Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-07-13
14:31:55 jaypipes bauzas: hmm? no, nothing to do with amending reqspec
14:32:04 bauzas it's just because of the affinity filter, but not sure I understand your problem honestly
14:32:21 bauzas jaypipes: you were telling dan you had a problem with that yesterday night :)
14:32:44 jaypipes bauzas: oh, yeah, sorry, that was a different rant :)
14:33:04 bauzas anway, if I can help you...
14:33:17 bauzas at least by sharing thoughts
14:33:19 jaypipes bauzas: my general annoyance with the scheduler is its tight coupling of state with decision-making.
14:34:00 jaypipes bauzas: And I'd much prefer to deal with things like num_instances > 0 outside of the filter scheduler.
14:34:46 bauzas jaypipes: the real problem is that since we really only supported FilterScheduler (ChanceScheduler being only for tests or crazy operators), we really didn't had a clear interface
14:35:09 bauzas so we built a pile of dependencies between the driver, the host manager and the filters
14:35:29 jaypipes bauzas: that and the unit tests are a pile of spaghetti that test way too much surface area and therefore if you change any little thing, a million tests crash and burn.
14:35:47 bauzas the code itself isn't very large, that's just the fact that we're poorly defining our internal interfaces
14:35:56 jaypipes that, and there's no reason for the WeighedHost class to begin with.
14:36:25 bauzas I just feel we need to be pragmatic for Pike and do the nice stuff later
14:36:30 jaypipes but whatever, I'm back off the cliff edge and making slow progress.
14:36:47 jaypipes bauzas: not sure what you mean by "nice stuff"./
14:36:55 bauzas honestly, our internal interfaces are... internal, so I don't care with us changing those a couple of times
14:37:26 bauzas jaypipes: I'm talking of getting results first, and then clean up our model later
14:37:50 jaypipes bauzas: yeah... that's what has caused all of the issues in the first place :) shoot first, ask questions later.
14:38:02 bauzas I know
14:38:26 bauzas do you remember I started changing those interfaces in 2014 ?
14:38:38 bauzas that's just a funny story
14:46:01 openstackgerrit Takashi NATSUME proposed openstack/nova master: Enable cold migration with target host(1/2) https://review.openstack.org/408955
14:47:02 jaypipes bauzas: I know you know. And I know you know that I know you know.
14:48:00 bauzas jaypipes: wake me before you know-know
14:48:05 bauzas aaaaaah
14:48:18 bauzas wake me *up before you know-know
14:48:45 cdent good effort though
14:49:00 openstackgerrit Stephen Finucane proposed openstack/nova master: trivial: Remove unnecessary function https://review.openstack.org/483433
14:51:52 jaypipes bauzas: hehe
14:52:53 mriedem so we're making wham! jokes now?
14:53:28 dansmith oh jeez, you'll get mriedem all worked up
14:53:36 mriedem well look who it is
14:53:46 mriedem someone has a 'wham!' keyword listener
14:53:54 dansmith heh
14:54:07 bauzas :)
14:55:04 efried mordred mriedem Now a good time to talk about ksa/sta/os-service-types releases in support of use-service-catalog-for-endpoints ?
14:56:13 mriedem i need some time
14:56:58 edleafe jaypipes: I've bookmarked that quote
14:58:24 jaypipes edleafe: ha! it was paraphrasing White Goodman from "Dodgeball: A True Underdog Story" (http://www.imdb.com/title/tt0364725/quotes)
14:58:38 jaypipes "I know you. You know you. And I know you know that I know you."
14:58:51 edleafe oh, not that quote. http://p.anticdent.org/38I
14:59:19 jaypipes edleafe: ooohhhhh, that one.
14:59:31 edleafe jaypipes: :)
15:02:52 mriedem stvnoyes: i've gone through the compute manager changes on https://review.openstack.org/#/c/463987/ - there is a definite bug in the way that _detach_volume is called
15:03:32 mriedem stvnoyes: beyond that, i'm concerned that since we don't always call rollback_live_migration_at_destination, we might not cleanup properly
15:04:52 mordred efried_bbiab, mriedem: ping me when you're ready
15:25:01 edleafe jaypipes: let me check my understanding of compute-land with you
15:25:13 edleafe jaypipes: There is a compute service. It has a single resource tracker. That RT can have 1 or more compute nodes
15:25:30 edleafe jaypipes: Normally it's one (VMs), but for Ironic it can be (and usually is) >1
15:25:36 edleafe jaypipes: is that correct?
15:25:40 jaypipes yup
15:25:56 edleafe jaypipes: ok thx
15:25:58 dansmith correct in the technical sense
15:26:04 jaypipes edleafe: always a single resource tracker. that resource tracker tracks >1 compute node objects in case of ironic.
15:26:05 dansmith not correct in the moral should-this-be-done sense
15:26:11 openstackgerrit Takashi NATSUME proposed openstack/nova master: List/show all server migration types (1/2) https://review.openstack.org/430608
15:26:28 edleafe dansmith: my soul is already unclean
15:26:52 bauzas edleafe: because jaypipes fixed that :)
15:27:10 bauzas edleafe: we previously had 1 RT per compute "node"
15:27:21 bauzas <3 jaypipes
15:28:19 edleafe bauzas: get a room, you two
15:28:45 jaypipes lol
15:29:05 isantosp Hi, I have an issue on nova newton, im trying to set ipv6 taking it from the port creation in neutron newton, apparently neutron is doing his job perfectly, the problem is when nova try to call port_crete_minimal, I have the following: ```
15:29:09 bauzas edleafe: you mean #openstack-loves-jaypipes IRC room ? :p
15:29:33 isantosp subnet_id_v4 = host_subnets['host']['available_random_subnet_v4']
15:29:35 isantosp port_req_body['port']['fixed_ips'] = [
15:29:37 isantosp {'subnet_id': str(subnet_id_v4)}]
15:29:39 isantosp yes
15:29:55 edleafe bauzas: there are several such rooms, I imagine
15:30:40 isantosp In the creation of the instance, i see this error: {"message": "'subnets'", "code": 500}
15:31:07 dansmith melwitt: I think you should roll on with the update to that patch. I'm to the tests, and my will to live is waning
15:31:35 isantosp and this one in the logs
15:31:38 isantosp File "/usr/lib/python2.7/site-packages/nova/network/neutronv2/api.py", line 392, in _create_port_minimal
15:31:40 isantosp host_subnets = port_client.show_host(instance.host)
15:31:42 isantosp KeyError: 'subnets'
15:32:46 isantosp any ideas? the line 392 that is pointed there, it has nothing to solve
15:36:15 openstackgerrit Alex Szarka proposed openstack/nova master: Transform the transformed notifications functional tests https://review.openstack.org/483448
15:40:34 sean-k-mooney jangutter: well what specifically is the gap. you can use https://github.com/Juniper/contrail-nova-vif-driver/blob/master/vif_plug_vrouter/vrouter.py as an example of an external plugin
15:40:47 openstackgerrit Takashi NATSUME proposed openstack/nova master: Enable cold migration with target host(2/2) https://review.openstack.org/408964
15:43:10 sean-k-mooney jangutter: for the sriov type you should call your plugins plug method with a VIFHostDevice and for the virtio forworder case you should use VIFVHostUser with a costom port profile if you need to pass vendor specific info.
15:43:54 openstackgerrit Takashi NATSUME proposed openstack/nova master: api-ref: Add parameters in cold migrate action https://review.openstack.org/410042
15:44:01 openstackgerrit Takashi NATSUME proposed openstack/nova master: api-ref: Add parameters in cold migrate action https://review.openstack.org/410042
15:45:22 openstackgerrit Takashi NATSUME proposed openstack/nova master: Add functional tests for cold migration to same host https://review.openstack.org/414926
15:45:31 openstackgerrit Takashi NATSUME proposed openstack/nova master: Add functional tests for cold migration to same host https://review.openstack.org/414926
15:48:17 openstackgerrit Alex Szarka proposed openstack/nova master: Transform the transformed notifications functional tests https://review.openstack.org/483448
15:48:27 sean-k-mooney jangutter: was your question more how do i tell nova to allocate the vf and call the plugin?
15:49:16 sean-k-mooney jangutter: without a review too look at its a little hard to advise what do next
15:52:31 openstackgerrit Alex Szarka proposed openstack/nova master: Transform instance.exists notification https://review.openstack.org/403660
15:52:50 openstackgerrit Alex Szarka proposed openstack/nova master: Transform instance.resize.error notifications https://review.openstack.org/417882
15:53:09 mriedem jaypipes: dansmith: can one of you put the priority reviews docs patch through https://review.openstack.org/#/c/470420/
15:53:45 dansmith sure
15:53:58 jangutter sean-k-mooney: your wish is my command, just as we speak the clearance came through.
15:54:52 jangutter So my next newbie question to the room is: should I throw away the old reviews and create new ones, or replace them? what's the status quo?
15:55:52 mriedem dansmith: https://review.openstack.org/#/c/445142/ also needs a rebase
15:56:29 dansmith mriedem: ac
15:56:30 dansmith k
15:56:31 sean-k-mooney jangutter: cool. am it really up to you. you can keep the change ids if the new code is a nautal evolution of the previous patches or abandon the old patchs and leave a link to the new patchset if its substially different
15:57:38 jaypipes jangutter: just Abandon the old patches.
15:57:45 jaypipes jangutter: patches are cheap :)

Earlier   Later