Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2019-01-22
22:24:57 openstackgerrit Matt Riedemann proposed openstack/nova master: Don't call begin_detaching when detaching volume from shelved vm https://review.openstack.org/624959
22:36:55 jaypipes mriedem: any reason this wasn't backported to ocata that I should know about? https://bugs.launchpad.net/nova/+bug/1750623
22:36:57 openstack Launchpad bug 1750623 in OpenStack Compute (nova) pike "rebuild to same host with different image shouldn't check with placement" [Medium,Fix committed] - Assigned to Lee Yarwood (lyarwood)
22:37:21 jaypipes mriedem: I don't see anything specifically about ocata but wondered if you remembered some difficulty getting that onto ocata I should be aware of...
22:48:06 mriedem jaypipes: ocata is in extended maintenance mode so it was probably just a matter of not doing it unless someone cared
22:49:20 mriedem i'm not opposed to backporting that
23:12:09 openstackgerrit Merged openstack/nova master: Use X-Forwarded-Proto as origin protocol if present https://review.openstack.org/631103
23:12:17 openstackgerrit Merged openstack/nova master: Extend NeutronFixture to return port with resource request https://review.openstack.org/630719
23:31:22 openstackgerrit Takashi NATSUME proposed openstack/nova master: Add minimum value in max_concurrent_live_migrations https://review.openstack.org/632240
#openstack-nova - 2019-01-23
00:21:59 openstackgerrit Merged openstack/nova master: Convert port to str when validate console port https://review.openstack.org/626853
02:09:14 openstackgerrit Merged openstack/nova master: Move simple execute call to processutils. https://review.openstack.org/621527
03:09:44 openstackgerrit yenai proposed openstack/nova master: Disconnect volume when encryption is failed https://review.openstack.org/632598
03:46:24 openstackgerrit zhufl proposed openstack/nova master: Add missing ws seperator between words https://review.openstack.org/632606
05:08:22 openstackgerrit Yongli He proposed openstack/nova master: Add server subresouce toplogy API https://review.openstack.org/621476
06:01:37 openstackgerrit yenai proposed openstack/nova master: Disconnect volume when encryption is failed https://review.openstack.org/632598
06:43:11 openstackgerrit zhufl proposed openstack/nova master: Add missing ws seperator between words https://review.openstack.org/632606
06:44:18 openstackgerrit Dakshina Ilangovan proposed openstack/nova-specs master: Nova changes to support Local Resource Management that uses Resource Management Daemon https://review.openstack.org/630817
07:40:25 openstackgerrit zhufl proposed openstack/nova master: Add missing ws seperator between words https://review.openstack.org/632606
08:58:46 openstackgerrit Zhenyu Zheng proposed openstack/nova master: Per-instance serial number https://review.openstack.org/619953
09:03:29 openstackgerrit yenai proposed openstack/nova master: Ignore VolumeAttachmentNotFound exception in compute.manager https://review.openstack.org/632657
09:04:17 gameon Does anyone run AMD & Intel in their OpenStack environments together? Is it inadvisable?
09:12:15 kashyap gameon: I'm not an operator, but as long as you ensure to schedule your guests in a "pool of AMD hosts" vs. "pool of Intel hosts", it should be fine?
09:12:29 kashyap To allow live migration compatibility
09:13:04 gameon kashyap: Thank you - I understand live migration wouldn't work, but the AMD pricing seems very keen these days!
09:13:15 kashyap E.g. also check out: https://docs.openstack.org/nova/rocky/user/aggregates.html
09:13:49 kashyap gameon: Heh, yeah. Also recall that last year we had a string of hardware CVEs
09:13:56 kashyap So ... that makes things "interesting".
09:14:40 kashyap gameon: FWIW, I gave a talk at the recently concluded Berlin Summit on that: https://www.openstack.org/summit/berlin-2018/summit-schedule/events/22105/effective-virtual-cpu-configuration-in-nova
09:15:29 kashyap The slides are fairly self contained, and should give a sense of what's involved
09:15:53 kashyap (For Operators, particularly this section is relevant: "x86: Updating to patched vCPU models")
09:17:26 gameon Thank you very much, that's really helpful. I was at the Summit but didn't catch your talk :)
09:17:38 kashyap gameon: Also Nova has this thing called "host aggregates", that lets the scheduler make more sensible decisions based on certain properties.
09:18:11 kashyap gameon: Yeah, no worries. There's too much input at these events, and human attention is limited :-)
09:56:49 kashyap mdbooth: Hi, when you get a moment, I'm stumped with this unit test failure (after MIN{LIBVIRT_QEMU}_VERSION bump): http://logs.openstack.org/07/632507/2/check/openstack-tox-py27/6f87de4/testr_results.html.gz
09:57:18 kashyap (See the first one: test_get_guest_config_ppc64_through_image_meta_spice_enabled)
09:57:28 kashyap testtools.matchers._impl.MismatchError: '<nova.virt.libvirt.config.LibvirtConfigMemoryBalloon object at 0x7f10ac940150>' is not an instance of LibvirtConfigGuestVideo
09:57:28 kashyap raise mismatch_error
09:57:28 kashyap [...]
09:57:31 kashyap [...]
09:57:43 openstackgerrit caoxufeng proposed openstack/nova master: oslo.config==6.8.0 requires rfc3986>=1.2.0 and requirements project had made rfc3986==1.2.0 in upper-constraints.txt,so modify it . https://review.openstack.org/632391
09:58:55 mdbooth kashyap: Looking at your patch
10:02:14 mdbooth kashyap: Ok, I can reproduce the failure locally
10:03:07 kashyap Nod
10:03:51 kashyap (There's a typo in the commit message: 1.3.1 --> 3.0.0)
10:04:01 kashyap (But that's besides the point.)
10:05:35 mdbooth kashyap: Ok, to start off with this is just a bad test.
10:05:53 mdbooth self._test_get_guest_config_ppc64(8)
10:06:17 mdbooth 8 is an uncommented bare constant, the best kind of constant
10:07:01 mdbooth It appears to describe the expected location of the video device in the ordered list of devices generates by _get_guest_config
10:07:17 mdbooth I can't imagine why that's a good idea
10:07:39 kashyap Yeah, there's many of those undocumented constants :-(
10:08:12 kashyap There are in total 33 failures, though. Still ploughing through
10:08:23 mdbooth So my expectation is that if we go dig in _get_guest_config, something in there is guarding the creation of devices based on libvirt or qemu version number
10:08:37 mdbooth Which breaks this test, which expects its device to be at exactly position 8
10:08:53 mdbooth It's a stupid test, but unfortunately you have to fix it now
10:08:58 kashyap Yeah, I have that open, and looking at it
10:09:28 kashyap Had the same thing last year -- spent weeks fixing all kinds of unit test failures due to the bump
10:10:26 kashyap mdbooth: Thanks for the hint
10:10:31 mdbooth It's almost as bad as the 'code diff' tests we used to write, but fortunately seem to be going out of fashion now.
10:23:45 kashyap mdbooth: So if I changed the position from 8 --> 7 "it succeeds":
10:23:46 kashyap + self._test_get_guest_config_ppc64(7)
10:23:46 kashyap - self._test_get_guest_config_ppc64(8)
10:23:57 kashyap That "solution" makes me feel physically stupid.
10:24:55 mdbooth kashyap: If you do that: https://goo.gl/images/6ehtJv
10:25:23 kashyap :D
10:26:16 mdbooth kashyap: Although it might be the pragmatic thing to do. The problem is that the correct thing to do here is to determine what the test was actually trying to test, and rewrite it so that it tests it.
10:26:35 mdbooth That would require mental effort from the reviewer which you may find hard to come by.
10:26:44 kashyap Yeah, I will freely admit that I'm "still in the process of grokking it all"
10:26:58 mdbooth From experience, I can say that nobody likes reviewing long series of test fixes.
10:27:04 kashyap mdbooth: Yeah, I don't think anyone gives a damn thing about it
10:27:17 kashyap mdbooth: I will for now do the "pragmatic thing" (I feel really dirty doing it)
10:27:38 mdbooth At least write a comment containing an apology and blaming somebody else ;)
10:27:41 kashyap Because I hate that I don't understand the root cause entirely. I probably will if I spend my whole day duking around various files
10:27:46 kashyap mdbooth: Yeah, good point.
10:27:56 kashyap "Leave the trace for the poor sucker who comes later."
10:42:09 openstackgerrit caoxufeng proposed openstack/nova master: oslo.config==6.8.0 requires rfc3986>=1.2.0 and requirements project had made rfc3986==1.2.0 in upper-constraints.txt,so modify it . https://review.openstack.org/632391
11:39:54 openstackgerrit Theodoros Tsioutsias proposed openstack/nova master: Add requested_networks to RequestSpec https://review.openstack.org/570201
11:39:55 openstackgerrit Theodoros Tsioutsias proposed openstack/nova master: [WIP] Enable rebuild for instances in cell0 https://review.openstack.org/570203
11:39:55 openstackgerrit Theodoros Tsioutsias proposed openstack/nova master: Add instance hard delete https://review.openstack.org/570202
12:47:55 kashyap mdbooth: Tiny progress!
12:48:27 kashyap I at least see a "pattern" that some of these tests seem to fail only when I update FAKE_QEMU_VERSION
12:48:34 kashyap If that is unchanged, then the test succeeds.
12:50:01 mdbooth kashyap: But you need to change it...
12:50:32 kashyap Yes, yes. I was isolating the problem
12:50:47 kashyap mdbooth: For example, see this small PDB session w/ & without my patch for guest.devices: http://paste.openstack.org/show/743172/
12:51:08 kashyap This is a different test that also fails due to:
12:51:10 kashyap self.assertEqual(2, len(guest.devices))
12:51:19 kashyap [...]
12:51:24 kashyap testtools.matchers._impl.MismatchError: 2 != 1
12:52:06 kashyap The fakelibvirt.py is "misbehaving" due to the version change.
12:59:59 mdbooth kashyap: There are way too many version checks for me to audit quickly. Honestly I'm not sure it's worth it, though. Could you not just change the test to scan the list of the devices for the one it's looking for?
13:05:47 kashyap I think I've got a lead; will update once done.
13:05:55 kashyap mdbooth: Please disregard me for now.
13:07:28 mdbooth kashyap: If I wanted to track it down quickly, I'd modify the unit test to spit out the device list, compare the generated devices with and without the bumped version number, then go find the guard in that code.
13:08:40 kashyap mdbooth: I don't know how to get the device list.
13:08:52 kashyap That's why I went the PDB route with and without my patch
13:09:16 kashyap And got that: dir(guest.devices[0]) output at the bottom: http://paste.openstack.org/raw/743168/
13:09:35 mdbooth kashyap: Right, so you've worked out what device was removed?
13:10:07 kashyap 'port' seems to be missing on the second iteration
13:10:19 kashyap mdbooth: And ... weirdly enough, see this comment from migi on the review:

Earlier   Later