Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-09-26
14:31:31 sdague regardless seeing if you can get an inspect console on the issue would be handy
14:32:20 jaypipes sdague: will do
14:32:36 jaypipes efried: you want me to response here or on the review?
14:32:38 efried jaypipes Okay, so in the example in the comment I linked above: does that work as stated?
14:33:05 efried jaypipes Wouldn't mind discussing a bit, then putting a summary in the review. Just to make sure we're talking about the same thing, save time.
14:33:30 jaypipes efried: your statement there, right after "How about" is spot on.
14:34:03 jaypipes efried: this is excellent:
14:34:04 jaypipes "With nested resource providers, traits defined on a parent RP are assumed to belong to all its child (descendant) RPs. However, traits defined on a child RP do not apply to the parent (ancestor) RPs. There is no implied sharing of traits within aggregates."
14:34:46 jaypipes efried: even more explicit would be pointing out that aggregates don't actually have *any* traits associated to themselves at all (there's no aggregate_metadata table like there is in Nova)
14:34:55 jaypipes efried: only resource providers have traits associated with them.
14:35:18 efried Rightright, point being that RP1 doesn't inherit any traits from RP2 just because they're in the same aggregate.
14:35:59 jaypipes efried: correct. it's worth spelling that out. aggregates are only grouping mechanisms, nothing more.
14:36:31 efried jaypipes Okay, cool. So traits are inherited in NRPs, downwards but not upwards. And the example below that sentence would work as described. I guess the implementation details aren't important, but I'm a bit curious how it would work if you didn't actually internally copy the traits from the parent to its children.
14:37:37 jaypipes efried: don't worry about the implementation details of the queries at this point.
14:38:43 efried jaypipes Roger that. So okay, it sounds like we're in agreement. Thanks for the talk.
14:39:50 jaypipes efried: just typing up in the review... gimme a few
14:39:59 efried ++
14:43:51 ratailor sdague, you around ?
14:43:51 jaypipes alex_xu: not sure if you're still up, but if you can make the few requested changes on https://review.openstack.org/#/c/497713/ from me and efried, I'm ready to +2 that one.
14:44:04 openstackgerrit Matthew Booth proposed openstack/nova master: Make etree.tostring() emit unicode everywhere https://review.openstack.org/507488
14:44:05 openstackgerrit Matthew Booth proposed openstack/nova master: libvirt: Don't VIR_MIGRATE_NON_SHARED_INC without migrate_disks https://review.openstack.org/507202
14:44:38 ratailor sdague could you have a look at https://review.openstack.org/#/c/504885/
14:46:12 openstackgerrit Balazs Gibizer proposed openstack/nova master: Moving more utils to ServerResourceAllocationTestBase https://review.openstack.org/499539
14:46:13 openstackgerrit Balazs Gibizer proposed openstack/nova master: Test resource allocation during soft delete https://review.openstack.org/495159
14:46:13 openstackgerrit Balazs Gibizer proposed openstack/nova master: factor out compute service start in ServerMovingTest https://review.openstack.org/503037
14:46:57 openstackgerrit Moshe Levi proposed openstack/nova master: Don't overwrite binding-profile https://review.openstack.org/505613
14:48:05 gibi sdague: I'm looking at https://review.openstack.org/#/c/505673 (qemu 2.10 support). Do we have a way to unit test the change you made in the libvirt/driver.py ?
14:52:02 openstackgerrit Rodolfo Alonso Hernandez proposed openstack/nova master: Change 'InstancePCIRequest' spec field https://review.openstack.org/449257
14:52:14 openstackgerrit Rodolfo Alonso Hernandez proposed openstack/nova master: Read Neutron port 'binding_profile' during boot https://review.openstack.org/507481
14:54:59 gibi if somebody want some easy patches to review then I'd like to suggest a test improvement series starts https://review.openstack.org/#/c/499539 and mriedem already +2 on it.
14:55:58 bauzas jaypipes: I did noticed
14:56:05 bauzas jaypipes: Firefox FTW
14:56:19 jaypipes bauzas: I'm on firefox...
14:56:31 bauzas when I say FTW, it's sarcastic
14:56:44 bauzas so yeah, I'm hitted too
14:57:09 bauzas I suspected a French regulation cause, but looks like it's not :p
14:58:07 mriedem gibi: good point, there is no unit test coverage of the libvirt driver change or the images change to use the new flag
14:58:14 mriedem gibi: -1 that sonofa
14:59:21 gibi mriedem: I think the image part is covered by https://review.openstack.org/#/c/505673/5/nova/tests/unit/virt/libvirt/test_utils.py@181
15:00:31 bauzas mriedem: not sure I love assigning a new module attribute for the QEMU version, but I guess it's probably the only easy way
15:00:51 bauzas mriedem: I'm thinking of cases where you don't reload your module and you're doomed
15:00:57 mriedem oh i see
15:01:21 bauzas but I guess upgrading qemu would be part of a nova minor upgrade as well, so that would require restarting n-cpu as well
15:01:33 mriedem bauzas: jamespage went an alternate route where the version was passed down through the many plumbing layers, and it was quickly unmanageable
15:01:53 mriedem another alternative is handling the error, parsing the output and retrying with the flag, which also sucks
15:01:56 bauzas mriedem: yeah, I was thinking the image module wasn't written like that, but unfortunately it is
15:02:06 bauzas it's just an helper module
15:02:35 bauzas so no way to pass a flag but just adding a global var which sucks in my mind but which sucks less than the other possibilities you envisaged
15:10:44 mriedem weird, seeing this in the logs when creating 500 servers using the fake driver
15:10:46 mriedem Sep 26 15:09:38 devstack nova-compute[30351]: INFO nova.scheduler.client.report [None req-f0c0e899-18e6-47f9-b13b-34829646d07e demo demo] Deleted allocation for instance 62604071-77c2-46e1-9f57-d7192edc3f82
15:10:46 mriedem /nova/compute/resource_tracker.py:1171}}
15:10:46 mriedem Sep 26 15:09:38 devstack nova-compute[30351]: DEBUG nova.compute.resource_tracker [None req-f0c0e899-18e6-47f9-b13b-34829646d07e demo demo] Instance 62604071-77c2-46e1-9f57-d7192edc3f82 has been deleted (perhaps locally). Deleting allocations that remained for this instance against this compute host: {u'resources': {u'VCPU': 1, u'MEMORY_MB': 512, u'DISK_GB': 1}}. {{(pid=30351) _remove_deleted_instances_allocations /opt/stack/
15:11:33 mriedem doesn't really matter for what i'm testing, but it's odd
15:15:14 mriedem stephenfin: what's your actual -1 on this now? https://review.openstack.org/#/c/457532/
15:15:18 mriedem i'd like to see this get in so i can use it in devstack
15:15:33 mriedem dtroyer told us to make the command names "openstack resource provider <action>"
15:15:42 mriedem to be consistent with everything else in osc
15:15:52 mriedem being new to osc, i'm going to follow his recommendations
15:16:19 stephenfin mriedem: There isn't one. Removed now
15:16:24 mriedem thanks
15:16:35 mriedem and yeah, it's a bit large, but it lays the base crud ops for RPs
15:16:48 mriedem plus the common framework stuff
15:21:42 cdent jaypipes: one of the other things I did almost immediately after you asked was limiting aggregate checks: https://review.openstack.org/#/c/489633/
15:23:00 sdague gibi: it's tested in images.py
15:23:08 sdague mriedem / gibi: https://review.openstack.org/#/c/505673/5/nova/tests/unit/virt/libvirt/test_utils.py
15:23:24 jaypipes cdent: cool, will check shortly.
15:23:55 mriedem sdague: i think he meant there is no unit test for the driver code setting the variable in https://review.openstack.org/#/c/505673/5/nova/tests/unit/virt/libvirt/test_driver.py
15:23:58 gibi sdague: that test sets the images.QEMU_VERSION directly, but you have some code in the driver.py that sets images.QEMU_VERSION
15:24:46 mriedem like test_next_min_qemu_version_ok
15:24:55 openstackgerrit Chris Dent proposed openstack/nova master: [placement] gabbi tests for shared custom resource class https://review.openstack.org/485209
15:25:03 mriedem gibi: you can -1 and i can add the test later
15:25:10 sdague mriedem: he already did
15:25:16 mriedem ok
15:26:01 gibi :)
15:27:02 mriedem dansmith: you said os reboot and restack would cleanup devstack?
15:27:29 dansmith mriedem: reboot and stack
15:27:38 mriedem ok
15:32:23 openstackgerrit Evgeny Antyshev proposed openstack/nova master: Add ploop procedures to privsep.libvirt https://review.openstack.org/507569
15:40:38 openstackgerrit Sean Dague proposed openstack/nova master: Support qemu >= 2.10 https://review.openstack.org/505673
15:40:47 sdague gibi / mriedem unit test added
15:41:55 gibi sdague: looking
15:45:20 gibi sdague: thanks, looks good, +2
15:45:47 sdague mriedem: ok, you next, reminding you of the "be faster and merge things" principle from the PTG :)
15:47:17 dansmith sdague: he's currently working on being fast and merge things on my instance list patch
15:47:28 dansmith sdague: but I'll look at yours so he can continue to be fast and merge on my patch
15:47:32 sdague :)
15:47:40 sdague it's pretty small and should be straight forward
15:50:36 mriedem sdague: i shouldn't approve my own change
15:51:48 sdague mriedem: it's barely your change any more
16:02:22 gibi I've just realized that I have to migrate the notification burndown chart from openshift 2 to 3 until end of September. This will be joyful
16:02:35 mdbooth stephenfin: Are the changes to https://review.openstack.org/#/c/507488/ evil?
16:03:07 mdbooth Thinking specifically of the 'if six.py2:' in libvirt/guest.py
16:03:26 mdbooth I still think it's saner to have unicode everywhere, and convert to something else at the point of use
16:04:12 dansmith cdent: are you working your way up that migration uuid series? if so, I'll hold off pushing that fix you just identified until you have a chance to nit out on anything else
16:04:27 stephenfin mdbooth: No, I'd probably do the same thing
16:04:40 dansmith meaning, you can identify more nit-ish things and I can fix them, vs. just pushing for nits later
16:04:59 cdent dansmith: yup (on phone at the moment though)
16:05:15 dansmith cdent: okay, np
16:06:55 stephenfin mdbooth: though I would like to see one of the tests slightly reworked, if that makes sense to you?
16:07:47 mdbooth stephenfin: Sure thing. I'll take a look in the morning.

Earlier   Later