Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-05-26
19:50:57 figleaf hey, we're growing carrots here!
19:51:02 cdent we've planted some
19:51:05 mriedem i was into alternate hosts before it was cool to be into alternate hosts
19:51:05 cdent but not ready yet
19:51:14 sdague oh, you know, we could add merged patches to the IRC bot here
19:51:23 mriedem sdague: yeah we should do that
19:51:26 sdague https://review.openstack.org/468527
19:51:28 mriedem i've liked patchbot in other channels
19:51:42 sdague mriedem: gerritbot does it
19:51:50 cdent alternate hosts is going to cause problems, I reckon
19:51:51 cdent we'll see
19:51:53 sdague just the nova channel doesn't have it turned on
19:51:57 sdague we do in others
19:52:01 mriedem yeah qa
19:52:04 mriedem qa is bot city
19:52:08 cdent have a good weekend all
19:52:09 mriedem +1
19:52:12 mriedem o/
19:52:15 sdague enjoy
19:52:46 sdague merged patches are nice to see things going in and feel like there is progress :)
19:55:33 efried leakypipes Been digging into get_inventory; can I confirm a couple things with you real quick?
19:58:41 efried instance was claimified. Have I got that right?
19:58:41 efried leakypipes First thing, I'm noticing the old get_available_resource included not just "how much is there" but "how much has been used". The latter doesn't appear to have a place in get_inventory(). I _think_ I've gleaned that this is because the layers outside of the driver (resource tracker? scheduler?) calculate that for me by summing up allocations for existing instances, said allocations having been created when the
19:58:49 mriedem sdague: progress is nice
19:58:51 mriedem fun is fun
20:02:34 efried leakypipes Second, on a similar note, I have a bunch of fields in get_available_resource that don't have standard ResourceClass affordance. I could create custom resource classes for those, but there doesn't seem much point if nothing is going to consume them. So do those just... disappear?
20:03:34 efried leakypipes E.g. https://github.com/openstack/nova/blob/master/nova/virt/powervm/host.py#L55-L64
20:07:55 openstack Launchpad bug 1691667 in OpenStack Compute (nova) "The metadata 8775 port has no authentication" [Undecided,New]
20:07:55 mriedem sdague: isn't this kind of a known thing? https://bugs.launchpad.net/nova/+bug/1691667
20:07:59 mriedem i don't really see it documented though
20:08:32 sdague yeh
20:09:12 sdague I thought there was host based authentication in there
20:09:26 sdague like it actually cares about request ip address
20:09:36 sdague but, that code is so old and wonky, idk
20:12:42 mriedem i might be thinking of it coming up during the summit
20:12:45 mriedem someone saying you can't use https
20:14:06 harlowja leakypipes so u going to finally make https://github.com/jaypipes/enamel a thing -P
20:14:45 sdague https://github.com/openstack/nova/blob/315ae692256939e6925db56c6ac833a237d46571/nova/api/metadata/handler.py#L130
20:15:09 sdague mriedem: we definitely validate against remote addr, or a proxy header equiv if you tell us to trust a proxy
20:15:37 sdague if you tell us to trust the proxy then you better not let your instances go straight to the server as well, all requests should only come from the proxy
20:15:49 sdague https://github.com/openstack/nova/blob/315ae692256939e6925db56c6ac833a237d46571/nova/api/metadata/handler.py#L131-L132
20:16:10 mriedem anyway ca'nt find that now
20:16:18 sdague mriedem: it's not over https
20:16:29 sdague but that's all limitations of the amazon implementation I thought
20:16:37 sdague also no ipv6 for same reason
20:18:02 sdague mriedem: I'll respond to the bug
20:28:50 sdague ok, headed out. I'm mostly out next week. Including some camping
20:28:52 sdague enjoy all
20:35:31 openstackgerrit Matt Wisch proposed openstack/nova master: Corrected compute_monitors setting help https://review.openstack.org/468538
20:45:18 openstack Launchpad bug 1689692 in OpenStack Compute (nova) "marker not found" [Undecided,New]
20:45:18 mriedem has it always been the case that if you specify a limit>what's left to page on after the marker we 404? https://bugs.launchpad.net/nova/+bug/1689692
20:45:24 mriedem that seems wrong
20:46:55 superdan limit or marker?
20:46:58 superdan marker makes sense to 404
20:51:37 mriedem the marker in this case exists
20:51:46 mriedem looks like all instances are ERROR state which means they should be in cell0
20:52:09 mriedem they are sending a limit which is 1 more than how many they have left in the page
20:55:10 mriedem superdan: i think i know
20:55:29 mriedem yup
20:55:58 mriedem so they are all in cell0
20:56:04 mriedem we get the instances from cell0 here https://github.com/openstack/nova/blob/master/nova/compute/api.py#L2554
20:56:14 mriedem with limit=3, we got 2 back b/c that was the end of the page
20:56:18 mriedem so we are left with limit 1
20:56:31 mriedem so we continue to page into the cells https://github.com/openstack/nova/blob/master/nova/compute/api.py#L2569
20:57:03 mriedem and we're going to fail here https://github.com/openstack/nova/blob/master/nova/compute/api.py#L2676
20:57:08 mriedem because the marker was in cell0
20:57:20 mriedem so the marker is never found in the other cells, but the limit said to keep checking
20:57:40 superdan we don't clear the marker once we find it?
20:57:46 superdan or maybe we do but not if we start on cell0?
20:58:15 mriedem i don't think we do if we start on cell0 yeah
20:58:30 superdan gotcha
21:02:39 mriedem writing a recreate functional test, should be simple
21:15:20 openstackgerrit Sean Dague proposed openstack/nova master: Migrate to oslo request_id middleware - mv 2.46 https://review.openstack.org/467998
21:18:54 sdague grrr, requirements conflict
21:19:00 sdague rebased and reapproved
21:21:07 mriedem damn this goes back to newton
21:21:19 mriedem functional regression test has got it recreated though
21:24:02 mriedem superdan: thar she blar ^ before you leave for 4 days
21:24:02 openstack bug 1689692 in OpenStack Compute (nova) "MarkerNotFound when limit>num(instances) and marker starts in cell0" [High,Triaged] https://launchpad.net/bugs/1689692 - Assigned to Matt Riedemann (mriedem)
21:24:02 openstackgerrit Matt Riedemann proposed openstack/nova master: Add recreate functional test for regression bug 1689692 https://review.openstack.org/468549
21:27:33 superdan mriedem: you're going to work on a fix too presumably?
21:27:40 superdan glad that was easy to repro,
21:27:43 openstack Launchpad bug 1693633 in openstack-manuals "Compute schedulers in Configuration Reference - AggregateImagePropertiesIsolation filter example" [Undecided,Incomplete]
21:27:43 mriedem sgordon: heh look familiar? https://bugs.launchpad.net/openstack-manuals/+bug/1693633
21:27:45 superdan but kinda uncool at the same ime
21:27:48 mriedem superdan: yeah
21:27:53 mriedem should be trivial
21:27:55 mriedem marker = None
21:28:11 superdan aye
21:40:18 mriedem superdan: those hypervisor stats backports are up btw
21:40:38 superdan okay but as soon as the ups man shows up with my weekend toys, I'm outta here
21:41:00 mriedem his truck broke down i heard
21:41:17 mriedem i have to leave soon too
21:44:13 mriedem crap of course there are failing compute api unit tests
21:48:29 openstackgerrit Ed Leafe proposed openstack/nova master: Have scheduler claim resources https://review.openstack.org/463993
21:48:30 openstackgerrit Ed Leafe proposed openstack/nova master: Remove selecting host based on host_subset_size https://review.openstack.org/465174
21:48:30 openstackgerrit Ed Leafe proposed openstack/nova master: Pass the number of hosts to return explicitly https://review.openstack.org/465767
21:48:31 openstackgerrit Ed Leafe proposed openstack/nova master: Return multiple hosts from select_destinations https://review.openstack.org/465768
21:48:31 openstackgerrit Ed Leafe proposed openstack/nova master: Pass a list of instance UUIDs to scheduler https://review.openstack.org/465171
21:49:03 cburgess superdan Toys?

Earlier   Later