://review.openstack.org/616037 | 23:26 |
| openstackgerrit | Merged openstack/nova-specs master: Remove doc about metadata from 'share-pci-between-numa-nodes' https://review.openstack.org/619807 | 23:48 |
| *** mriedem has quit IRC | 23:58 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!