Posted	Nick	Remark
#openstack-nova - 2017-04-13
21:58:54	sean-k-mooney	i belive the resolution involved allowing multiple port bindings during a live migrate
21:59:28	mriedem	dtp: i left some comments/questions,
21:59:37	mriedem	dtp: i'm green in this area so maybe my questions don't make sense
21:59:41	jangutter	sean-k-mooney: would you know some keywords I can google?
21:59:53	dtp	mriedem thanks!
22:00:57	mriedem	melwitt: i'm +2 on https://review.openstack.org/#/c/456227/ - we need to get that in for p1
22:01:01	mriedem	but no pressure
22:01:04	sean-k-mooney	jangutter: i belive its https://review.openstack.org/#/c/309416/
22:01:26	melwitt	mriedem: hah. looking at it now
22:02:06	jangutter	sean-k-mooney: thanks!
22:03:59	jangutter	would anyone else have cycles to do the final review on https://review.openstack.org/#/c/456386 ?
22:06:46	openstackgerrit	Matt Riedemann proposed openstack/nova-specs master: Blacklist yasfb 0.6.0 https://review.openstack.org/456769
22:06:57	sean-k-mooney	jaypipes: perhaps you could take a skim over jangutter path https://review.openstack.org/#/c/456386 if you have time. form an os-vif point of view it makes sense to me.
22:12:01	melwitt	mriedem: I think I'm missing how HostMappingNotFound is raised from things that don't appear to raise it
22:12:18	mriedem	melwitt: the decorator
22:12:29	mriedem	melwitt: which one specifically?
22:12:59	melwitt	mriedem: like service_get_by_compute_host, but I see the decorator now. duh
22:13:05	mriedem	ok
22:13:35	melwitt	meaning duh to me, I wasn't seeing where the lookup was coming from
22:13:52	mriedem	melwitt: i took that as a personal 'duh' to me
22:13:59	mriedem	i was highly offended
22:14:00	melwitt	I knew it!
22:19:25	melwitt	mriedem: I don't see how it can be raised from the compute_node_get here https://review.openstack.org/#/c/456227/4/nova/api/openstack/compute/hypervisors.py@192 and then things could fail at L195 not raising the webob exception
22:26:44	mriedem	melwitt: hmm yeah i don't see HostMappingNotFound raised from there either
22:27:14	melwitt	not harmful for the except block, but harmful not to have the try-except around the other call that can raise it
22:27:47	mriedem	well, it's not good to be handling exceptions that won't happen either
22:27:55	mriedem	melwitt: want to -1?
22:28:07	melwitt	yeah, I will. I'm just looking through the tests now
22:28:34	melwitt	yes, agreed not good
22:50:07	mriedem	dansmith: i don't want to pester but are we missing anything on https://review.openstack.org/#/c/456227/ ?
22:50:50	dansmith	gdi you people
22:51:45	dansmith	yeah I think I might've mixed up compute_get and service_get
22:51:53	dansmith	ffs
22:52:20	dansmith	I think I even explained it right in one of the earlier sets and then did the wrong thing
22:52:37	melwitt	got too excited
22:54:23	mriedem	it would be nice if we could just put a try/except around the entire os-hypervisors/os-services/os-hosts APIs
22:54:44	mriedem	heh, you know, we actually could
22:55:49	melwitt	now that would be sweet
22:56:50	mriedem	well we do have a catch all place in the wsgi layer
22:57:32	mriedem	this i think https://github.com/openstack/nova/blob/master/nova/api/openstack/wsgi.py#L396
22:58:08	mriedem	so, if we wanted to be lazy, we could just handle HostMappingNotFound there and return httpnotfound
22:58:11	mriedem	melwitt: dansmith: ^
22:58:15	mriedem	kind of a cop out though
22:58:33	dansmith	you'd rather backport that change? :)
22:58:43	mriedem	this is going to ocata?
22:59:00	mriedem	https://review.openstack.org/#/c/442162/ was pike
22:59:21	dansmith	oh yeaq
22:59:25	dansmith	I was thinking the p-1
22:59:31	mriedem	ok,
22:59:33	dansmith	you'd rather do that?
22:59:43	dansmith	I think I have this changed, fwiw
22:59:44	mriedem	i thought red hat backported multi-cell support for rhosp ocata :)
22:59:51	mriedem	how about we do the safe thing now
22:59:57	mriedem	and look at the lazy way later
22:59:58	claudiub|2	mriedem: hellou. just peeking around here for a bit. this blueprint was approved in ocata, and had the implementation up for review, but it didn't merge in ocata: https://blueprints.launchpad.net/nova/+spec/hyper-v-nested-virtualization
23:00:08	melwitt	it seems like doing that would be weird. though it is doing like Forbidden and Invalid that way but not NotFound. dunno the history on that
23:00:10	openstackgerrit	Dan Smith proposed openstack/nova master: Fix hypervisors api missing HostMappingNotFound handlers https://review.openstack.org/456227
23:00:15	claudiub|2	can it get re-approved? :D
23:00:22	mriedem	claudiub|2: for you? anything.
23:00:25	claudiub|2	<3
23:00:37	claudiub|2	seems like an early easter. :)
23:00:38	mriedem	melwitt: i think it's basically just a catch all for global types of failures
23:00:43	mriedem	which hostmappingnotfound kind of is
23:00:55	dansmith	mriedem: yeah that one is pretty easy to catch and always do the right thing
23:01:04	mriedem	claudiub|2: jesus hasn't come up and seen his shadow yet so don't count your chickens
23:01:05	dansmith	it's fairly specific about what happened and what should be done
23:01:07	melwitt	yeah. I was just noting it for whatever reason intentionally left out the base class for NotFound
23:01:07	mriedem	this easter could be different
23:01:19	dansmith	mriedem: could be super confusing for someone trying to track down the 404 though
23:01:25	mriedem	dansmith: yeah i totally agree
23:01:39	mriedem	dansmith: i never knew that existed until about a year ago when laski pointed it out
23:01:45	mriedem	^ is generally my story for everything
23:01:46	claudiub|2	well, it is a bit cloud-y lately
23:01:51	melwitt	laski, that rascal
23:03:29	claudiub|2	also, i've submitted the sr-iov bp: https://blueprints.launchpad.net/nova/+spec/hyperv-sriov no implementation yet i'm afraid, apparently there's a whole guide to follow just to configure the nics properly.
23:04:24	mriedem	claudiub|2: sounds about right
23:04:31	mriedem	claudiub|2: do you get a long weekend?
23:05:05	claudiub|2	well, technically i'm free on monday, i'll still be around
23:05:16	claudiub|2	i don't like vacations too much. :D
23:05:23	mriedem	you can read the sriov manual
23:05:42	claudiub|2	yeah, that is in my plan. :D
23:06:14	claudiub|2	and question: can we live-migrate vms with sr-iov?
23:06:19	mriedem	claudiub|2: no
23:06:23	mriedem	resource tracker issue
23:06:24	claudiub|2	hm, interesting
23:06:26	mriedem	*issues
23:06:31	mriedem	cfriesen knows about it
23:06:40	claudiub|2	is that a nova limitation, or a hypervisor limitation?
23:09:47	claudiub|2	so, alaski is like beetlejuice - say his name too many times and he shows up. :D
23:10:53	alaski	I'm glad I appear to have explained things to folks in the past, because I don't remember much these days
23:15:26	melwitt	you missed out on some cerebral shelve feature talk today
23:20:37	alaski	I hope it was good. I am very much not proud of introducing that feature.
23:25:38	melwitt	hm ... define good. mriedem found that upon shelve offload, volumes are not disconnected and then also when network is cleaned up, that only does something for nova-net through the network_api, the neutron implementation is a no-op pass
23:27:00	melwitt	so it pretty much relies on the unshelve putting the instance back on the same host it was shelved offloaded from, which seems really unlikely
23:29:15	alaski	yeah, I would not expect that to happen much in practice
23:33:17	gmann	mriedem: thank i almost forgot about that bug, ll check patch
23:37:04	cfriesen	claudiub|2: the first issue is that we need to get resource tracking in general working over live migration. (see https://review.openstack.org/#/c/244489 and the follow-on patches)
23:39:23	cfriesen	do people actually use shelve? If so, why hasn't anyone noticed this?
23:40:22	alaski	maybe not using volumes, or neutron?
23:40:41	alaski	it did seem to have a small following, despite discouragement about using it

Earlier   Later