| *** tetsuro has joined #openstack-placement | 00:11 | |
| openstackgerrit | Jie Li proposed openstack/nova-specs master: Support volume-backed server rebuild https://review.openstack.org/532407 | 03:46 |
|---|---|---|
| *** diga has joined #openstack-placement | 04:00 | |
| *** tetsuro has quit IRC | 04:04 | |
| *** tetsuro has joined #openstack-placement | 05:11 | |
| openstackgerrit | Jie Li proposed openstack/nova-specs master: Support volume-backed server rebuild https://review.openstack.org/532407 | 06:04 |
| *** takashin has left #openstack-placement | 07:01 | |
| *** tssurya has joined #openstack-placement | 07:46 | |
| *** s10 has joined #openstack-placement | 09:17 | |
| *** e0ne has joined #openstack-placement | 09:24 | |
| *** cdent has joined #openstack-placement | 09:31 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: Correct lower-constraints.txt and the related tox job https://review.openstack.org/614559 | 10:46 |
| *** diga has quit IRC | 11:16 | |
| mnaser | just a heads up | 11:21 |
| mnaser | we cant really continue adding support for os_placement if the placement-manage command is not there | 11:21 |
| mnaser | how do we init dbs? | 11:21 |
| cdent | mnaser: it's nearly merged: https://review.openstack.org/#/c/600161/ | 11:30 |
| cdent | it's last in a stack of quite a lot of stuff to do database migrations and reviews have been slow to come | 11:30 |
| cdent | but there was a surge of effort yesterday | 11:30 |
| cdent | I'm working on matt's suggestions on ^ today so it shoud happen before the end of the week | 11:31 |
| *** s10 has quit IRC | 11:57 | |
| *** s10 has joined #openstack-placement | 12:04 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: Add a placement-manage CLI https://review.openstack.org/600161 | 12:20 |
| cdent | efried, mriedem, edleafe : I add some docs to ^ | 12:20 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Adapt placement fixtures for external use https://review.openstack.org/617935 | 12:46 |
| belmoreira | cdent efried considering the number of requests from the compute nodes to placement is there any other relevant patch that I should have a look? https://review.openstack.org/#/c/615606 https://review.openstack.org/#/c/615646 https://review.openstack.org/#/c/615677 | 13:03 |
| *** s10 has quit IRC | 13:05 | |
| *** s10 has joined #openstack-placement | 13:21 | |
| *** diga has joined #openstack-placement | 13:22 | |
| *** mriedem has joined #openstack-placement | 13:41 | |
| cdent | belmoreira: sorry, was away, the whole stack beginning with https://review.openstack.org/#/c/615606 is the main stuff | 14:26 |
| mriedem | cdent: teeny tiny things in https://review.openstack.org/#/c/600161/4 | 14:30 |
| mriedem | er https://review.openstack.org/#/c/600161/ | 14:30 |
| cdent | ✔ | 14:30 |
| mriedem | that will make life easier when placement-status is added | 14:30 |
| * cdent shakes fist | 14:31 | |
| mriedem | which if someone else hasn't started, i can do this morning | 14:31 |
| cdent | If we must :) | 14:31 |
| mriedem | there will be a tc spanking if the goal is not accomplished | 14:32 |
| *** dtantsur has joined #openstack-placement | 14:32 | |
| * cdent drops trou | 14:32 | |
| dtantsur | hi folks! | 14:32 |
| dtantsur | I'm trying to understand the Placement API design and how it maps into Ironic stuffs | 14:33 |
| dtantsur | given that any Ironic node is represented by exactly one instance of a custom resource class | 14:33 |
| dtantsur | to reserve a node I need: 1. GET /resource_providers?resources=CUSTOM_BAREMETAL:1&required=list,of,traits | 14:33 |
| dtantsur | 2. POST /allocations with suitable UUID? | 14:34 |
| cdent | thanks for coming dtantsur | 14:34 |
| cdent | first | 14:34 |
| cdent | before you keep going on too far | 14:34 |
| cdent | I need to caution you that "reserving" is not a thing that placement officially does | 14:34 |
| cdent | if you make an allocation (usually PUT /allocations/{uuid}) you are saying "I am consuming this" | 14:34 |
| cdent | re high level overview, some of the info at https://docs.openstack.org/placement will help, but it is currently in flux as we move things around | 14:35 |
| dtantsur | cdent: I have hard time understanding reserving vs consuming here | 14:35 |
| cdent | consuming means: this thing is in use and the resources in its inventory are actually being used | 14:36 |
| dansmith | well, | 14:36 |
| efried | belmoreira: https://review.openstack.org/#/c/615695/ and https://review.openstack.org/#/c/615677/ are going to be the big ones for reducing placement traffic. But they're built on top of | 14:36 |
| efried | https://review.openstack.org/#/c/615646/ | 14:36 |
| efried | https://review.openstack.org/#/c/615641/ (merged in master) | 14:36 |
| efried | https://review.openstack.org/#/c/615606/ (merged in master) | 14:36 |
| cdent | reserving means: I expect to want this thing soon, and while I am not using them right now, I don't want any one else to | 14:36 |
| dansmith | it means "we assume you're using it" | 14:36 |
| dansmith | actually consuming those resources would be actively using the memory, if we're being pedantic | 14:37 |
| cdent | dansmith: right, that's why I said officially, because I was coming round to: if you want it to mean "hold", it would work | 14:37 |
| dtantsur | okay, I see the difference in the VM case | 14:37 |
| *** s10 has quit IRC | 14:37 | |
| dansmith | dtantsur: it's not really any different.. in the vm case, we "consume" well before we actually consume, just like the ironic case | 14:38 |
| dtantsur | how do you avoid the instances landing on a compute that only has space for one of them? | 14:38 |
| dtantsur | ah! | 14:38 |
| dansmith | dtantsur: I think the word "reserve" has a time component for a lot of people, which is why it wasn't used in placement terminology | 14:38 |
| dtantsur | sure, I don't want it to associate with Blazer :) | 14:38 |
| cdent | yeah, it reminds too much of making a reservation for dinner later | 14:38 |
| mriedem | the ironic driver does reserve all of the inventory on a node while it's being cleaned | 14:38 |
| mriedem | if that's what you're thinking of | 14:39 |
| dtantsur | (omg, I forgot about it) no, I rather mean "consume" here. | 14:39 |
| dtantsur | actually it would be awesome if somebody could read https://review.openstack.org/#/c/617953/ and tell me how deeply screwed I'll be when I come to a Placement backend for that. | 14:40 |
| cdent | dtantsur: I'll read that later in the day, I need to get this placement-manage stuff doneish | 14:40 |
| mriedem | dtantsur: so that is a proxy to placement? | 14:41 |
| mriedem | oh backends... | 14:41 |
| dtantsur | mriedem: this is one of the cases. the other case is hardcore standalone ironic with nothing else. | 14:41 |
| mriedem | well, | 14:42 |
| mriedem | it sounds like you're looking for the placement allocations api, which if modeled that way then the backend for placement just works, and the backend for a db just needs to implement the placement api | 14:42 |
| mriedem | you could also use the PlacementDirect tool for the db backend | 14:43 |
| mriedem | "The conductor creates an *Allocation* object in the database with | 14:43 |
| mriedem | ``state=allocating`` and ``conductor_affinity=<host name>``." | 14:43 |
| dtantsur | PlacementDirect? any reference to that? | 14:43 |
| mriedem | is where things probably start to break down | 14:43 |
| dtantsur | why? | 14:43 |
| mriedem | https://github.com/openstack/placement/blob/master/placement/direct.py | 14:44 |
| mriedem | allocations in placement don't have state | 14:44 |
| mriedem | or affinity | 14:44 |
| cdent | dtantsur: I suggest you avoid PlacementDirect unless there's some deep reason you actually need it | 14:44 |
| mriedem | there are aggregates in placement... | 14:44 |
| dtantsur | mriedem: this affinity thing is for ironic HA bookkeeping, you can ignore it here, I think | 14:44 |
| dtantsur | essentially to say "oh, this allocation was in progress by that conductor.. and now it's dead wut du?" | 14:45 |
| mriedem | why can't standalone ironic just depend on placement? | 14:47 |
| mriedem | the cinder guys said that standalone cinder requiring placement would also be a show stopper for some | 14:47 |
| dansmith | it would need to behave differently if consumed by nova then right? | 14:47 |
| mriedem | i guess standalone is truly standalone | 14:47 |
| dansmith | unless it's just double accounting everything | 14:47 |
| mriedem | that's noted in the spec, under the nova driver impact section | 14:47 |
| dtantsur | I hear you :( yes, there are people who want ironic and nothing else. some even asked me if we can avoid amqp.. | 14:47 |
| dansmith | especially for the zero day seed case, | 14:48 |
| dansmith | completely standalone ironic makes a lot of sense I think | 14:48 |
| dtantsur | dansmith: I guess we'll need a configuration option for who reports to Placement: nova-compute or ironic | 14:48 |
| dtantsur | ah, you're not talking about reporting | 14:48 |
| dansmith | dtantsur: I guess I missed what the use case is here | 14:48 |
| dtantsur | dansmith: an example that comes up very often is cluster API in k8s | 14:49 |
| dtantsur | some folks want to have a bm provisioning tool supported, and it may be ironic | 14:49 |
| dtantsur | but they don't want a lot of openstack | 14:49 |
| * dansmith stares wide-eyed at that spec | 14:50 | |
| dansmith | I guess I need to read | 14:50 |
| mriedem | so the k8s cluster API ~= nova? | 14:50 |
| dtantsur | mriedem: I'm not expert at all, but I think it's kind of "give me a cluster" API. the Nova backend is in progress, but some folks are also interested in a pure BM thing. | 14:52 |
| mriedem | nova backend for k8s? | 14:52 |
| dtantsur | mriedem: it's k8s installer, not k8s itself. i.e. a way to install k8s on openstack (= nova) | 14:53 |
| mriedem | oh | 14:53 |
| * dtantsur needs to polish his knowledge of k8s before it becomes no longer hypeish | 14:53 | |
| mriedem | sure, if there is one marketing message i got from the summit it's that k8s on ironic is getting people all excited | 14:54 |
| cdent | rather | 14:55 |
| mriedem | that and if i put 'edge' in any of my talks they will be accepted | 14:55 |
| dtantsur | ++ | 14:55 |
| cdent | k8s on ironic @ the edge | 14:55 |
| dtantsur | I'm doing it now for 2 conferences, let's see how it goes | 14:55 |
| mriedem | heh | 14:55 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Add a placement-manage CLI https://review.openstack.org/600161 | 14:56 |
| cdent | mriedem: I think that ^ gets it | 14:56 |
| mriedem | come on steve get it | 14:57 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Adapt placement fixtures for external use https://review.openstack.org/617935 | 14:58 |
| cdent | ah, go me, le sigh | 14:59 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Add a placement-manage CLI https://review.openstack.org/600161 | 15:00 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Adapt placement fixtures for external use https://review.openstack.org/617935 | 15:00 |
| mriedem | +2 | 15:01 |
| cdent | dansmith, efried : your eyes on https://review.openstack.org/600161 would be great | 15:01 |
| *** efried is now known as efried_pto | 15:02 | |
| efried_pto | cdent: ^ (back this afternoon) | 15:02 |
| cdent | aye aye | 15:02 |
| mriedem | added stephenfin to the bottom on | 15:02 |
| mriedem | *one | 15:02 |
| mriedem | i forgot to do that last night | 15:02 |
| belmoreira | efried cdent: thanks. Will add them when they move to master. | 15:07 |
| mriedem | my dad: "i sent a video to your mother and she can't open it. if i send it to you and you send it to her, will she be able to view it then?" | 15:11 |
| mriedem | yes dad, that's how file type conversion works | 15:11 |
| cdent | tell him to print it | 15:12 |
| mriedem | cdent: did you see the comment on https://review.openstack.org/#/c/600162/ about changing the depends-on to the changeid so the 3rd party CI using zuulv2 will work | 15:12 |
| mriedem | ? | 15:12 |
| cdent | yes, but I hadn't got around to doing it because there's something like 3 or 4 different patches that would need to be updated | 15:13 |
| cdent | and that was enough effort to cause me to "later" and then "forget" | 15:14 |
| mriedem | i think it's just the devstack patch that needs updating? | 15:14 |
| cdent | it depends on which bit of third party ci we're caring about. I was also thinking in terms of the nova changes to use the placement fixture: https://review.openstack.org/#/c/617941/ . Oh, I guess I fixed that one already | 15:15 |
| cdent | I'll change the devstack one now | 15:15 |
| * cdent wants everyone (especially vmware) to move to zuulv3 | 15:17 | |
| dtantsur | zuulv3 ftw. absolute awesomeness. | 15:49 |
| *** tssurya has quit IRC | 15:55 | |
| mriedem | why isn't requirements.txt in the deps for tox targets in placement? | 16:06 |
| mriedem | hmm i guess it's not in nova's tox.ini either... | 16:08 |
| cdent | mriedem: usedevelop=True makes sure that python setup.py develop is called and pbr makes sure that requirements.txt is installed | 16:10 |
| cdent | (when setup.py develop is called) | 16:10 |
| cdent | so if requirements.txt was listed explicitly we'd try to install twice | 16:11 |
| cdent | this bit me while fixing the lower-constraints stuff | 16:11 |
| cdent | (where "this" is pbr doing things) | 16:11 |
| mriedem | sorry was arguing about the merits of traditional stuffing with my wife | 16:13 |
| mriedem | hmm, guess i need to figure out why oslo.upgradecheck isn't getting installed them | 16:14 |
| mriedem | *then | 16:14 |
| cdent | traditional stuffing versus? | 16:16 |
| cdent | I don't think that tox is aware of changes to requirements.txt all by itself, did you -r? | 16:17 |
| mriedem | versus some spinach/artichoke/cream cheese abomination | 16:17 |
| mriedem | it'll be like the turkey has a urinary tract infection | 16:17 |
| mriedem | i did -r yeah | 16:17 |
| mriedem | listen, if i can't do away with mashed potatoes in place of roasted yukon goldies, i'll be damned if i'm not getting traditional stove top style stuffing | 16:18 |
| *** e0ne has quit IRC | 16:21 | |
| cdent | i agree | 16:22 |
| cdent | it's not thanksgiving without sagey bready stodgy yumness | 16:22 |
| mriedem | it's what the pilgrims would have wanted | 16:23 |
| mriedem | we're hosting this year which is rare, and i don't particularly care for a dried out bird, so i suggested we just make steaks and stuff people actually like to eat | 16:26 |
| mriedem | but was shot down | 16:26 |
| openstackgerrit | Matt Riedemann proposed openstack/placement master: Add placement-status upgrade check command https://review.openstack.org/619035 | 16:30 |
| openstackgerrit | Matt Riedemann proposed openstack/placement master: Consider root id is None in the database case https://review.openstack.org/619041 | 17:01 |
| mriedem | ^ is the proper cherry pick of that bug fix from nova | 17:03 |
| edleafe | mriedem: yeah, mine was a placeholder until the nova patch merged so we didn't forget it | 17:04 |
| *** dtantsur is now known as dtantsur|afk | 17:19 | |
| *** e0ne has joined #openstack-placement | 17:31 | |
| *** e0ne has quit IRC | 17:44 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: Allow placement to start without a config file https://review.openstack.org/619049 | 17:48 |
| openstackgerrit | Chris Dent proposed openstack/placement master: WIP: Optionally migrate database at service startup https://review.openstack.org/619050 | 17:48 |
| cdent | mriedem: pep8 issues on https://review.openstack.org/619035 otherwise very nice | 17:54 |
| *** e0ne has joined #openstack-placement | 17:57 | |
| openstackgerrit | Artom Lifshitz proposed openstack/nova-specs master: Re-propose numa-aware-live-migration spec https://review.openstack.org/599587 | 18:22 |
| *** efried_pto is now known as efried | 18:28 | |
| mriedem | sob | 18:42 |
| openstackgerrit | Matt Riedemann proposed openstack/placement master: Add placement-status upgrade check command https://review.openstack.org/619035 | 18:53 |
| cdent | +2 to that. get efried on that stuff and we be stylin | 18:56 |
| efried | I'm doing what, reviewing ---^ ? | 18:56 |
| cdent | and its parent and cousin | 18:57 |
| cdent | Many, if not most, of the things on https://etherpad.openstack.org/p/placement-extract-stein-4 are at least started. Might be time to refactor to a -5 soon | 19:20 |
| *** e0ne has quit IRC | 19:22 | |
| efried | cdent: https://review.openstack.org/#/c/600161/ +A, with a question inline for a fup if relevant. | 19:38 |
| efried | cdent: what else specifically? | 19:38 |
| * cdent looks | 19:38 | |
| cdent | efried: both of the two above that | 19:39 |
| efried | ... | 19:39 |
| cdent | of the two the 'fixtures' one is probably more immediately useful, but then after that you'll probably want to go see: https://review.openstack.org/#/c/617941/ | 19:41 |
| cdent | getting the cli merged will allow a big cascade of other stuff to merge and test, so is a major step forward. go us. | 19:45 |
| openstackgerrit | Merged openstack/placement master: Add a placement-manage CLI https://review.openstack.org/600161 | 19:51 |
| * cdent drinks | 19:53 | |
| openstackgerrit | Jack Ding proposed openstack/nova-specs master: Flavor Extra Spec and Image Properties Validation https://review.openstack.org/618542 | 19:54 |
| efried | cdent: https://review.openstack.org/#/c/618708/ - see note, please -2. | 20:00 |
| cdent | roger | 20:01 |
| efried | cdent: Questions in https://review.openstack.org/#/c/617935/ if you please. | 20:12 |
| * cdent pleases | 20:13 | |
| efried | cdent, mriedem: question in https://review.openstack.org/#/c/619035/ | 20:19 |
| efried | cdent: one more question in https://review.openstack.org/#/c/617935/ but +W. | 20:22 |
| * cdent defers to mriedem on 619035, I'm not clear | 20:22 | |
| mriedem | replied | 20:23 |
| mriedem | (py35) osboxes@osboxes:~/git/placement$ pip show openstack-placement | 20:24 |
| mriedem | Name: openstack-placement | 20:24 |
| mriedem | Version: 0.0.1.dev10832 | 20:24 |
| mriedem | when we release, that will become 1.0.0 | 20:24 |
| openstackgerrit | Merged openstack/placement master: Adapt placement fixtures for external use https://review.openstack.org/617935 | 20:31 |
| cdent | yay | 20:32 |
| efried | cdent: Do you agree with mriedem that we should start placement at 1.0.0? I would say there's an argument to start it at <version it was the last time it existed in nova>+1. Thoughts? | 20:35 |
| cdent | yes 1 | 20:35 |
| cdent | it's the pattern other extracted things have followed | 20:35 |
| efried | So basically the placement service will go from 18.0.0 to 1.0.0, sort of. | 20:35 |
| efried | okay, fair enough. | 20:36 |
| cdent | I think it's more sensible to people who will be coming to placement fresh | 20:36 |
| efried | I guess. Seeing 19.0.0 they be like, where are the previous versions? | 20:36 |
| cdent | right | 20:37 |
| openstackgerrit | Merged openstack/placement master: Add placement-status upgrade check command https://review.openstack.org/619035 | 20:52 |
| *** efried is now known as efried_back_mond | 22:43 | |
| *** efried_back_mond is now known as efried_back_mon | 22:43 | |
| cdent | mriedem: you care to speculate why this is now happening on the devstack change: http://logs.openstack.org/07/619107/3/check/devstack/6e59b7f/job-output.txt.gz#_2018-11-20_22_48_36_173407 | 23:01 |
| cdent | I can't figure out which thing has changed | 23:01 |
| mriedem | hmm | 23:05 |
| mriedem | no changes to devstack-gate https://github.com/openstack-infra/devstack-gate/commits/master | 23:05 |
| mriedem | not seeing anything recent or related in openstack-zuul-jobs https://github.com/openstack-infra/openstack-zuul-jobs/commits/master | 23:06 |
| cdent | yeah, I did a similar search around and was bewildered | 23:06 |
| mriedem | i'd ask in infra | 23:07 |
| cdent | heading that way shortly | 23:08 |
| mriedem | hmm https://github.com/openstack-dev/devstack/blob/master/.zuul.yaml#L262 | 23:11 |
| mriedem | maybe something changed in zuul and how it handles required-projects overrides | 23:12 |
| mriedem | the devstack-gate change doesn't take effect on this job i don't think since it doesn't use devstack-gate | 23:12 |
| mriedem | so my guess would be we need to add placement to that required projects list in the devstack job... | 23:12 |
| mriedem | but then why did i need to add it to openstack-zuul-jobs, | 23:13 |
| mriedem | oh that was the legacy base job template though | 23:13 |
| mriedem | http://git.openstack.org/cgit/openstack-infra/openstack-zuul-jobs/tree/zuul.d/jobs.yaml#n957 | 23:13 |
| mriedem | so how the devstack change was working before, idk | 23:13 |
| mriedem | but my guess is that's the problem | 23:13 |
| cdent | yeah, that's what's bewildering | 23:13 |
| cdent | i'll try it | 23:14 |
| *** takashin has joined #openstack-placement | 23:34 | |
| cdent | mriedem: " i think it's because previous patchsets had a depends-on on a change in placement, so it was being pulled in.", said corvus | 23:36 |
| mriedem | ah | 23:40 |
| mriedem | well that makes sense at least | 23:40 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Clean up use of Database fixture https://review.openstack.org/619121 | 23:41 |
| mriedem | teeny mistake in the devstack change commit message | 23:42 |
| mriedem | it's nearly midnight for you, want me to update this? | 23:42 |
| cdent | well it wouldn't be me if I didn't make some of those | 23:42 |
| cdent | yes, please, that would be great | 23:42 |
| cdent | I've just started yet another fup on something else | 23:43 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Allow placement to start without a config file https://review.openstack.org/619049 | 23:56 |
| openstackgerrit | Chris Dent proposed openstack/placement master: WIP: Optionally migrate database at service startup https://review.openstack.org/619050 | 23:56 |
| * cdent waves goodnight | 23:58 | |
| *** cdent has quit IRC | 23:59 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!