| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-04-13 | |||
| 10:04:59 | mdbooth | How about we just explicitly don't support it? | |
| 10:05:16 | mdbooth | It wouldn't be a regression, as it previously blew away your data anyway | |
| 10:05:23 | lyarwood | mdbooth: yeah it's hackaround or block it tbh | |
| 10:05:44 | lyarwood | mdbooth: there are downstream users jumping up and down asking for this | |
| 10:05:57 | mdbooth | I think the fix should be to pass a different path to qemu | |
| 10:06:38 | mdbooth | lyarwood: Hmm, we're still risking their data, though | |
| 10:06:55 | mdbooth | Surely cinder operations are going to be broken while Nova is doing its thing | |
| 10:06:59 | mdbooth | For eg | |
| 10:07:24 | lyarwood | mdbooth: yeah with this renaming they will be | |
| 10:07:46 | mdbooth | Ideally we'd leave it alone entirely, create a dev mapper device on top of it, then use the dev mapper device instead of the original device, which remains untouched | |
| 10:08:31 | mdbooth | IIRC we're only doing this hack because the existing code doesn't allow the encryptor to modify the path used by other Nova code | |
| 10:09:20 | lyarwood | mdbooth: yeah because we do all of this _after_ creating the libvirtdiskconfig objects iirc | |
| 10:09:25 | mdbooth | I'm really not comfortable with Nova moving cinder volumes around physically, especially when we don't tell cinder about it | |
| 10:09:48 | lyarwood | ack that's ah great point that I missed tbh | |
| 10:09:52 | lyarwood | a* | |
| 10:10:09 | mdbooth | What's the plan for a real fix, btw? | |
| 10:10:56 | lyarwood | well, my QEMU 2.6 native LUKS spec landed yesterday so that's one way of `fixing` this | |
| 10:12:03 | mdbooth | lyarwood: Of course. | |
| 10:12:04 | lyarwood | really not sure what we can do for the existing encryptors aside from block this for file based volumes for now | |
| 10:13:28 | mdbooth | lyarwood: Honestly my short term fix would be to raise an unsupported exception instead of deleting the user's volume | |
| 10:14:24 | mdbooth | lyarwood: Long term qemu native luks | |
| 10:14:40 | lyarwood | mdbooth: cool, I'll post a new change for that later today. | |
| 10:14:59 | mdbooth | If we're going to fix it as it is, I think a Nova refactor to allow us to switch out the path | |
| 10:15:42 | mdbooth | Disgusting hack potential: change the nfs volume driver to use a loopback device instead of the file directly | |
| 10:28:28 | johnthetubaguy | cdent: so the PUT request has no body in your spec, is that correct? | |
| 10:28:51 | cdent | johnthetubaguy: that's right, because there's nothing useful to put in it | |
| 10:29:14 | johnthetubaguy | cdent: saves the need to check it matches the URL or fail, makes sense I think | |
| 10:30:26 | openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Use plain routes list for flavor endpoint instead of stevedore[1] https://review.openstack.org/456549 | |
| 10:31:30 | cdent | johnthetubaguy, bauzas: I'll make a new version in few minutes to include the info that bauzas asked for | |
| 10:40:53 | sfinucan | cdent: In the absence of jaypipes (wherever he's gone to), could you look at https://review.openstack.org/#/c/361140/ again and see if there's any must-have changes needed today? | |
| 10:41:13 | cdent | sfinucan: sure, will do right after fixing this other spec | |
| 10:41:23 | sfinucan | cdent: spot on | |
| 10:51:26 | openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Use plain routes list for flavor endpoint instead of stevedore[1] https://review.openstack.org/456549 | |
| 10:53:25 | sfinucan | pkoniszewski: You around? Question on the 'live_migration_scheme' and 'live_migraton_inbound_addr' options | |
| 10:53:52 | pkoniszewski | sfinucan: yes, what's up? | |
| 10:54:08 | openstackgerrit | Chris Dent proposed openstack/nova-specs master: A spec for idempotent custom resource class creation https://review.openstack.org/453732 | |
| 10:54:18 | sfinucan | pkoniszewski: What was your rationale behind df334b4 ? | |
| 10:54:31 | sfinucan | I'm guessing it's a usability thing, but was there anything more subtle than that? | |
| 10:54:55 | sfinucan | pkoniszewski: https://github.com/openstack/nova/commit/df334b4f414b8c341709df37ca10065d3b50fcef for reference | |
| 10:56:36 | pkoniszewski | sfinucan: so my point was to deprecate and finally remove live_migration_uri which is used for hacking the interface used for live migrations, now there is live_migration_inbound_addr to choose an interface | |
| 10:56:45 | pkoniszewski | also rationale behind df334b4 was to keep an option to change scheme for live migration, some of operators use SSH instead of TCP | |
| 10:56:57 | pkoniszewski | i didn't want to force anyone to switch to TCP from SSH | |
| 10:57:48 | openstack | Launchpad bug 1671288 in OpenStack Compute (nova) "live_migration_uri to live_migration_scheme SSH settings" [Undecided,Confirmed] - Assigned to Stephen Finucane (stephenfinucane) | |
| 10:57:48 | sfinucan | pkoniszewski: Sounds fair. Unfortunately though some operators also use extraparamters, e.g. '?command=xyz' https://bugs.launchpad.net/nova/+bug/1671288 | |
| 10:58:20 | pkoniszewski | yeah, i noticed that like two weeks ago | |
| 10:58:27 | pkoniszewski | so it's a trade off right now | |
| 10:58:57 | sfinucan | pkoniszewski: trade-off between...? :) | |
| 11:00:42 | sfinucan | pkoniszewski: I was going to add a 'live_migration_extraparamters' DictOpt to allow folks to specify those options if they want them | |
| 11:00:48 | pkoniszewski | keep live_migration_uri and document how it can be used/let people google how to use it, or implement a few configuration options to cover operators use cases (like SSH username and key) and remove live_migration_uri entirely | |
| 11:01:02 | pkoniszewski | or a DictOpt, it also sounds good | |
| 11:01:08 | sfinucan | ...but then I though, wouldn't it be easier to keep live_migration_uri and mark it as 'advanced' | |
| 11:01:16 | sfinucan | https://docs.openstack.org/developer/oslo.config/opts.html#oslo_config.cfg.Opt | |
| 11:01:40 | sfinucan | pkoniszewski: seems we're weighing up similar trade-offs so :) | |
| 11:03:20 | johnthetubaguy | sfinucan: pkoniszewski: I was thinking about that before as well | |
| 11:03:35 | johnthetubaguy | honestly, folks seem used to using the URL, maybe we just keep it | |
| 11:03:50 | johnthetubaguy | and add a description about how you change SSH params in there | |
| 11:04:16 | sfinucan | right, I'll revert the deprecation so, mark it as advanced, and expand the help with a link to the libvirt docs | |
| 11:04:29 | sfinucan | that'll keep both standard and "advanced" operators happy | |
| 11:04:47 | sfinucan | with minimal code changes (read: easy backporting) | |
| 11:04:48 | johnthetubaguy | I am tempted to include the SSH case in the description | |
| 11:05:00 | sfinucan | johnthetubaguy: e.g.? | |
| 11:05:03 | johnthetubaguy | makes sense to have the same key for migration and livemigration | |
| 11:05:13 | johnthetubaguy | sfinucan: it was in one of the bugs somewhere | |
| 11:05:49 | johnthetubaguy | sfinucan: here we go live_migration_uri = qemu+ssh://nova@%s/system?keyfile=/var/lib/nova/.ssh/id_rsa&no_tty=1&no_verify=1 | |
| 11:10:37 | johnthetubaguy | sfinucan: yeah, maybe leave those stranger bits off the example | |
| 11:26:43 | openstackgerrit | Bryan Song proposed openstack/nova master: Migrate VMs failed when anti-affinity group was removed from VMs. https://review.openstack.org/451664 | |
| 11:27:53 | voelzmo | Hey there | |
| 11:28:18 | voelzmo | A while back, there was some blueprint about updateable metadata for servers. Is that still a thing that's being pursued? | |
| 11:31:29 | voelzmo | s/metadata/userdata/ | |
| 11:31:36 | voelzmo | I guess this might be the one https://blueprints.launchpad.net/nova/+spec/userdata-modification | |
| 11:32:11 | openstackgerrit | Bryan Song proposed openstack/nova master: Migrate VMs failed when anti-affinity group was removed from VMs. https://review.openstack.org/451664 | |
| 11:35:43 | openstackgerrit | Stephen Finucane proposed openstack/nova master: conf: Don't deprecate the 'live_migration_uri' https://review.openstack.org/456571 | |
| 11:35:44 | openstackgerrit | Stephen Finucane proposed openstack/nova master: conf: Convert 'live_migration_inbound_addr' to HostAddressOpt https://review.openstack.org/456573 | |
| 11:35:44 | openstackgerrit | Stephen Finucane proposed openstack/nova master: conf: Gather 'live_migration_scheme', 'live_migration_inbound_addr' https://review.openstack.org/456572 | |
| 11:35:57 | sfinucan | johnthetubaguy, pkoniszewski, owalsh: ^ | |
| 11:43:56 | openstackgerrit | Lee Yarwood proposed openstack/nova master: encryptors: Switch to os-brick encryptor classes https://review.openstack.org/391597 | |
| 11:45:35 | openstackgerrit | Bryan Song proposed openstack/nova master: Migrate VMs failed when anti-affinity group was removed from VMs. https://review.openstack.org/451664 | |
| 11:50:45 | EmilienM | hey folks, this patch is really critical for TripleO CI to run correctly again: https://review.openstack.org/#/c/456227/ | |
| 11:51:04 | openstack | Launchpad bug 1682001 in tripleo "[promotion] nova hypervisors/detail fails when host is not mapped to any cell" [Critical,In progress] - Assigned to Oliver Walsh (owalsh) | |
| 11:51:04 | EmilienM | https://bugs.launchpad.net/tripleo/+bug/1682001 | |
| 11:57:33 | openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Use plain routes list for flavors-extraspecs endpoint instead of stevedore https://review.openstack.org/456585 | |
| 12:07:51 | openstackgerrit | NidhiMittalHada proposed openstack/python-novaclient master: Nova volume-attach doesn't care for device name https://review.openstack.org/454048 | |
| 12:09:05 | peter-hamilton | johnthetubaguy: if you have any further questions on the cert validation spec just ping me | |
| 12:09:43 | johnthetubaguy | peter-hamilton: cool, I should hit that one now | |
| 12:14:22 | johnthetubaguy | peter-hamilton: do we want to return a 400 error when a user gives us a list of certs, but our cert checking is turned off? | |
| 12:16:23 | peter-hamilton | johnthetubaguy: good question; does it make sense to throw a warning and store the certs with the instance anyway? | |
| 12:17:00 | johnthetubaguy | peter-hamilton: I am bit worried the user would assume the cloud has done the verification, but it hasn't actually done it | |
| 12:17:04 | peter-hamilton | johnthetubaguy: that way cert validation could be enabled at a later date and boot would still work? | |
| 12:17:42 | johnthetubaguy | although we do allow folks to specify the cert list when the image doesn't require verification anyways | |
| 12:17:52 | johnthetubaguy | its probably more consistent to just log a warning | |
| 12:18:05 | johnthetubaguy | I mean the user doesn't see that I guess | |
| 12:18:08 | peter-hamilton | johnthetubaguy: i could go either way i think | |
| 12:18:16 | johnthetubaguy | peter-hamilton: honestly, me too | |
| 12:18:43 | johnthetubaguy | wonders which side of the fence to sit | |
| 12:18:54 | peter-hamilton | johnthetubaguy: should we hash this out during implementation? | |
| 12:19:35 | johnthetubaguy | peter-hamilton: thats probably the right call | |
| 12:20:17 | peter-hamilton | johnthetubaguy: i'll make a note of it for when i get to the API patches | |
| 12:26:30 | openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Use plain routes list for flavors-access endpoint instead of stevedore https://review.openstack.org/456589 | |
| 12:31:43 | openstackgerrit | Yasufumi Ogawa proposed openstack/nova-specs master: Merge "Libvirt options for DPDK" https://review.openstack.org/456593 | |
| 12:43:59 | dane-fichter | johnthetubaguy: Can you take a look at peter's spec when you get a chance? https://review.openstack.org/#/c/357151/ | |