Earlier  
Posted Nick Remark
#openstack-nova - 2022-08-03
09:41:42 sean-k-mooney[m] 'device_id': '',
09:41:42 sean-k-mooney[m] 'device_owner': '',
09:41:46 sean-k-mooney[m] is acttully detach
09:42:08 amorin ack, we should maybe split that in 2 separate function
09:42:15 amorin where detach would call unbind
09:42:24 sean-k-mooney[m] that or add a detach kwarg
09:42:47 amorin ok, maybe easier
09:43:16 amorin I think I have all info, I'll do some tests, open the bug, and come back
09:43:17 sean-k-mooney[m] did this cause you any issues or did you just notice it and find it ood
09:43:51 amorin it's causing us issues because we are monitoring the ports attached to a compute, and we found some bound ports without instances
09:44:25 amorin and then, digging into that, we were thinking that it's odd to keep the binding :)
09:44:39 sean-k-mooney[m] that should not result in the ports being bound on the compute node
09:44:44 sean-k-mooney[m] sorry not bound
09:44:48 sean-k-mooney[m] but present on ovs
09:44:53 amorin it's not anymore
09:44:58 amorin so i'ts good
09:44:58 sean-k-mooney[m] and or "plugged"
09:45:21 sean-k-mooney[m] ya okj
09:45:30 sean-k-mooney[m] just making sure we were not leaking the neutron ports
09:45:42 sean-k-mooney[m] we acctuly do call unplug_vifs
09:45:46 sean-k-mooney[m] during shelve offload
09:45:54 amorin yes, everything is actually fine on the compute
09:46:01 sean-k-mooney[m] incidentally shelve offload is where it should be unbound
09:46:05 sean-k-mooney[m] not shelve itslef
09:46:11 amorin it's just that the database (ml2_port_binding) does not reflect what is on the compute
09:46:24 amorin yes
09:46:30 sean-k-mooney[m] by defualt we automaticaly shelve offload with a time out of 0 seconds
09:46:44 sean-k-mooney[m] yep that makes sense
09:46:49 amorin true, we kept this parameter to 0 here
09:47:34 sean-k-mooney[m] personally i would prefer to eventurlaly remove the config option and always just offload but that proably wont happen
09:48:00 amorin what's the purpose of delaying the offload?
09:48:55 sean-k-mooney[m] in rare cases if you unshleve quickly enough it can sometimes be useful to delay. the use case was for bustable instances
09:49:14 sean-k-mooney[m] basically in some cases where shelve/unshele is used to scale in/out
09:49:28 sean-k-mooney[m] its nice to have say a 15min delay incase the load spike
09:49:42 sean-k-mooney[m] as its faster to unshleve if the vms is still in place
09:50:08 sean-k-mooney[m] in partice that was only ever useful for vms without ceph or boot form volume
09:50:08 amorin ok, that's sound like a weird usecase to me
09:50:19 amorin shelve / unshelve is not suppose to happen a quick way
09:50:31 sean-k-mooney[m] well that depends
09:50:36 sean-k-mooney[m] shelve should be very very quick
09:50:45 sean-k-mooney[m] if you are using rbd or boot form volume
09:50:55 sean-k-mooney[m] sicne there is no data to copy
09:51:18 sean-k-mooney[m] but ya it was a pretty niche usecase which is why i would prefer to simplify the code
09:51:29 amorin ack
09:51:36 sean-k-mooney[m] that or remove the config option but make it an api parmater
09:51:54 sean-k-mooney[m] so the user can say delay for up to x time
09:52:18 sean-k-mooney[m] i do not like config driven api behavior shich this currently is
10:47:27 opendevreview Merged openstack/nova master: Remove the PowerVM driver https://review.opendev.org/c/openstack/nova/+/850346
11:23:09 sean-k-mooney[m] stephenfin: your lines of code stats will never not be negitive ^
11:23:57 gibi don't encourage him, he will find a way to replace the content of the nova repo with a simple readme file :D
11:23:58 sean-k-mooney[m] cells v1, nova networks xen and power vm.
11:24:06 sean-k-mooney[m] whats next on the list :)
11:24:44 sean-k-mooney[m] oh i for got the docs project :)
11:25:04 sean-k-mooney[m] @gibi: that would be one way to close out all the bugs
11:25:57 gibi what an elegant way :)
11:25:58 sean-k-mooney[m] @gibi speaking of which i marked the custom device_owner bug as invalid
11:26:17 gibi heh, I guessed you would
11:26:26 gibi and I agree
11:28:58 gibi their use case can be solved without the requested change
11:30:14 sean-k-mooney[m] if we wanted to support it we should do it properly as a minor feature with the filtering of events on the neutron side and filtering of port on the nova side so taht they are never stored in our netrowk info cache ectra
11:30:48 opendevreview Merged openstack/nova master: [trivial] Simplify dict get call by removing unused default https://review.opendev.org/c/openstack/nova/+/850450
11:37:40 gibi sean-k-mooney[m]: I agree
12:12:33 gibi during live migration what triggers the creation of the domain on the dest host? Is it the call to migrateToURI3 from the source host? or nova somehow pre-create a domain on the dest host?
12:12:46 sean-k-mooney[m] yes
12:13:06 sean-k-mooney[m] its created by libvirt as a result of our call to libvirt on the source node
12:13:11 gibi thanks
12:13:16 sean-k-mooney[m] via the migrateToURI3 function
12:13:39 gibi so even though our func test env simulate libvirt it does not simulate that 14:12 < gibi> during live migration what triggers the creation of the domain on the dest host? Is it the call to migrateToURI3 from the source host? or nova somehow pre-create a domain on the dest host?
12:13:44 gibi 14:12 < sean-k-mooney[m]> yes
12:13:45 gibi again
12:13:47 gibi sorry
12:13:53 sean-k-mooney[m] correct
12:14:04 sean-k-mooney[m] we just mock out that call
12:14:07 gibi so even though our func test env simulate libvirt it does not simulate that migrateToURI3 triggers the domain creation on the dest
12:14:15 gibi ack
12:14:43 gibi that explains some func test failures I see with https://review.opendev.org/c/openstack/nova/+/851832/3
12:14:47 sean-k-mooney[m] we dont really simulate the domins in general in the libvirt fixture
12:14:58 sean-k-mooney[m] at least not for the move ops like that
12:15:18 sean-k-mooney[m] we proably could mock it differntly adn have it do that to be fair
12:15:38 gibi three is some vm tracking in the fixture but yes, that seems to be incomplete
13:13:54 opendevreview Balazs Gibizer proposed openstack/nova master: Do not mutate migration context for rollback_live_migration_at_destination https://review.opendev.org/c/openstack/nova/+/851832
13:13:54 opendevreview Balazs Gibizer proposed openstack/nova master: Prevent instance.save() under mutated migration context https://review.opendev.org/c/openstack/nova/+/850746
13:22:06 opendevreview Balazs Gibizer proposed openstack/placement master: Clarify trait filtering in the API doc https://review.opendev.org/c/openstack/placement/+/825501
13:25:10 opendevreview Balazs Gibizer proposed openstack/nova stable/ussuri: Reproduce bug 1944759 https://review.opendev.org/c/openstack/nova/+/836733
13:25:11 opendevreview Balazs Gibizer proposed openstack/nova stable/ussuri: Store old_flavor already on source host during resize https://review.opendev.org/c/openstack/nova/+/836734
13:31:13 opendevreview Balazs Gibizer proposed openstack/nova master: Poison /sys access via various calls in test https://review.opendev.org/c/openstack/nova/+/844627
13:31:14 opendevreview Balazs Gibizer proposed openstack/nova master: Add compute restart capability for libvirt func tests https://review.opendev.org/c/openstack/nova/+/850510
13:31:14 opendevreview Balazs Gibizer proposed openstack/nova master: Rename [pci]passthrough_whitelist to device_spec https://review.opendev.org/c/openstack/nova/+/843834
13:31:15 opendevreview Balazs Gibizer proposed openstack/nova master: Rename exception.PciConfigInvalidWhitelist to PciConfigInvalidSpec https://review.opendev.org/c/openstack/nova/+/843861
13:31:15 opendevreview Balazs Gibizer proposed openstack/nova master: Rename whitelist in tests https://review.opendev.org/c/openstack/nova/+/843862
13:31:16 opendevreview Balazs Gibizer proposed openstack/nova master: Basics for PCI Placement reporting https://review.opendev.org/c/openstack/nova/+/846187
13:31:16 opendevreview Balazs Gibizer proposed openstack/nova master: Extend device_spec with resource_class and traits https://review.opendev.org/c/openstack/nova/+/846218
13:31:17 opendevreview Balazs Gibizer proposed openstack/nova master: Reject PCI dependent device config https://review.opendev.org/c/openstack/nova/+/846435
13:31:18 opendevreview Balazs Gibizer proposed openstack/nova master: Reject mixed VF rc and trait config https://review.opendev.org/c/openstack/nova/+/846436
13:31:18 opendevreview Balazs Gibizer proposed openstack/nova master: Ignore PCI devs with physical_network tag https://review.opendev.org/c/openstack/nova/+/846219
13:31:20 opendevreview Balazs Gibizer proposed openstack/nova master: Reject devname based device_spec config https://review.opendev.org/c/openstack/nova/+/846466
13:31:20 opendevreview Balazs Gibizer proposed openstack/nova master: Support [pci]device_spec reconfiguration https://review.opendev.org/c/openstack/nova/+/846470
13:31:22 opendevreview Balazs Gibizer proposed openstack/nova master: Stop if tracking is disable after it was enabled before https://review.opendev.org/c/openstack/nova/+/847009
13:31:22 opendevreview Balazs Gibizer proposed openstack/nova master: Move provider_tree RP creation to PciResourceProvider https://review.opendev.org/c/openstack/nova/+/850546
13:31:24 opendevreview Balazs Gibizer proposed openstack/nova master: Retry /reshape at provider generation conflict https://review.opendev.org/c/openstack/nova/+/851358

Earlier   Later