Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-04-14
15:33:58 mriedem and a misconfiguration
15:34:05 openstackgerrit Matt Riedemann proposed openstack/nova master: Add release note and update cell install guide for multi-cell limitations https://review.openstack.org/456923
15:34:12 mriedem melwitt: ^
15:35:40 mriedem EmilienM: owalsh: ^
15:38:26 efried This thing about having to run a new discovery command to make stuff work... Could it not just be made to work out of the box if you only have one cell?
15:47:49 efried [irony: when --help doesn't automatically page, and you have to say --help|less]
15:50:14 mriedem efried: we have a config option if you want things discovered automatically,
15:50:20 mriedem but it's a bit heavy for large deployments
15:50:27 mriedem dan has a series up to improve that
15:50:32 jangutter Is there a way to refer to today's IRC chat in a spec?
15:50:34 efried mriedem Okay, neat.
15:50:40 mriedem jangutter: yeah, esc
15:50:41 mriedem *Sec
15:50:43 efried http://eavesdrop.openstack.org/irclogs/%23openstack-nova/
15:50:47 efried jangutter ^^
15:50:48 mriedem efried: https://blueprints.launchpad.net/nova/+spec/discover-hosts-faster
15:51:03 mriedem jangutter: http://eavesdrop.openstack.org/irclogs/%23openstack-nova/%23openstack-nova.2017-04-14.log.html#t2017-04-14T14:44:06
15:52:24 jangutter Whatever you write is now memorialized forever. Or until cats reclaim their rightful place on top of the food chain.
15:54:10 mriedem our cat can easily be defeated with a bag of chips
15:54:47 openstackgerrit Jan Gutter proposed openstack/nova-specs master: Spec to enable Netronome SmartNIC support https://review.openstack.org/456386
15:57:26 openstackgerrit Matt Riedemann proposed openstack/nova master: Tell people that the nova-cells man page is for cells v1 https://review.openstack.org/456925
15:57:29 jangutter Another newbie question. I've replied to a bunch of comments on an older review. They're in "draft" in gerrit now, am I too late?
15:57:41 mriedem jangutter: no, you can still publish those
15:57:43 efried jangutter No - go to that patch set and they'll show up.
15:58:15 efried Still not entirely understanding why a compute node can't automatically 'discover' itself into a cell if there's only one cell...
15:58:31 mriedem efried: it would require an upcall from the compute to the api layer, which we don't want
15:58:35 mriedem upcalls are the devil
15:58:43 mriedem s/the/a/
15:58:53 openstackgerrit Ildiko Vancsa proposed openstack/nova master: Add Cinder v3 detach call to _terminate_volume_connections https://review.openstack.org/456896
15:59:12 mriedem efried: also, that would require either an upcall or putting the nova api db credentials in the compute config
15:59:16 mriedem another thing we don't want
16:00:20 efried mriedem And all of that is different from how a compute node registered itself B.C. (before cells)?
16:00:52 mriedem yes
16:00:58 mriedem in the before times,
16:01:01 mriedem there was no api db
16:01:08 mriedem there were no host mappings in the api db
16:01:36 mriedem when you start nova-compute, it creates it's own compute_nodes table record for that host (happens in ResourceTracker.update_available_resource eventually)
16:01:57 mriedem when you start the service it also creates a record in the services table for the nova-compute binary
16:02:05 mriedem those all get mangled up in the api somehow
16:02:08 mriedem feeblepipes loves it
16:02:32 mriedem but with no cells, i.e. regular old stock nova, there was just a single db and everything was in there
16:03:05 mriedem efried: are you going to be in boston?
16:03:10 efried mriedem Yessir.
16:03:45 mriedem then put https://www.openstack.org/summit/boston-2017/summit-schedule/events/18727/ on your schedule
16:04:18 efried mriedem Roger wilco, and thanks.
16:12:46 EmilienM mriedem: ack
16:33:37 feeblepipes jangutter: yes, your summary of the requested changes above is good.
16:36:09 jangutter feeblepipes: if all else fails there's also that list of emulated vNICs that qemu uses: confusingly it's called "hw_vif_model"
16:36:23 feeblepipes jangutter: yeah, I know :(
16:43:06 feeblepipes peter-hamilton: approved.
16:43:30 peter-hamilton feeblepipes: awesome, thanks!
16:43:46 feeblepipes peter-hamilton: np. sorry again for the delay.
16:44:02 peter-hamilton feeblepipes: no problem, i hope you recover quickly
16:44:24 feeblepipes peter-hamilton: about 50% today. which is better than yesterday's 10% :)
16:44:52 peter-hamilton feeblepipes: yep, i've been there
16:45:05 peter-hamilton feeblepipes: thanks for the comment on the history btw
16:45:14 peter-hamilton feeblepipes: i find it helps to summarize decisions and the back-and-forth over the months
16:45:30 feeblepipes yeah, it actually makes it a bit easier for those of us kinda coming into a spec late in its lifetime.
16:45:30 peter-hamilton feeblepipes: no one else seems to use it for much
16:45:42 feeblepipes ++
16:52:44 jangutter Any more actions on https://review.openstack.org/456386 I need to take?
16:56:52 feeblepipes jangutter: did you push a new revision with wording around the "genericness" of parameters passed via os-vif/Neutron port binding profile?
16:57:49 jangutter feeblepipes: yeah, that's been updated, plus a link to the chat.
16:58:56 feeblepipes jangutter: awesome. reviewing again.
17:06:54 feeblepipes jangutter: k, +2 from me. mriedem, feel free to give 456386 another spin.
17:07:37 jangutter feeblepipes: thanks!
17:07:44 feeblepipes jangutter: np
17:08:03 feeblepipes dansmith: did you see johnthetubaguy's request for your review on https://review.openstack.org/#/c/407237/?
17:37:54 owalsh mriedem: yea, the related change in tripleo was to handle this
18:08:02 mriedem feeblepipes: jangutter: done
18:08:28 feeblepipes mriedem: thanks, Matt. and apologies for the delay on these after spec freeze.
18:08:33 mriedem p
18:08:35 mriedem *np
18:08:48 mriedem you get an exception because jay was sick
18:08:53 mriedem was/is
18:08:53 feeblepipes mriedem: I'm trying to get the GPU spec finalized and the one about listing migrations.
18:09:10 mriedem feeblepipes: i have not looked at the gpu spec and probably won't
18:09:20 mriedem i plan on looking at the listing migration types one though
18:09:24 feeblepipes mriedem: k
18:09:27 mriedem because that api annoys me
18:09:35 feeblepipes mriedem: I have a number of concerns on it, yes.
18:09:45 mriedem lack of gpu complexity in nova doesn't annoy me
18:09:58 jangutter feeblepipes, mriedem, you have my eternal gratitude!
18:10:36 mriedem feeblepipes: e.g. https://bugs.launchpad.net/nova/+bug/1668747
18:10:36 openstack Launchpad bug 1668747 in OpenStack Compute (nova) "api-ref: os-migrations is missing response details and information about v2.23" [Medium,Confirmed]
18:10:42 feeblepipes jangutter: kein Problem!
18:11:18 feeblepipes jangutter: or maybe "geen probleem!"? :)
18:11:37 jangutter It's the second one!
18:11:41 feeblepipes aha!
18:12:44 jangutter Fair warning, don't let us Afrikaans people start teaching you phrases. You might end up learning the ultimate language to insult someone in.
18:13:03 feeblepipes :)
18:23:52 openstackgerrit Chris Dent proposed openstack/nova master: Update resource tracker to PUT custom resource classes https://review.openstack.org/456915
18:32:44 mriedem feeblepipes: while you're close to passing out or vomiting, let me throw this on you https://review.openstack.org/#/c/453272/
18:35:38 feeblepipes mriedem: what's the difference between the instance action of evacuate vs the instance action of rebuild?
18:36:09 mriedem with evacuate the source compute host is dead
18:36:13 mriedem with rebuild it's not
18:36:24 feeblepipes mriedem: how would the user know which is which?
18:36:31 mriedem users don't evacuate
18:36:33 mriedem since it's a migratoin
18:36:34 mriedem *migratoin

Earlier   Later