| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-06-29 | |||
| 12:41:17 | gaoyuan | bauzas: this is the code, I deleted something to make it clearer | |
| 12:41:47 | mriedem | how was i able to do this then? https://review.openstack.org/#/c/473627/8/nova/scheduler/filter_scheduler.py@197 | |
| 12:42:16 | gibi | bauzas: ups sorry, here is the correct URL https://review.openstack.org/#/c/419730 | |
| 12:42:31 | gaoyuan | bauzas: what I want, is to set a break point, in the actual code, to see the value of, let's say, connection_info | |
| 12:43:41 | mriedem | edleafe: ok removed my -1 | |
| 12:43:45 | mriedem | edleafe: going to rebase https://review.openstack.org/#/c/473627/ ? | |
| 12:44:55 | bauzas | edleafe: I have a shit ton of review backlog, any urgent things I should look at for placement? | |
| 12:45:02 | edleafe | mriedem: maybe because of https://github.com/openstack/nova/blob/master/nova/db/sqlalchemy/api_models.py#L207 ? | |
| 12:45:03 | bauzas | like the above ? | |
| 12:45:31 | edleafe | bauzas: I would focus on Jay's series on allocation candidates | |
| 12:45:45 | bauzas | edleafe: that's the obvious one | |
| 12:45:45 | edleafe | bauzas: he's changing how scheduler drivers work | |
| 12:45:53 | sfinucan | mriedem: Assuming gibi hasn't already asked you, could you look at the discussion here https://review.openstack.org/#/c/448779/30/nova/notifications/objects/instance.py | |
| 12:46:04 | edleafe | bauzas: well, you missed that discussion on Tuesday | |
| 12:46:17 | mriedem | edleafe: the object doesn't know anything about that | |
| 12:46:19 | bauzas | edleafe: tl;dr then ? | |
| 12:46:25 | mriedem | edleafe: plus sqla columns default to nullable=True | |
| 12:46:44 | edleafe | bauzas: drivers now return the entire list of HostState objects to the scheduler manager | |
| 12:46:50 | sfinucan | wait - I just saw the "is in China" bit. At some point, in that case | |
| 12:47:00 | edleafe | bauzas: The claim is now done in the manager, not the driver | |
| 12:47:08 | mriedem | sfinucan: yeah gibi mentioned it above | |
| 12:47:33 | bauzas | edleafe: which claim ? | |
| 12:47:45 | bauzas | edleafe: you mean consume_from_req() ? | |
| 12:47:53 | bauzas | or did we merged something while I was in my cave? | |
| 12:47:57 | edleafe | bauzas: No, the posting of the allocation | |
| 12:48:12 | bauzas | I should look at code then | |
| 12:48:38 | edleafe | bauzas: that might help :) | |
| 12:49:34 | bauzas | that's where stupid paperwork rules about country-specific regulations ruins my productivity | |
| 12:50:42 | mriedem | edleafe: rebase that extra specs change and i'll +2 before be | |
| 12:50:44 | mriedem | *Bed | |
| 12:50:58 | edleafe | mriedem: on it | |
| 12:52:52 | openstackgerrit | Moshe Levi proposed openstack/os-vif master: unplug_vf_passthrough: don't try to delete representor netdev https://review.openstack.org/478820 | |
| 12:53:27 | bauzas | gaoyuan: well, you can introspect connection_info whenever you want | |
| 12:54:23 | bauzas | that snippet you gave me just implies the assertion of what's passed to the function | |
| 12:56:02 | gaoyuan | bauzas: maybe another time, I am chasing my bus | |
| 12:56:09 | bauzas | k | |
| 12:58:32 | moshele | mriedem: hi | |
| 12:58:41 | moshele | mriedem: can we release a new os-vif version? In os-vif I introduce hardware offload support for openvswitch https://review.openstack.org/#/c/460278/ | |
| 12:58:55 | moshele | mriedem: I need os-vif to be release so I can continue work on the nova side - https://review.openstack.org/#/c/398265/ | |
| 12:59:20 | mriedem | moshele: sure | |
| 12:59:36 | mriedem | moshele: has anyone proposed a new release in the releases repo? | |
| 13:00:11 | moshele | mriedem: I am not familiar with the release procedure | |
| 13:00:24 | edleafe | mriedem: the merge conflict is not simple. So don't wait up just for that patch | |
| 13:00:32 | mriedem | moshele: ok i'll do it | |
| 13:00:40 | mriedem | edleafe: i'll be up through the nova meeting | |
| 13:00:43 | mriedem | so another 90 min at least | |
| 13:00:54 | mriedem | i'm fing hardcore china style these days | |
| 13:02:21 | moshele | mriedem: thanks :) | |
| 13:04:06 | bauzas | edleafe: I'm confused | |
| 13:04:08 | mriedem | moshele: https://review.openstack.org/478923 | |
| 13:04:24 | bauzas | edleafe: did we merged a POST /allocation call to placement somewhere in the scheduler code yet? | |
| 13:04:43 | bauzas | can't find it in my git log | |
| 13:04:48 | bauzas | or I'm blind | |
| 13:05:17 | bauzas | I only see https://review.openstack.org/#/c/478291/ new which merged that night | |
| 13:05:24 | moshele | mriedem: ok thanks | |
| 13:05:39 | mriedem | bauzas: no not merged yet | |
| 13:05:53 | mriedem | bauzas: current series starts here https://review.openstack.org/#/c/475448/ | |
| 13:06:03 | bauzas | mriedem: the below edleafe's quote confused me | |
| 13:06:04 | bauzas | (14:47:00) edleafe: bauzas: The claim is now done in the manager, not the driver | |
| 13:06:17 | mriedem | in the series | |
| 13:06:20 | edleafe | bauzas: AFAIK, the allocations is a PUT, if that helps | |
| 13:06:22 | mriedem | as in, is now done in the series | |
| 13:06:28 | openstackgerrit | Takashi NATSUME proposed openstack/nova master: api-ref: Add X-Openstack-Request-Id description https://review.openstack.org/474847 | |
| 13:07:17 | bauzas | well, I know I'm French, but when I use the "merge" verb, it has a very clear meaning to me :) | |
| 13:07:29 | bauzas | anyway | |
| 13:07:41 | edleafe | bauzas: for background, the discussion moved from gerrit to IRC at around 14:20 here: http://eavesdrop.openstack.org/irclogs/%23openstack-nova/%23openstack-nova.2017-06-27.log.html | |
| 13:07:51 | bauzas | I loved those off days, but I hate how it hitted me | |
| 13:08:18 | edleafe | And to clarify, *in Jay's design*, the claim is now done in the manager | |
| 13:10:50 | openstackgerrit | Ildiko Vancsa proposed openstack/nova master: WIP/POC Implement new attach Cinder flow https://review.openstack.org/330285 | |
| 13:12:55 | bauzas | edleafe: just for my brain sanity, could we please use the word "proposed" instead of "done" if we speak of changes up ? :) | |
| 13:13:24 | bauzas | emotional lift, y'know :p | |
| 13:13:56 | gcb | mriedem: Can you revisit https://review.openstack.org/#/c/466787. it blocks oslo.messaging stable/newton release | |
| 13:21:09 | bauzas | edleafe: so, FWIW, after reading all of this, including https://review.openstack.org/#/c/478291/ I honestly think it's totally understandable to return HostStates instead of freeform dicts | |
| 13:21:30 | bauzas | the interface was very poor within the scheduler code, so I don't think it hurts | |
| 13:21:41 | mriedem | gcb: ok, i got really confused by the various backports for those changes, and they weren't consistent about the change id and some were abandoned, i.e. https://review.openstack.org/#/c/474456/ | |
| 13:22:01 | bauzas | the only drawback I would have said would be about custom drivers, but mriedem already got the word | |
| 13:22:11 | bauzas | which is that we no longer support custom drivers | |
| 13:22:32 | mriedem | bauzas: i already reported a docs bug abou that | |
| 13:22:39 | mriedem | since the docs tell you how to write your own scheduler driver | |
| 13:22:47 | bauzas | that sounds reasonable | |
| 13:22:56 | bauzas | we could still opt-in custom drivers | |
| 13:23:02 | mriedem | bauzas: https://bugs.launchpad.net/nova/+bug/1701105 | |
| 13:23:02 | openstack | Launchpad bug 1701105 in OpenStack Compute (nova) "Customizing the OpenStack Compute (nova) Scheduler in Operations Guide" [Medium,Confirmed] | |
| 13:23:03 | bauzas | after all, it's just adding a choice to an opt | |
| 13:23:26 | mriedem | if someone wants to hack a custom driver they are going to hack the config options, sure | |
| 13:23:32 | bauzas | but I strongly disapprove this | |
| 13:23:38 | mriedem | we just shouldn't advertise this as a built in capability that we support | |
| 13:23:42 | bauzas | given it's merely not interoperable | |
| 13:24:04 | bauzas | honestly, we did cut the possibility and nobody complainted | |
| 13:24:08 | bauzas | complained | |
| 13:24:18 | bauzas | if someone comes with a good usecase about that, we could discuss | |
| 13:24:32 | bauzas | but for the moment, no need to open a debate like this | |
| 13:25:06 | bauzas | and most of the operators wanting to run a specific scheduler can still use FilterScheduler and custom filters that call out to a 3rd party scheduler | |
| 13:26:42 | openstackgerrit | Merged openstack/python-novaclient master: Cleanup duplicated methods https://review.openstack.org/476493 | |
| 13:26:49 | mriedem | gcb: +W | |
| 13:27:08 | mriedem | bauzas: or they aren't upgraded to ocata yet | |
| 13:27:10 | mriedem | but anyway | |
| 13:27:40 | bauzas | mriedem: well, actually, that leaves me wonder one thing | |
| 13:27:59 | bauzas | mriedem: in the past, you could just disable all filters but just run your custom one | |
| 13:28:10 | bauzas | and then litterally do all nasty things you had in mind | |