Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2018-01-02
15:54:43 mdbooth mriedem: That is an excellent point. That should at least be tested.
15:54:43 mriedem also, another reason we don't have constraints on these things at first
15:54:49 mdbooth And I wouldn't be surprised to find it's broken.
15:54:51 mriedem mozart89: no
15:55:01 kashyap Oops, the 'i' in "Performance impact" _have_ to be capital; or else pep8 fails. /me corrected it out of OCD, and realized the template itself was like htat
15:55:43 mdbooth There's no constraint on it right now. The goal is to put a constraint on it in the first release where we positively assert that this migration has already run.
15:56:01 mdbooth i.e. if it fails, it's the operator's fault.
15:56:41 mriedem i was thinking of the unique constraint https://review.openstack.org/#/c/242602/28/nova/db/sqlalchemy/migrate_repo/versions/374_bdm_uuid.py
15:56:45 mriedem but i see uuid is nullable=True
15:56:51 mriedem which is the constraint you're talking about
15:57:21 mriedem the BlockDeviceMapping.uuid is not nullable though
15:57:24 mriedem which is probably good enough
15:58:26 mdbooth mriedem: Ah, actually this is still ok because it never loads the object. This is purely a db operation. I'll still add a test for it, if for no other reason than to make it obvious it's intentional.
15:59:32 mozart89 mriedem: I'm not using any NUMA Topology ( CPU pinning and e.t.c). But when i'm triggering "live migration" it stops with message "Require both a host and instance NUMA topology to fit instance on host".
16:00:24 mriedem mozart89: that's just a debug message isn't it?
16:00:55 mriedem mozart89: that's just a debug message that is output when the NUMATopologyFilter runs
16:01:03 mriedem and is likely not your issue
16:04:51 kashyap jaypipes: Is it true---to fix this (http://paste.openstack.org/show/636270/)---I have to explicitly change the order pf py27/py34 in tox.ini?
16:05:13 mriedem kashyap: no
16:05:20 mriedem rm -rf .testrepository
16:05:31 kashyap Yep, just did that.
16:05:47 stephenfin mriedem, kashyap: That isn't necessary on master anymore, I thought
16:05:53 jaypipes why is there a py27/py35 env for a specs repo?
16:05:54 stephenfin Since the switch to stestr
16:06:22 mozart89 mriedem: ok, thank you for direction :)
16:06:25 stephenfin Oh, this is nova-specs. nvm me
16:06:44 kashyap stephenfin: Yeah, it seems to be PEBKAC
16:06:46 mriedem there is no py27/py35 targets for nova-specs
16:07:21 mriedem Require both a host and instance NUMA topology
16:07:23 mriedem oops
16:17:51 kashyap mriedem: When you get a sec, is this also my env prob - http://paste.openstack.org/show/636286/
16:18:30 stephenfin kashyap: Stick 'basepython=python2' in the requisite tox target
16:18:47 stephenfin I'd say your system Python is python3.x
16:18:53 mriedem yeah ^
16:18:55 kashyap stephenfin: I wanted to do that, but I felt: "wait a moment, it should work with existin master"
16:18:58 kashyap Yeah, it is
16:19:15 stephenfin The gate/Ubuntu uses Python 2 so we'd never see it there
16:20:11 stephenfin So it needs basepython or the requisite changes to be Python3 compatible. Latter is better long term, naturally (one year closer to 2020)
16:21:33 openstackgerrit Stephen Finucane proposed openstack/nova master: Address nits in I46d483f9de6776db1b025f925890624e5e682ada https://review.openstack.org/530462
16:22:05 stephenfin gibi_away, jaypipes: So that ^ should _definitely_ work now, if you fancy sticking it on your review queue (it's cleanup for the PCI policy thing)
16:23:41 kashyap stephenfin: That's it; thanks.
16:24:13 mriedem kashyap: push a patch so others don't hit it
16:24:18 kashyap `tox -epep8 -- -HEAD` succeeds; ( mriedem - `tox -e fast8` was a joke, right
16:24:22 kashyap mriedem: Okido
16:24:28 mriedem kashyap: tox -e fast8 is a nova repo thing
16:24:30 mriedem not nova-specs
16:24:39 kashyap Ah, okay.
16:25:52 kashyap mriedem: Oops on, actually "that's not it". Fast switching between repos, I ran it on Nova upstream
16:26:40 kashyap stephenfin: That's my tox.ini - http://paste.openstack.org/show/636299/
16:26:48 kashyap That should more than do, right?
16:26:53 kashyap (Again, this is for nova-specs)
16:30:38 stephenfin kashyap: Should do, yup
16:30:42 stephenfin You can drop line 21
16:30:47 jaypipes stephenfin: done
16:31:08 stephenfin kashyap: Also line 3-5, but that's another change
16:31:09 mriedem ildikov: are you going to rebase https://review.openstack.org/#/c/267587/ on top of https://review.openstack.org/#/c/529695/ ?
16:31:15 stephenfin jaypipes: Thank you, sir
16:31:20 kashyap stephenfin: Yep
16:31:58 ildikov mriedem: I added it in 'Depends-On'
16:32:15 ildikov mriedem: I can make a chain if you would prefer that
16:32:35 ildikov ?
16:33:30 stephenfin mdbooth: Given the similarities to your own BDM patch, wanna take a shot at https://review.openstack.org/#/c/530487/ this week, once you're feeling better?
16:34:26 mriedem ildikov: the chain is better when it's in the same repo
16:34:33 mriedem IMO
16:35:47 ildikov mriedem: ok, will make it a chain
16:37:09 mriedem stephenfin: is this part of the pci policy bp i just marked as complete? https://review.openstack.org/#/c/530487/
16:37:19 mriedem https://blueprints.launchpad.net/nova/+spec/share-pci-between-numa-nodes
16:37:41 stephenfin mriedem: Nope, nothing to do with that. Just a nice-to-have that was talked about here before Xmas
16:37:50 mriedem ok i seem to remember that
16:38:11 mriedem pcidevice uuid was added for nested RPs?
16:38:22 stephenfin Yup
16:38:24 mriedem ok
16:39:58 kashyap stephenfin: Hmm, bizarre, despite that .tox.ini file you saw, it is still going to 3.6:
16:40:11 kashyap [...] nova-specs/.tox/docs/lib/python3.6/site-packages/yasfb/formatter.py", line 451, in format_rss2_file
16:40:14 kashyap handle.write(string.encode('utf-8'))
16:40:17 kashyap TypeError: write() argument must be str, not byte
16:40:26 kashyap Is there any other way I can force py2.7?
16:40:28 clarkb kashyap: you may need to run with tox -r for it to rebuild the tox env
16:40:44 clarkb if it already ran and built an older env with different python version it won't rebuilt it unless you pass -r
16:41:00 kashyap clarkb: I just nuked the .tox and re-ran, to precisely avoid that
16:41:16 kashyap But let be double-check for sure
16:45:33 openstackgerrit Ildiko Vancsa proposed openstack/nova master: [api] Allow multi-attach in compute api https://review.openstack.org/271047
16:45:33 openstackgerrit Ildiko Vancsa proposed openstack/nova master: [libvirt] Allow multiple volume attachments https://review.openstack.org/267587
16:46:28 ildikov mriedem: you were right as always :) done ^^
16:50:23 kashyap clarkb: Thanks; that was it; after re-building, this succeeds `tox -e py27`
16:52:40 openstackgerrit Kashyap Chamarthy proposed openstack/nova-specs master: Add UEFI Secure Boot support for QEMU/KVM guests, using OVMF https://review.openstack.org/506720
16:55:39 cdent jaypipes, edleafe, efried, jroll: I'm home now and mostly free.
16:57:54 openstackgerrit Merged openstack/python-novaclient master: Updated from global requirements https://review.openstack.org/530697
17:00:50 openstackgerrit Matthew Booth proposed openstack/nova master: Add test for assignment of uuid to a deleted BDM https://review.openstack.org/530774
17:04:38 mdbooth mriedem: ^^^
17:04:43 mriedem ack
17:10:43 mdbooth stephenfin: That pci uuid patch screams refactor with library code to me. You?
17:11:08 stephenfin mdbooth: It's getting there, but I could only find the two occurrences
17:11:08 mdbooth I mean, it's a whole bunch of cut/paste.
17:11:17 stephenfin One more and I'd have dragged it out
17:11:35 mdbooth Ok, if it's only 2.
17:11:46 mdbooth Still borderline, though.
17:13:48 mdbooth I just need legacy-tempest-dsvm-py35 to complete for the top patch, then the top 2 will merge.
17:13:59 mdbooth It's taking a suspiciously long time, though :(
17:14:16 mdbooth Perhaps my goat didn't find favour.

Earlier   Later