Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2020-09-18
14:47:26 mnaser aka if shared storage: skip ssh
14:54:24 gibi mnaser: I might be wrong but I remember nova detects that the instance is on share path by checking that the same file exists on both the source and the dest
14:54:44 mnaser gibi: yeah i think that might be the only reason it still does it
14:55:04 mnaser i think breaking out early if the VM is using volumes only would make sense cause that would eliminate the whole ssh key thing for us
14:55:09 mnaser the only reason we still have it is just for that
14:55:14 gmann gibi: bauzas lyarwood what is consensus for Focal migration? same as we discussed in meeting?
14:55:24 sean-k-mooney mnaser: we could do that via an rpc i guess
14:55:56 mnaser sean-k-mooney: can one n-cpu rpc to another n-cpui
14:56:03 sean-k-mooney yes
14:56:12 sean-k-mooney it happens alot in move operations
14:56:26 sean-k-mooney well not a lot but we do rpcs between teh computes invovled
14:57:12 gibi mnaser: hm, only volume instances still can have config drives
14:57:29 mnaser gibi: right but i think we dont actually move that we just rebuild it on the target with a cold migrate
14:57:30 mnaser i _think_
14:57:59 gibi most probably yes
14:58:33 gibi so, yeah we can stop doing that ssh shared storage check if the instance is only using volumes
14:58:33 mnaser https://github.com/openstack/nova/blob/cff7382fb3de4eef7aeddc34b6ae3409cbe50ea0/nova/virt/libvirt/driver.py#L8340-L8376
14:58:39 mnaser hmm this doesnt seem to use ssh
15:00:39 mnaser i cant seem to honestly find the code that does the ssh to remote host..
15:00:41 gibi then i don't know what step needs the ssh
15:01:14 mnaser i should go look for a traceback
15:01:23 gibi gmann: we merged the libvirt min version bump, and we go with what we discussed on the meeting yesterday as far as I understand
15:02:04 gmann gibi: ok, thanks for confirmation I will check and reply the ML after my breakfast
15:02:26 openstack bugzilla.redhat.com bug 1624709 in openstack-packstack "[OSP 13] Need to change the permission of /etc/nova/migration/identity after installation" [High,Closed: errata] - Assigned to ratailor
15:02:26 mnaser ok i dont have the actual traceback but it looks like we try to mkdir remotely? https://bugzilla.redhat.com/show_bug.cgi?id=1624709
15:04:00 openstackgerrit Elod Illes proposed openstack/nova stable/queens: libvirt: Provide VIR_MIGRATE_PARAM_PERSIST_XML during live migration https://review.opendev.org/747976
15:04:13 mnaser maybe a cold migrate does move the configdrive after all
15:06:31 sean-k-mooney shelve recreates it as does cross cell resize
15:06:48 sean-k-mooney but for same cell resize/migrate i think we copy them
15:08:35 gibi mnaser: I only see the RemoteFilesystem.create_dir used for non shared instances
15:08:38 gibi driver.py#L690
15:08:40 gibi https://github.com/openstack/nova/blob/90777d790d7c268f50851ac3e5b4e02617f5ae1c/nova/virt/libvirt/driver.py#L10347-L10349
15:10:48 gibi but during livemigrate we copy the config drive https://github.com/openstack/nova/blob/90777d790d7c268f50851ac3e5b4e02617f5ae1c/nova/virt/libvirt/driver.py#L9731
15:11:00 openstackgerrit Ghanshyam Mann proposed openstack/nova master: DNM Testin Focal with integrated-gate-compute on Bionic https://review.opendev.org/752711
15:51:45 openstackgerrit Balazs Gibizer proposed openstack/nova master: Clean up the DynamicVendorData constructor https://review.opendev.org/752718
15:51:45 openstackgerrit Balazs Gibizer proposed openstack/nova master: Use cell targeted context to query instance metadata https://review.opendev.org/752459
15:53:19 gibi that was all from me for this week. See you on Monday
15:53:21 gibi o/
16:12:06 openstackgerrit Claudiu Belu proposed openstack/nova master: hyperv: Configures chassis asset tags for VMs. https://review.opendev.org/752723
16:23:47 sean-k-mooney stephenfin: look like this issue im facing is that the fakelibvirt dirver is incomplete
16:24:45 sean-k-mooney stephenfin: when we generate teh fake instance domain its not provided with the info related to the vm pci devices just that the nic is type direct or whatever.
16:25:06 stephenfin Hmm, yeah, that's not surprising
16:25:25 sean-k-mooney we are also not generating network node_dev elements for any of the pci PF/VFs
16:25:51 sean-k-mooney im going to hard code a few thigs to see if i can make some minimal chagnes to get it to pass
16:26:03 sean-k-mooney but we likely need to flesh this out more in the future
16:26:41 stephenfin yeah, I did the minimal amount necessary to get the sr-iov migration tests working
16:27:02 stephenfin but it's full of holes, for sure. I just didn't have SR-IOV hardware handy to inspect the flow and copy the XML from
16:27:32 sean-k-mooney so the issue is i need to backprot this to queens
16:27:43 sean-k-mooney so im hesitent to depend on your refactors
16:27:53 sean-k-mooney but they might also fix some of the issues
16:28:37 sean-k-mooney this is what causes the detach to fail to find the device currently
16:28:40 sean-k-mooney https://github.com/openstack/nova/blob/master/nova/tests/unit/virt/libvirt/fakelibvirt.py#L1826-L1830
16:28:52 sean-k-mooney that ses the souce_dev to fake_pf_interface_name
16:29:08 sean-k-mooney but in https://review.opendev.org/#/c/602432/17/nova/tests/unit/virt/libvirt/fakelibvirt.py
16:29:30 sean-k-mooney both before and after my change we dont set source_dev when generatign the vm domain
16:30:10 sean-k-mooney im going to hard code if for now but im concerned then next thing ill hit is the pci adress dont match adresses form the host pci_info object
16:30:24 sean-k-mooney since in both cases we are just incrementing the adresses
16:30:46 sean-k-mooney ill corss that bridge later i guess.
16:34:12 sean-k-mooney the issue with that fixture is that the comment and assumtion is wrong
16:34:26 sean-k-mooney we also use this with macvtal to get the vf name not jsut the parent
16:34:57 sean-k-mooney and we also dont have the required info to work this out when we generate the domain stuff currently.
17:03:16 openstackgerrit Merged openstack/nova master: test_evacuate: Wait until subnode is down before starting tests https://review.opendev.org/751065
17:14:07 sean-k-mooney stephenfin: ... https://github.com/openstack/nova/blob/master/nova/tests/unit/virt/libvirt/fakelibvirt.py#L1087-L1093
17:14:27 sean-k-mooney detach device in the fake libvirt fixture only support disks
17:14:57 sean-k-mooney so i dont know how/why gibis func test could be passing on master.
19:19:51 lyarwood gmann: sorry was afk, are you still around?
19:20:31 lyarwood gmann: AFAIK the plan is to still move to focal in V and skip the tests
19:20:51 lyarwood gmann: in the meantime we wanted to try to move bionic onto the train UCA to allow us to move the MIN_ versions in the libvirt driver
19:21:10 lyarwood gmann: https://review.opendev.org/#/c/752641/ needs to land first in stable/ussuri and then master to allow grenade to work correct
19:21:14 lyarwood correctly*
19:21:24 lyarwood then the rest of https://review.opendev.org/#/q/topic:bump-libvirt-qemu-victoria+(status:open+OR+status:merged) can land
19:22:16 lyarwood gmann: when do you think th focal change is actually going to land btw?
19:56:57 gmann lyarwood: hi, sorry i was away
19:59:05 gmann lyarwood: ok, and I am keeping tempest-integrated-compute job on bionic to have those test coverage running somewhere
19:59:34 gmann lyarwood: let me review the devstack patches
20:00:12 gmann plan for Focal migration is pending on Barbican issue which is the only blocker left https://storyboard.openstack.org/#!/story/2007732
20:09:34 openstackgerrit Ghanshyam Mann proposed openstack/nova master: DNM: testing Train UCA https://review.opendev.org/752759
21:40:12 openstackgerrit Merged openstack/nova stable/ussuri: Correctly disable greendns https://review.opendev.org/751621
23:03:02 openstackgerrit Merged openstack/nova master: test_evacuate.sh: Support libvirt-bin and libvirtd systemd services https://review.opendev.org/752650
23:03:12 openstackgerrit Merged openstack/nova stable/rocky: Improve CinderFixtureNewAttachFlow https://review.opendev.org/749222
#openstack-nova - 2020-09-19
01:38:16 gmann lyarwood: +2 on both devstack patches,
01:42:27 openstackgerrit Merged openstack/nova stable/queens: libvirt: Provide VIR_MIGRATE_PARAM_PERSIST_XML during live migration https://review.opendev.org/747976
14:20:08 openstackgerrit Merged openstack/nova stable/rocky: Robustify attachment tracking in CinderFixtureNewAttachFlow https://review.opendev.org/749223
16:58:08 openstackgerrit Merged openstack/nova stable/rocky: compute: Use source_bdms to reset attachment_ids during LM rollback https://review.opendev.org/749224
#openstack-nova - 2020-09-20
08:42:44 openstackgerrit Mamduh proposed openstack/os-vif master: Fix - os-vif fails to get the correct UpLink Representor https://review.opendev.org/745895
11:50:36 aj_mailing join #openstack
21:59:03 openstackgerrit Merged openstack/nova stable/rocky: compute: refactor volume bdm rollback error handling https://review.opendev.org/749225
21:59:11 openstack bug 1889108 in OpenStack Compute (nova) rocky "failures during driver.pre_live_migration remove source attachments during rollback" [Undecided,In progress] https://launchpad.net/bugs/1889108 - Assigned to Lee Yarwood (lyarwood)
21:59:11 openstackgerrit Merged openstack/nova stable/rocky: Add regression tests for bug #1889108 https://review.opendev.org/749226
21:59:14 openstackgerrit Merged openstack/nova stable/rocky: compute: Don't delete the original attachment during pre LM rollback https://review.opendev.org/749227
#openstack-nova - 2020-09-21
01:10:54 openstackgerrit Yao wang proposed openstack/nova master: Use VIR_DOMAIN_XML_INACTIVE for detach volume https://review.opendev.org/749489
01:40:33 openstackgerrit Yingji Sun proposed openstack/nova stable/train: Set different VirtualDevice.key https://review.opendev.org/752843
02:37:18 openstackgerrit Yao wang proposed openstack/nova master: Use VIR_DOMAIN_XML_INACTIVE for detach volume https://review.opendev.org/749489
08:01:16 bauzas good morning Nova
08:10:47 gibi good morning bauzas
08:23:12 openstackgerrit Elod Illes proposed openstack/nova stable/pike: libvirt: Provide VIR_MIGRATE_PARAM_PERSIST_XML during live migration https://review.opendev.org/747978
08:27:39 kashyap bauzas: When is the deadline? 01-Oct?
08:27:47 bauzas yes...
08:35:59 brinzhang gibi: can you review this patch https://review.opendev.org/#/c/749052?
09:09:03 kashyap lyarwood: Morning, I've tried a rapid succession of attaching 20 disks (via `attach-device`) while there's consistent 40% of load on the CPUs ... can't reproduce so far.

Earlier   Later