Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2018-06-06
16:12:17 bauzas mriedem: ok, so that will work, because I wasn't seeing any exception coming from the virt layer handled by the dest host
16:12:26 stephenfin namely, the logical NIC where all tunneled traffic ingresses and egresses from the compute node
16:12:42 bauzas mriedem: the spec is a bit missing of details, but okay
16:12:44 stephenfin logical because it can actually be two or more physical NICs bonded together
16:13:02 efried stephenfin: Ah. And those multiple NICs may live on multiple NUMA nodes.
16:13:07 openstackgerrit Murali Annamneni proposed openstack/nova master: Enables MySQL Cluster Support for Nova https://review.openstack.org/446643
16:13:13 stephenfin you're only allowed one tunnel endpoint in neutron. Every tunneled network must use that endpoint
16:13:18 stephenfin efried: Correct
16:13:34 stephenfin Ditto for physnet-type networks (e.g. layer 2)
16:14:08 stephenfin You could have bonded NICs spread across the various host NUMA nodes
16:14:25 efried k, I get that.
16:14:48 efried Thanks stephenfin.
16:15:10 stephenfin no problemo
16:21:57 jaypipes mriedem: +W
16:22:35 bauzas jaypipes: snap, I litterally +Wd 10 secs after
16:22:39 bauzas but I have comments on it
16:26:11 mriedem bauzas: i replied to your comments about pre live migratoin
16:26:16 mriedem i don't really understand the concern
16:27:19 bauzas mriedem: it's a pure implementation thing
16:27:41 bauzas mriedem: it's just the fact we pass straight to the compute the migrate_data object that we build by the conductor
16:27:43 bauzas or the api
16:27:47 stephenfin efried: You're not on openstack-oslo but https://review.openstack.org/#/c/564822/
16:27:52 mriedem "you somehow need to hook into this and call the virt driver so that you can raise the exception."
16:27:56 mriedem bauzas: ^ is all over rpc call
16:28:05 mriedem so if the virt driver raises an exception, it goes back to the source and fails
16:28:10 bauzas mriedem: but the source host never touches it *before* the dest compute amends it
16:28:26 stephenfin efried: I spotted the same issue on oslo.logging (I think) earlier this week
16:28:30 mriedem bauzas: it doesn't need to
16:28:34 bauzas mriedem: sure, but the spec was saying "we pass the source params to the dest so it can fail there"
16:28:37 mriedem the dest host populates that information and sends it back to the source
16:29:01 mriedem bauzas: where does it say that?
16:29:10 mriedem it just says, implement pre_live_migratoin for the vmware driver
16:29:16 mriedem like any virt driver that supports live migratoin
16:33:22 bauzas mriedem: it just says "implement pre migration checks"
16:33:38 bauzas oh shit
16:33:41 bauzas s/of/to
16:34:13 openstackgerrit Merged openstack/nova-specs master: VMware: add support for live migration https://review.openstack.org/299207
16:34:16 bauzas "which will carry foo, bar, baz *of* the target"
16:34:24 bauzas and not "to the target"
16:35:21 bauzas mriedem: I think I misunderstood the point
16:35:27 bauzas because of s/of/to
16:38:08 openstackgerrit Stephen Finucane proposed openstack/nova master: objects: Add NUMANetworkInfo object https://review.openstack.org/564439
16:38:09 openstackgerrit Stephen Finucane proposed openstack/nova master: conf: Add '[neutron] physnets' and related options https://review.openstack.org/564440
16:38:10 openstackgerrit Stephen Finucane proposed openstack/nova master: libvirt: Start populating NUMACell.network_info field https://review.openstack.org/564441
16:39:24 bauzas mriedem: tbc, the exception that would be raised by the virt driver because of incompatible vcluster datasets would be here https://github.com/openstack/nova/blob/47423dfb7fb22924b6b06b99bb571c49bbc54af8/nova/compute/manager.py#L6097-L6101
16:39:34 bauzas but I'm overengineering
16:41:31 mriedem bauzas: maybe you mean this? https://review.openstack.org/#/c/270116/10/nova/virt/vmwareapi/vmops.py@1623
16:42:18 bauzas mriedem: *facepalms*
16:42:28 bauzas I searched for the implementation
16:42:42 bauzas but since they used a fucking wrong gerrit name, I was fcked
16:42:55 mriedem i think on the dest, the driver just stores what the cluster and data store is, so the source has to find a compatible esxi host in the same cluster and datastore
16:43:10 mriedem if it can't, then it's a failure
16:43:33 bauzas note for myself : read the fscking launchpad whiteboard and don't assume people use gerrit correctly
16:44:13 mriedem johnthetubaguy: you can drop the procedural -2 on this now https://review.openstack.org/#/c/270116/
16:45:16 bauzas mriedem: yup, I just feel https://review.openstack.org/#/c/270116/10/nova/virt/vmwareapi/driver.py@277 needs to return the object they said
16:45:46 bauzas but anyway, let's voice on the change
16:50:34 mriedem bauzas: that's this https://review.openstack.org/#/c/270116/10/nova/virt/vmwareapi/driver.py@304
16:50:42 mriedem it's the same object
16:51:12 mriedem conductor calls check_can_live_migrate_destination to get migrate_data, passes it to live_migration, which passes it to pre_live_migration, which passes it bakc
16:51:13 mriedem *back
16:51:17 mriedem it's like the live migration hot potato
16:52:47 bauzas yeah, it's just yet another way of doing the same check
16:53:38 dansmith mriedem: melwitt: tssurya: assume we're having a cells meeting?
16:53:42 bauzas if we were using the result of check_can_live_migrate_dest() on the compute by calling the source compute too before we run the migration, that would save us one roundtrip
16:56:18 openstackgerrit Mathieu Gagné proposed openstack/nova-specs master: Use generic metadata term instead of config drive https://review.openstack.org/572837
16:58:18 tssurya dansmith: I guess so ?
16:58:54 dansmith let's do it just to say we did
17:15:41 openstackgerrit Merged openstack/nova-specs master: Use generic metadata term instead of config drive https://review.openstack.org/572837
17:19:51 mgagne got 3 nova-specs change merged in 1 day. more in 1 day than 2 whole years? /s =)
17:19:56 dansmith heh
17:20:00 mriedem lyarwood: btw, connection_info has access_mode but the volume attachment records in cinder have attach_mode
17:20:04 mriedem so that's probably where i got it
17:20:20 mriedem mgagne: you are now at your quota limit for nova specs
17:21:12 mgagne haha, for the next 2 years, see you at OpenStack Xylophone
17:21:18 dansmith heh
17:22:21 mriedem openstack xenophobe coming to a denver conference center near you!
17:22:45 mriedem lyarwood: so on https://review.openstack.org/#/c/552078/ i'm inclined to leave it attach_mode to mirror what's in the cinder api
17:23:00 mriedem lyarwood: given that, i'm not sure if there is additional stuff you need for that spec
17:23:23 mriedem i also realize i'm talking to a dev/null here and hopefully you'll come out of your FFU dungeon for awhile to reply
17:25:11 mgagne mriedem: I got Queens up with CellsV1 so there is that. I didn't try to boot an instance but services were not complaining. Now working my way up from Mitaka to Queens.
17:25:44 mriedem nice
17:25:49 mriedem just cell0?
17:25:53 mriedem or 1:1 cell mappings?
17:33:11 mgagne cell0 since it's required and old setup with cells.json
17:35:25 openstackgerrit Surya Seetharaman proposed openstack/nova-specs master: Handling a down cell https://review.openstack.org/557369
17:37:23 tssurya mriedem, melwitt, dansmith: ^^
17:38:17 tssurya jaypipes: whenever you have the time would appreciate your expert opinion on https://review.openstack.org/#/c/560042/
17:38:34 tssurya thanks in advance
17:41:20 dansmith jaypipes: yeah I cc'd you on the above a couple weeks ago looking for your comment from a db perspective, so would appreciate a read from you
17:45:17 openstackgerrit Curt Moore proposed openstack/nova-specs master: Add spec for downloading images via RBD https://review.openstack.org/572805
17:46:04 jaypipes dansmith, tssurya: k, will do a review.
17:55:25 openstackgerrit Jan Gutter proposed openstack/nova-specs master: Spec to implement vRouter HW offloads https://review.openstack.org/567148
17:59:26 openstackgerrit Jeremy Freudberg proposed openstack/nova master: DNM Revert "Use ksa session for cinder microversion check" https://review.openstack.org/572853
18:32:51 openstackgerrit Merged openstack/nova master: Add granular placement policy rules for aggregates https://review.openstack.org/570621
18:56:30 jaypipes dansmith, tssurya: k, reviewed. sorry for the delay on that.
18:56:48 tssurya jaypipes: thanks,!
18:58:40 jaypipes tssurya: np. mriedem's heal_allocations patch is a really good example of doing that batch processing stuff.
18:59:08 tssurya absolutely! I will use similar logic then :)
18:59:45 jaypipes tssurya: cool. just remember the key to that is to ensure you pass a sort_key consistently so you get a stable paging system.
19:00:08 mriedem jaypipes: btw, i've got a +2 on heal_allocatoins now if you want to send it home

Earlier   Later