Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-04-13
14:56:51 mriedem with newer libvirt/qemu
14:57:06 mriedem which we were told by some (cough cough) it would fix all of our problems :)
14:58:58 kashyap mriedem: I never make blanket promises, certainly to _you_ that it will "fix all the problems"
14:59:26 mriedem https://review.openstack.org/456656
14:59:28 kashyap s/to/not to/
14:59:39 mriedem kashyap: i'm just teasing
14:59:49 kashyap I know, I forgot to add a smiley
15:07:34 owalsh kashyap: that looks like a packaging issue, nc not installed
15:08:48 jangutter :O "nc not installed" conjures the same horror in me as "gcc not installed".
15:10:16 sdague mriedem: the fix is merged now, right?
15:10:53 mriedem clarks?
15:10:56 mriedem that merged last night
15:10:57 mriedem this is something else
15:11:14 sdague mriedem: https://review.openstack.org/#/c/456388/
15:11:21 sdague no, it didn't merge until this morning
15:11:50 sdague also, it seems totally weird to me that libvirt would shell out to nc
15:12:56 dansmith sdague: it has definitely used nc for things
15:13:26 dansmith sdague: used to be you couldn't connect a fedora virt-manager to an ubuntu host without installing netcat, or something like that
15:13:33 dansmith the reverse maybe, I forget
15:19:15 sdague so I think this is probably the problematic commit for us - https://libvirt.org/git/?p=libvirt.git;a=commitdiff;h=22eaee8e011645162e4b00d313bf2abdd4eb54e5
15:19:30 sdague probably good to get an ubuntu packaging bug openned up on that
15:20:17 sdague clarkb: did you do that yet?
15:21:52 owalsh sdague: that's for qemu+libssh, the qemu+ssh driver always requried netcat
15:25:28 sfinucan owalsh: Regarding https://review.openstack.org/456571, are you saying we need user and port options even if we don't deprecate 'live_migration_uri'
15:27:10 sdague it looks like the libvirt-daemon recommends netcat-openbsd, but does not require it
15:27:12 sdague Recommends: qemu-kvm | qemu (>= 0.9.1), libxml2-utils, netcat-openbsd
15:27:38 owalsh sfinucan: need to use live_migration_inbound_addr to set the IP network for live migration. We can't do this with live_migration_uri AFAIK
15:29:51 owalsh sfinucan: I assume that's one of the main reasons for deprecating live_migration_uri
15:30:44 owalsh sfinucan: but to use SSH we need to pass user, maybe port too, and almost certainly extraoptions, and there is no way to set this when using live_migration_inbound_addr
15:37:19 sfinucan owalsh: So for 'live_migration_uri', you could conceivably set it to something like 'xen+ssh://rjones@%s/?no_verify=1'. That has both a user and extra_options
15:37:46 sfinucan owalsh: Do we want the same functionality for whatever's using live_migration_inbound_addr?
15:39:49 owalsh sfinucan: yes, currently do set the user via live_migration_uri, but we can't control which network it users. %s is always replaced with the remote hostname
15:42:00 sfinucan owalsh: ...and live_migration_inbound_addr provides an alternative to that, but it doesn't allow you to a username or extra_specs currently, right?
15:43:59 sfinucan owalsh: If we allowed you to _not_ specify '%s' for 'live_migration_uri' and instead set a static destination/IP network, would that achieve what you want?
15:44:20 cdent edleafe: are your resource tracker related functional tests sort of on hold?
15:44:46 owalsh sfinucan: you have to specify %s - it's replaced with the hostname of the live miragation target
15:45:02 sfinucan owalsh: That's what I'm saying. What if I removed that requirement
15:45:26 sfinucan Or are the two complimentary somehow?
15:46:35 owalsh so live_migration_uri = foo://bar@%s/ will tranlate to foo://bar@nova1/ when migrating to nova1, and foo:/bar@nova2/ when migrating to nova2
15:48:07 owalsh when using the new approach when migrating to nova1 we get live_migration_inbound addr etc.. for nova1 to build the url to migrate to it
15:48:15 owalsh IIRC anyway
15:49:35 owalsh it been a while since I looked at it but I think there was consensus that the new options are required
15:49:47 openstackgerrit Artom Lifshitz proposed openstack/nova master: Plumbing for tagged nic attachment https://review.openstack.org/415622
15:49:48 openstackgerrit Artom Lifshitz proposed openstack/nova master: Delete nic metadata when detaching interface https://review.openstack.org/435387
15:49:49 openstackgerrit Artom Lifshitz proposed openstack/nova master: Libvirt support for tagged volume attachment https://review.openstack.org/418100
15:49:52 openstackgerrit Artom Lifshitz proposed openstack/nova master: Delete disk metadata when detaching volume https://review.openstack.org/435082
15:49:55 openstackgerrit Artom Lifshitz proposed openstack/nova master: Plumbing for tagged volume attachment https://review.openstack.org/391941
15:49:56 openstackgerrit Artom Lifshitz proposed openstack/nova master: API support for tagged device attachment https://review.openstack.org/376021
15:49:57 openstackgerrit Artom Lifshitz proposed openstack/nova master: Libvirt support for tagged nic attachment https://review.openstack.org/415520
15:57:30 edleafe cdent: yeah, more because I got distracted with other stuff while waiting for more feedback
15:58:18 cdent edleafe: yeah, just checking. I had a reminder to look at them but didn't have the current brain to say much in this round.
15:58:44 edleafe cdent: ok, remind me next week if you think of it
15:58:53 cdent ✔
15:59:22 mriedem dansmith: before you change things in https://review.openstack.org/#/c/456227/ i've left some comments
15:59:29 mriedem looks like some were at the same time you were replying to sfinucan
15:59:30 openstackgerrit Dan Smith proposed openstack/nova master: Fix hypervisors api missing HostMappingNotFound handlers https://review.openstack.org/456227
15:59:32 dansmith hah
15:59:33 mriedem doh
15:59:34 mriedem gdi
15:59:41 dansmith was literally just pushing when you said that
16:02:43 ildikov mriedem: johnthetubaguy: are you avaiable for the Cinder-Nova meeting?
16:02:54 mriedem yeah
16:18:16 openstackgerrit Daniel Pawlik proposed openstack/nova master: Changed network_info variable in test_create_configdrive https://review.openstack.org/436953
16:21:11 dtp good morning. can we get a ruling on https://blueprints.launchpad.net/nova/+spec/conf-network-metadata-always-static-ip, please?
16:23:38 kashyap mriedem: That live mig job failure is due to missing `nc` (Netcat), which seems to be resolved by clarkb's -- https://review.openstack.org/#/c/456388/ ("Install netcat for libvirt live migration")
16:23:43 kashyap That merged about 3 hours ago
16:23:58 mriedem kashyap: ok, i had thought that merged last night but forgot about the devstack dependency
16:24:05 mriedem i'll check logstash in a bit to see if it's dropping off
16:24:33 kashyap mriedem: See! You seem to blame libvirt at the drop of a hat :P
16:24:52 kashyap You actually didn't
16:25:17 ngupta Hi everyone, new to this channel. I want to know where does OpenStack stand with supporting encryption for Block Storage using the RBD driver.
16:27:11 openstackgerrit OpenStack Proposal Bot proposed openstack/nova master: Updated from global requirements https://review.openstack.org/455960
16:34:34 mriedem dansmith: is it just me, or is it weird that we don't terminate volume connections on the host when we shelve offload?
16:34:41 mriedem since when you unshelve, you can end up on another host
16:36:35 dansmith mriedem: yeah.. I assume that'd be a bug
16:36:49 mriedem like, the most latent bug ever...
16:37:00 dansmith I think there might've been some things that were left in place on the off chance it landed on the same host, but can't imagine that's right
16:37:00 mriedem i keep thinking i'm missing something obvious
16:37:11 dansmith unless we call to the old compute somehow, but we've lost that info I think
16:37:23 dansmith yeah, seems too obviously broken
16:37:52 mriedem conductor manager for unshelve calls out to scheduler to get new hosts
16:37:55 openstackgerrit Jianghua Wang proposed openstack/nova-specs master: Support VGPU https://review.openstack.org/450122
16:38:03 mriedem plus, you could have unshelved from a host that is now down or gone
16:38:11 mriedem *shelved
16:38:17 dansmith yeah
16:41:40 mriedem oh btw, pike-1 milestone today
16:41:46 mriedem i had forgotten about the need for the tag
16:42:35 jianghuaw_ jaypipes, sbauza: could you help to look at this VGPU spec?
16:42:36 jianghuaw_ https://review.openstack.org/450122
16:42:37 melwitt_ ngupta: I found this old spec about it https://review.openstack.org/#/c/239798/ and the consensus there mentions the way forward for rbd encrypted volumes should be with qemu native rbd encryption, but I don't think anyone worked on it yet
16:43:11 mriedem dansmith: also looks like we don't unbind ports from the host on shelve offload
16:43:20 mriedem this is all just very odd
16:43:59 openstackgerrit Stephen Finucane proposed openstack/nova master: WIP! conf: Add three new '[libvirt] live_migration_*' options https://review.openstack.org/456571
16:44:00 openstackgerrit Stephen Finucane proposed openstack/nova master: conf: Gather 'live_migration_scheme', 'live_migration_inbound_addr' https://review.openstack.org/456572
16:44:00 melwitt mriedem: yeah, that's "intentional"? I noticed it just leaves everything there as-is, network connection, volume connection, etc
16:44:00 openstackgerrit Stephen Finucane proposed openstack/nova master: conf: Convert 'live_migration_inbound_addr' to HostAddressOpt https://review.openstack.org/456573
16:44:45 sfinucan owalsh: I've to go to tag, so that's my WIP patches. Lemme know if you think that's the right direction. Will finish them tonight or Tuesday
16:45:37 melwitt it assumes it will go back on the same host, but I don't know if it does or can do anything to guarantee that. with the network part, I could see it wanting to keep the IP allocated so that it can't be given out (and network possibly run out of IPs) when the person comes back to unshelve. same concept as keeping the quota occupied during offload
16:46:59 owalsh sfinucan: cool, thanks
16:47:10 dansmith mriedem: must be missing something

Earlier   Later