Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2018-04-17
13:15:04 openstackgerrit Stephen Finucane proposed openstack/nova master: Add CPUWeigher https://review.openstack.org/379525
13:15:40 Shilpa stephenfin: do we need to inform respective projects to modify the same, or what action we need to take here?
13:15:46 gibi mriedem: the versioned notification works based on the values of the Aggregate ovo so if we want to send the update.start notification with the already updated data then we have to update the Aggregate ovo before the start notification. But that would mean the start notification is sent _after_ the actual update so it is not a start notification any more
13:16:12 gibi mriedem: a bit of chicken and egg
13:17:12 stephenfin Shilpa: We probably do, yes. However, why does that affect the Tempest job?
13:17:15 mriedem gibi: well, or add a 'new_metadata' field to the aggregate payload, something like that
13:17:21 Shilpa stephenfin: is because of any refrence we are missing and cauing temst failure for https://review.openstack.org/#/c/550172?
13:17:23 mriedem or 'metadata_updates'
13:18:46 mriedem gibi: granted, the legacy notification is a bit dumb in that it sends the start notification with the updates but it also sends the end notification with the same updates, but not the full aggregate meta, so the legacy notification is also arguably incomplete
13:19:20 mriedem since i don't know how these are consumed, i don't want to rathole on it too much and that's why i didn't -1
13:19:43 Shilpa stephenfin: i am still finding root cause of tempst failure, and it was failed for for patch set 1 too, but missing logs there since got over written because of patch set 2
13:19:44 mriedem if the consumer wants/needs to know the delta, i guess they can diff the metadata from the start/end notifications
13:20:13 stephenfin Shilpa: That configuration value definitely works if you test manually?
13:20:47 stephenfin Shilpa: Hmm, looks like we're not passing the token correctly http://logs.openstack.org/72/550172/2/check/tempest-full/f6945b6/controller/logs/screen-n-novnc.txt.gz
13:22:12 gibi mriedem: I would go for the solution where any update.start notification sends information about the state _before_ the update and the update.end sends the new state in the payload. In some cases we deviated from that and added the diff directly to the payload instance.update old_task_sate new_task_state but there we don't have start and end notification to use
13:22:17 Shilpa stephenfin: i have checked these logs and here seems token value is empty
13:24:23 Shilpa stephenfin: from where we can confirm, vnc 1.0.0 get downloaded successfully? i checked http://logs.openstack.org/72/550172/2/check/tempest-full/f6945b6/job-output.json.gz, but how once can confirm downloading of latest VNC version is successful?
13:24:55 mriedem gibi: yeah, +W
13:25:21 gibi mriedem: I think the proposed versioned aggregate.update_metadata is a good start. If somebody needs the explicity diff in the payload we can add that later
13:25:24 gibi mriedem: thanks
13:25:59 gibi mriedem: btw, I will have to go offline for a while this afternoon but I will be back for the notification subteam meeting
13:26:02 openstackgerrit Matthew Booth proposed openstack/nova master: Rename recreate to evacuate in driver signatures https://review.openstack.org/560900
13:26:15 stephenfin Shilpa: Not a clue, I'm afriad
13:32:52 sahid mriedem: still not convinced about https://review.openstack.org/#/c/553072/ ?
13:33:54 Shilpa stephenfin: one more findings, http://logs.openstack.org/72/550172/2/check/tempest-full/f6945b6/controller/logs/screen-n-cond-cell1.txt.gz, here novncproxy_base_url is not reflected what is expected, still referring vvnc_auto.html
13:34:09 mriedem sahid: just forgot about it
13:34:42 sahid mriedem: oh ok no worries
13:37:29 Shilpa stephenfin: i have seen workflow -1, what fix in nova you are expecting here?
13:38:30 stephenfin Shilpa: I noticed all the existing examples used 'path=websockify/' (i.e. added the '/')
13:38:47 stephenfin Just curious to see if that affects things. Perhaps the urlparse module needs it
13:38:55 stephenfin I'm just poking stuff
13:41:14 Shilpa stephenfin: ok, will wiat for Zuul , thank you.
13:44:48 bauzas mriedem: any help would be appreciated for https://review.openstack.org/#/c/561923/
13:44:57 bauzas mriedem: looks like it's blocking some internal CI
13:50:07 mriedem bauzas: want to remove the note from the commit message? https://review.openstack.org/#/c/561925
13:50:41 dansmith sahid: should these be vhost-user in the config help string as well? https://review.openstack.org/#/c/484997/19/nova/conf/libvirt.py
13:51:08 mriedem sahid: left comments in https://review.openstack.org/#/c/553072/
13:52:17 dansmith sahid: or is driver vhost-user and type is vhostuser?
13:53:03 bauzas mriedem: thanks
13:55:28 sahid dansmith: right, in the config doc i'm talking of the backend which is vhost-user. the 'vhostuser' name is for the libvirt interface type only
13:55:42 sahid i will update that point
13:55:46 dansmith sahid: okay, that's terribly confusing, but not your fault
13:55:56 dansmith sahid: also, why did you abandon the bottom patch?
13:56:45 dansmith sahid: I was just saying he noted it should be vhost-user not vhostuser
13:56:51 dansmith but he never commented on it I guess
13:56:52 openstackgerrit Matthew Booth proposed openstack/nova master: Rename recreate to evacuate in driver signatures https://review.openstack.org/560900
13:56:52 sahid because i understood that we want only change the driver name if rx or tx queue size is updated, if not so we keep the old behavior and touch nothing
13:56:53 openstackgerrit Matthew Booth proposed openstack/nova master: Remove redundant _do_check_can_live_migrate_destination https://review.openstack.org/561948
13:56:54 openstackgerrit Stephen Finucane proposed openstack/nova master: doc: Don't confuse CPU pinning/NUMA as Hyper-V only https://review.openstack.org/561949
13:56:55 openstackgerrit Stephen Finucane proposed openstack/nova master: doc: Remove 'admin/numa' doc https://review.openstack.org/561950
13:57:32 sahid hum so i missunderstood
13:58:12 dansmith sahid: okay I think that setting vhost-user when we're sure it's vhost-user is a good cleanup, unrelated to the vhost driver change at the top, which is what I was referring to
13:58:25 dansmith but it sounds like he's come around on that second bit anyway
13:59:23 dansmith sahid: if you don't mind restoring that bottom patch but s/vhostuser/vhost-user/ I'll be +2 on both
13:59:33 dansmith sahid: and then I'll try to get jaypipes to review it in trade for something else :)
13:59:54 sahid dansmith: yep actually i agreed with you about to fix that point
14:00:01 dansmith *gasp* :)
14:00:20 sahid just not sure i understand what you want with the vhost thing? do you want me to change something?
14:01:24 dansmith the vhost thing is in the second patch, in that you're overriding driver to vhost in more cases today than before the patch, which sean-k-mooney originally said he didn't want to change, but I think he is okay with it now, based on comments, so nothing to change
14:02:08 sahid ok cool but that is said I only update it if rx or tx queue size is asked
14:02:18 sahid anyway let me fix all of that and i will ping you
14:02:28 sahid thanks for your help
14:02:34 dansmith sahid: ah, you're right, that must be why :)
14:03:11 sahid :)
14:03:51 sahid mriedem: ok thanks i will take a look and see whether i can help sean if he does not have time to update the code
14:04:26 bhagyashris efried: Hi,
14:04:28 openstackgerrit Matt Riedemann proposed openstack/python-novaclient master: Microversion 2.61 - support extra_specs in flavor API https://review.openstack.org/561520
14:04:32 efried bhagyashris: Howdy!
14:07:02 mriedem bauzas: trade for you https://review.openstack.org/#/q/topic:bug/1746483+branch:stable/ocata
14:07:19 bauzas mriedem: roger, ack this
14:07:39 bauzas mriedem: you freed me from some internal blackhole, I can pay you off
14:08:39 bauzas oooooh I remember that nasty bug with IsolatedHosts :)
14:09:05 openstackgerrit Matt Riedemann proposed openstack/nova stable/queens: Document how to disable notifications https://review.openstack.org/561954
14:09:36 sean-k-mooney dansmith: ya i would prefer not to set teh driver name element if we dont need too however im fine with it for now
14:10:02 dansmith sean-k-mooney: yep, thanks for that review
14:17:43 openstackgerrit Sylvain Bauza proposed openstack/nova master: doc: BFV instances and IsolatedHostsFilter https://review.openstack.org/543494
14:18:07 bauzas mriedem: FWIW, just made the amendment you asked me 2 months ago https://review.openstack.org/#/c/543494/
14:20:51 openstackgerrit Sylvain Bauza proposed openstack/nova stable/pike: ironic: Get correct inventory for deployed node https://review.openstack.org/561925
14:21:33 mriedem ok
14:22:06 mriedem bauzas: also, gerrit UI won't add the cherry-picked from line in the commit message unless the change you're cherry picking *from* is merged
14:22:39 bauzas I see, in case of a merge commit
14:25:10 openstackgerrit Matt Riedemann proposed openstack/nova master: doc: BFV instances and IsolatedHostsFilter https://review.openstack.org/543494
14:29:48 openstackgerrit Matt Riedemann proposed openstack/nova stable/pike: libvirt: Block swap volume attempts with encrypted volumes prior to Queens https://review.openstack.org/543569
14:39:26 openstackgerrit Merged openstack/python-novaclient master: Stop printing flavor details on successful flavor-delete https://review.openstack.org/561634
14:43:57 kashyap Folks, when someone gets time, can you beat some sense into my (broken) test and me: http://paste.openstack.org/show/719390/
14:44:27 kashyap This is the last of the two unit test failures I need to fix for the "first batch" of the version bump clean-ups
14:44:36 efried kashyap: Which line is the exception hitting on?
14:45:09 kashyap efried: Line-56, that's what I'm messing up
14:45:21 kashyap efried: For context, this is the original: https://github.com/openstack/nova/blob/master/nova/tests/unit/virt/libvirt/test_driver.py#L9171,L9219
14:46:03 kashyap I am replacing the usage of the migrateToURI2() with migrateToURI3() in that test, as a result of post version bump clean-up.
14:46:29 efried kashyap: Uh, the exception hits when you say
14:46:29 efried mock_updated_guest_xml.return_value = ''
14:46:29 efried ?
14:46:40 kashyap In the original: target_xml is supplied to mock object to validate return value - test_mock.XMLDesc.return_value = target_xml
14:47:30 kashyap efried: Can't pin-point to that, but I know that's a mistake there. Still haven't ran PDB on that to narrow down, but was duking around other bits
14:48:21 kashyap efried: For context, this is the other local change that I've fixed for a different test: https://kashyapc.fedorapeople.org/0001-tests-unit-Fix-test_live_migration_update_graphics_x.patch
14:48:39 kashyap It's (the current broken test: 'test_live_migration_with_valid_target_connect_addr') in the same vein
14:48:53 efried kashyap: Is there code change behind this, or are you just tweaking the test case?
14:48:56 kashyap We're not doing any Mox stuff in this current test, though.
14:49:22 kashyap efried: I'm tweaking it as a result of: https://review.openstack.org/#/q/topic:bump_min_libvirt_and_qemu+(status:open+OR+status:merged)

Earlier   Later