Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-11-17
20:24:46 openstackgerrit Ed Leafe proposed openstack/nova master: Refactor the code to check for sufficient hosts https://review.openstack.org/520242
20:24:46 openstackgerrit Ed Leafe proposed openstack/nova master: Return Selection objects from the scheduler driver https://review.openstack.org/495854
20:24:47 openstackgerrit Ed Leafe proposed openstack/nova master: Modify select_destinations() to return objects and alts https://review.openstack.org/510159
20:24:47 openstackgerrit Ed Leafe proposed openstack/nova master: Change RPC for select_destinations() https://review.openstack.org/516707
20:24:48 openstackgerrit Ed Leafe proposed openstack/nova master: Make conductor pass and use host_lists https://review.openstack.org/511358
20:24:48 openstackgerrit Ed Leafe proposed openstack/nova master: Move the claim_resources method to scheduler utils https://review.openstack.org/511357
20:26:00 openstack Launchpad bug 1732947 in OpenStack Compute (nova) "volume-backed instance rebuild with no image change is still going through scheduler" [Undecided,In progress] - Assigned to Chris Friesen (cbf123)
20:26:00 mriedem dansmith: ok. i had to step away for a bit to clear my head - hadn't taken a break since i started today. i think i'm going to write a regression test for https://bugs.launchpad.net/nova/+bug/1732947 since i think rebuild + bfv with new image is something we need to fix anyway
20:26:13 mriedem and rebuild + bfv with original image for that matter
20:26:27 dansmith okay
20:27:54 openstackgerrit Dan Smith proposed openstack/nova master: WIP: Alternate fix for validating image on rebuild https://review.openstack.org/521186
20:28:29 dansmith mriedem: when you get a sec, let me explain my thinking in that test to make sure it's sound
20:30:47 mriedem i think i get it
20:31:04 mriedem we still have the allocation doubling up issue, but that's a separate problem
20:32:44 dansmith yeah
20:33:09 dansmith so if this is what you want, I need to go change the rest of the filters that are placement related to skip like this one does
20:36:57 dansmith mriedem: there's also the question of just doing this, or putting a straight revert in front
20:37:07 dansmith as this is, I'd have to backport the broken one first before I can backport this
20:37:24 dansmith which I can do for sure, but I'd vote to cleanly revert and re-fix
20:38:38 openstackgerrit Eric Fried proposed openstack/nova master: ProviderTree.uuid_set() https://review.openstack.org/520243
20:38:38 openstackgerrit Eric Fried proposed openstack/nova master: Scheduler set_inventory_for_provider does nested https://review.openstack.org/520643
20:38:39 openstackgerrit Eric Fried proposed openstack/nova master: SchedulerReportClient._get_providers_in_aggregates https://review.openstack.org/521097
20:38:39 openstackgerrit Eric Fried proposed openstack/nova master: SchedulerReportClient._get_providers_in_tree https://review.openstack.org/520663
20:38:40 openstackgerrit Eric Fried proposed openstack/nova master: ProviderTree.populate_from_iterable https://review.openstack.org/520756
20:38:40 openstackgerrit Eric Fried proposed openstack/nova master: WIP: Scheduler[Report]Client.get_provider_tree https://review.openstack.org/521098
20:38:41 openstackgerrit Eric Fried proposed openstack/nova master: WIP: Use update_provider_tree from resource tracker https://review.openstack.org/520246
20:38:41 openstackgerrit Eric Fried proposed openstack/nova master: WIP: ComputeDriver.update_provider_tree() https://review.openstack.org/521187
20:38:44 fried_rice leakypipes Nits all fixed up ^
20:39:07 leakypipes fried_rice: ah, you went with uuid_set(). noice.
20:39:13 fried_rice leakypipes I took the hint :)
20:39:23 leakypipes fried_rice: it was a gentle hint. :)
20:39:38 mriedem dansmith: comments inline
20:40:00 mriedem dansmith: and yeah, not sure about the revert at this point
20:40:53 fried_rice leakypipes This -1 was on the wrong patch btw - it stuck around after the rebase: https://review.openstack.org/#/c/520663/
20:41:12 fried_rice leakypipes ...just so you didn't bypass that guy for thinking you'd already seen it.
20:41:47 leakypipes fried_rice: gotcha.
20:41:59 leakypipes fried_rice: soon as tests complete, will re-review the series.
20:42:06 fried_rice leakypipes Dig
20:42:29 mriedem i'll work on the recreate test for https://bugs.launchpad.net/nova/+bug/1732947 and come back
20:42:29 openstack Launchpad bug 1732947 in OpenStack Compute (nova) "volume-backed instance rebuild with no image change is still going through scheduler" [Undecided,In progress] - Assigned to Chris Friesen (cbf123)
20:45:43 openstackgerrit Matt Riedemann proposed openstack/nova master: Add regression test for rebuild with new image doubling allocations https://review.openstack.org/521153
20:45:54 dansmith mriedem: abstractmethod implies that we change the base filter to inherit from ABCMeta which is more change
20:46:18 dansmith mriedem: we could also flip this and make the filters all skip if we're doing this policy check, unless they're opt-in as policy filters
20:46:40 dansmith mriedem: and then just have them set POLICY=True in their class definition
20:46:58 mriedem yeah i was thinking we could avoid the abc if we just have a default in the base and override in the subclasses
20:47:03 mriedem like we do for calling placement for the scheduler drivers
20:47:25 dansmith mriedem: and you want the default to be not-policy so we just override the ones that are policy?
20:47:55 mriedem if that means fewer overrides, then i think yes
20:48:04 dansmith I think it will
20:48:05 mriedem we want the default to be whatever the majority of filters want
20:48:06 mriedem ok
20:48:16 dansmith either way, I think the overrides should be in the policy ones, not the resource ones
20:48:28 mriedem yeah, like the affinity filters right?
20:48:44 dansmith er, I guess I have the override in the placementy one right now
20:48:55 dansmith lemme try to actually count how many overrides in each case
20:49:04 mriedem placementy == ram/disk/core filters?
20:49:28 dansmith yeah, mostly resource filters, except computefilter isn't really resourcey but is in that bucket
20:49:38 mriedem yeah that one is tricky
20:50:05 mriedem i was thinking about what you said about the computefilter not working for rebuild now,
20:50:09 mriedem but that's only if you're rebuilding with a new image,
20:50:21 mriedem and rebuilding with a new image on a disabled compute, that might not be the worst thing if we don't allow it
20:50:27 mriedem but it is a change
20:51:25 dansmith I think there are about 9 policy filters
20:51:51 dansmith maybe the best thing to do would be to just mark them all as either policy or not
20:58:00 artom We still a need default - can't operators have custom filters?
20:58:12 mriedem they can
20:58:14 mriedem but,
20:58:17 mriedem release note
20:58:30 artom Really, that's enough?
20:58:41 mriedem there isn't an api contract on out of tree things
20:58:43 jape2 Hi. is this the right place to ask for help regarding hyperv and dvr support for openstack pike?
20:58:44 dansmith artom: yeah, still default to false (i.e. skip) but explicitly mark all our own filters
20:59:12 artom mriedem, right, but we should still avoid breaking them, especially in easy cases like this
20:59:18 mriedem jape2: try #openstack-hyper-v
20:59:24 artom Anyways, dansmith nipped that in the bud
20:59:24 jape2 thanks!
21:03:30 openstackgerrit Eric Fried proposed openstack/nova master: Fix accumulated nits in refactor series https://review.openstack.org/521189
21:06:10 fried_rice figleaf dansmith mriedem leakypipes Anything for the weekly placement update email?
21:06:22 mriedem bugs
21:07:06 mriedem no idk, i haven't been too involved with placement lately
21:07:41 fried_rice figleaf I see movement on the series starting https://review.openstack.org/#/c/499239/ -- but still pinned -2. Is this ready to be unwedged yet?
21:07:51 mriedem no
21:10:37 openstackgerrit Dan Smith proposed openstack/nova master: WIP: Alternate fix for validating image on rebuild https://review.openstack.org/521186
21:11:01 dansmith mriedem: ^ updated approach, will start looking to see what unit test hell I've signed up for
21:11:54 dansmith artom: ^
21:21:43 mriedem this is pretty funny http://www.kttc.com/story/36870998/2017/11/17/police-string-of-burglaries-in-ne-rochester-lands-one-man-behind-bars
21:21:51 mriedem my kids school was in lockdown yesterday because of this joker
21:22:28 mriedem " She asked him what he wanted and he just threw a shoe at her and left."
21:23:00 cfriesen stealing someone's glasses? spraying deodorant on someone? this is a mentally ill guy, not a criminal mastermind
21:26:02 fried_rice mriedem This lightweight should rob my house. No taser required.
21:26:16 cfriesen mriedem: about that rebuild with boot-from-vol case, looks like a lot of the rebuild unit tests don't set up root_bdm properly. taking a look
21:26:38 mriedem cfriesen: he also got an umbrella and some air jordans
21:26:57 mriedem cfriesen: i'm writing a functional regression test for it
21:26:57 cfriesen there you go...set for life
21:27:07 cfriesen sweet
21:32:51 cfriesen mriedem: okay, so looks like there are no blockdevicemappings set up for these tests. could we set up a suitable blockdevicemapping for instance server1?
21:34:03 mriedem i don't grok the last part of your question - my functional test will probably create a volume-backed instance, using the CinderFixture
21:34:49 cfriesen I'm trying to figure out the most efficient way to solve all these failing unit tests. there's an instance that is created that is boot-from-vol, but compute_utils.get_root_bdm(context, instance) returns None
21:35:02 cfriesen sorry, boot-from-image
21:35:44 artom dansmith, left some minor comments

Earlier   Later