Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2018-01-16
17:32:49 mdbooth But we *will* have the secret
17:32:55 stephenfin Just those two so and I'm good
17:33:07 efried stephenfin We can sell you a POWER9 system to test on.
17:33:24 lyarwood mdbooth: I think nova/virt/libvirt/migration.py would take care of that with the XML update
17:33:37 mdbooth Looking
17:33:38 lyarwood mdbooth: we get_volume_config on the dest and then update the volume XML with anything new
17:34:12 lyarwood mdbooth: I had to look into that for the native to native case
17:34:27 lyarwood mdbooth: didn't think of the encryptor to native case but that *should* work
17:34:33 mdbooth Reading... thanks.
17:38:03 openstackgerrit Merged openstack/nova master: conf: Do not inherit image signature props with snapshots https://review.openstack.org/527046
17:39:26 BKaminski Time for my third shift self to head to sleep. If anyone who is familiar with the XenAPI drivers get a chance, would love if you could take a look at my small commit :) https://review.openstack.org/#/c/533168/ Have a great day evreyone.
17:40:28 mdbooth lyarwood: Time for another sanity check: I think that might not work, because get_updated_guest_xml() runs on the source, not the dest.
17:40:57 lyarwood mdbooth: odd, that wouldn't make sense
17:41:11 lyarwood mdbooth: well, to me at least :)
17:41:24 lyarwood mdbooth: I assumed it wanted the updated volume XML from the dest POV
17:41:24 mdbooth It's used to generate the destination xml to pass to the destination in initiating the migration
17:41:57 mdbooth It runs from _live_migration_operation, which I'm pretty sure is on the source
17:42:44 mdbooth But we do call _connect_volume() on the destination
17:42:58 mdbooth in pre_live_migration
17:44:14 lyarwood right but the volume secret lookup would return the src hosts secret UUID
17:44:17 mdbooth And use_native_luks will be true on the destination, so we'll create the host secret
17:44:45 lyarwood that's pretty useless, updating instance XML using the src host
17:44:57 mdbooth No, it's the other way round
17:45:13 mdbooth When we call _attach_encryptor during pre_live_migration on the destination
17:45:27 mdbooth We will create the host secret on the destination
17:46:15 mdbooth Also... we won't actually attach the encryptor
17:46:25 lyarwood right, but _get_volume_config is being called on the src, that does the secret lookup
17:46:39 mdbooth No it doesn't, because the source hasn't been upgraded
17:47:11 lyarwood ah right, in that old to native use case, you're correct
17:47:12 mdbooth I think we're uncovering a design defect in live migration here, tbh
17:47:22 mdbooth But it is what it is
17:47:50 openstackgerrit Eric Fried proposed openstack/nova master: SchedulerReportClient.set_traits_for_provider https://review.openstack.org/532564
17:47:50 openstackgerrit Eric Fried proposed openstack/nova master: ProviderTree.snapshot => ProviderSnapshot https://review.openstack.org/533244
17:47:51 openstackgerrit Eric Fried proposed openstack/nova master: Raise on API errors getting aggregates/traits https://review.openstack.org/526540
17:47:51 openstackgerrit Eric Fried proposed openstack/nova master: Track associated sharing RPs in report client https://review.openstack.org/526539
17:47:52 openstackgerrit Eric Fried proposed openstack/nova master: Track tree-associated providers in report client https://review.openstack.org/526541
17:47:52 openstackgerrit Eric Fried proposed openstack/nova master: ProviderTree.populate_from_iterable https://review.openstack.org/520756
17:47:53 openstackgerrit Eric Fried proposed openstack/nova master: SchedulerReportClient.get_provider_tree https://review.openstack.org/521098
17:47:53 openstackgerrit Eric Fried proposed openstack/nova master: SchedulerReportClient.set_aggregates_for_provider https://review.openstack.org/533808
17:47:54 openstackgerrit Eric Fried proposed openstack/nova master: WIP: SchedulerReportClient.update_from_provider_tree https://review.openstack.org/533821
17:47:54 openstackgerrit Eric Fried proposed openstack/nova master: ComputeDriver.update_provider_tree() https://review.openstack.org/521187
17:47:55 openstackgerrit Eric Fried proposed openstack/nova master: Fix nits in update_provider_tree series https://review.openstack.org/531260
17:47:55 openstackgerrit Eric Fried proposed openstack/nova master: WIP: Use update_provider_tree from resource tracker https://review.openstack.org/520246
17:47:59 efried mgoddard This whole thing has been rearranged and rebased ^
17:48:20 efried mgoddard I noticed you put some stuff on top (or in the middle, or something) - it'll need to be rebased accordingly.
17:48:24 mdbooth efried: Zuul can take another one for the team ;)
17:48:49 mdbooth lyarwood: So I'm not 100% sure what the state would be on the destination
17:48:57 efried mdbooth The way zuul has been treating me lately, I don't have too much sympathy.
17:49:03 mdbooth efried: Hehe
17:49:08 efried Course, zuul is in the position of power, there.
17:49:38 efried jaypipes See above: snapshot comes first now, followed by set_traits_for_provider so mgoddard can use it for the ironic traits bp.
17:49:59 mdbooth lyarwood: Given that this would be a hopefully ephemeral situation, I wonder if we can detect this in pre_live_migration and fail early.
17:50:47 mdbooth lyarwood: i.e. the fix is to not allow live migration host->native encryption
17:51:00 lyarwood mdbooth: well, this breaks native to native as well right?
17:51:09 mdbooth It goes away when they finish the upgrade
17:51:14 mdbooth lyarwood: Does it?
17:51:14 lyarwood mdbooth: if we can't lookup the UUID of the new secret on the dest and udpate the XML we can't migrate
17:51:38 lyarwood mdbooth: we'd just end up pointing to the src secret UUID that doesn't exist on the dest
17:51:45 mdbooth lyarwood: I thought we were using the volume id as the secret uuid, no?
17:52:16 lyarwood mdbooth: no, the lookup code uses the volume_id, the actual secret has a seperate UUID you need to provide in the XML
17:52:24 efried oh gdi. Stupid pep8 mistakes. And on the bottom patch, of course.
17:53:00 mdbooth efried: LOL. FWIW: 'git rebase -i master -x "tox -e pep8 -- -HEAD"' is your friend
17:53:23 efried ooo
17:53:29 mdbooth Will run pep8 against each of your patches individually
17:53:32 jaypipes efried: acl
17:53:34 efried That's slightly easier than the way I was gonna do it.
17:53:36 jaypipes ack
17:55:40 mdbooth lyarwood: Well, without checking the libvirt api specifically, sounds like that's broken :(
17:55:53 mdbooth This must have come up in a different context.
17:56:42 lyarwood mdbooth: I wonder if we can create a secret with the same UUID on the dest?
17:57:30 esberglu stephenfin: Responded to your OVS questions
17:57:36 mdbooth lyarwood: This works currently, right?
17:57:56 mdbooth i.e. We can't just not support it, because that would be a regression.
17:58:12 lyarwood mdbooth: LM with encrypted volumes? Yeah AFAIK it does.
17:58:25 mdbooth lyarwood: You'd need a config variable to switch it off ;) You should add one of those.
17:58:35 lyarwood \o/
17:58:55 lyarwood mdbooth: brb
17:59:02 mdbooth I have to run, but I don't believe this is beyond the wit of man.
17:59:16 mdbooth lyarwood: Catch up tomorrow afternoon?
17:59:26 openstackgerrit Mark Goddard proposed openstack/nova master: Add get_traits() method to ComputeDriver https://review.openstack.org/532287
17:59:26 openstackgerrit Mark Goddard proposed openstack/nova master: WIP: Send traits to ironic on server boot https://review.openstack.org/508116
17:59:27 openstackgerrit Mark Goddard proposed openstack/nova master: Implement get_traits() for the ironic virt driver https://review.openstack.org/532288
17:59:27 openstackgerrit Mark Goddard proposed openstack/nova master: Call get_traits() in the resource tracker https://review.openstack.org/532290
17:59:40 openstackgerrit Eric Fried proposed openstack/nova master: SchedulerReportClient.set_traits_for_provider https://review.openstack.org/532564
17:59:40 openstackgerrit Eric Fried proposed openstack/nova master: ProviderTree.snapshot => ProviderSnapshot https://review.openstack.org/533244
17:59:41 openstackgerrit Eric Fried proposed openstack/nova master: Raise on API errors getting aggregates/traits https://review.openstack.org/526540
17:59:41 openstackgerrit Eric Fried proposed openstack/nova master: Track associated sharing RPs in report client https://review.openstack.org/526539
17:59:42 openstackgerrit Eric Fried proposed openstack/nova master: Track tree-associated providers in report client https://review.openstack.org/526541
17:59:42 openstackgerrit Eric Fried proposed openstack/nova master: ProviderTree.populate_from_iterable https://review.openstack.org/520756
17:59:43 openstackgerrit Eric Fried proposed openstack/nova master: SchedulerReportClient.get_provider_tree https://review.openstack.org/521098
17:59:43 openstackgerrit Eric Fried proposed openstack/nova master: SchedulerReportClient.set_aggregates_for_provider https://review.openstack.org/533808
17:59:44 openstackgerrit Eric Fried proposed openstack/nova master: WIP: SchedulerReportClient.update_from_provider_tree https://review.openstack.org/533821
17:59:44 openstackgerrit Eric Fried proposed openstack/nova master: ComputeDriver.update_provider_tree() https://review.openstack.org/521187
17:59:45 openstackgerrit Eric Fried proposed openstack/nova master: Fix nits in update_provider_tree series https://review.openstack.org/531260
17:59:45 openstackgerrit Eric Fried proposed openstack/nova master: WIP: Use update_provider_tree from resource tracker https://review.openstack.org/520246
17:59:48 efried There, ffs. mgoddard Sorry dude
18:00:39 lyarwood mdbooth: sure, thanks again
18:00:47 melwitt jaypipes: scheduler question for you ... when the filter scheduler does the compute node prune based on placement before going through the scheduler filters, does it take into consideration overcommit values set by aggregate?
18:01:28 jaypipes melwitt: no
18:01:53 jaypipes melwitt: placement has no concept of "aggregate overcommit"

Earlier   Later