Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2018-08-08
12:43:03 holser_ lyarwood
12:43:05 artom OK, this is starting to get annoying, and consistently reproducible - I can't get a devstack compute to "join" another devstack allinone - it errors out with "Didn't find service registered by hostname after 60 seconds"
12:43:12 artom What am I doing wrong?
12:43:19 lyarwood holser_: \o_ hey
12:43:37 lyarwood holser_: https://review.openstack.org/589548 - can you also create a bug for this so we can backport it into stable?
12:43:38 holser_ concerning https://review.openstack.org/#/c/589548/ the valid case is u'[2001:0db8:85a3:0000:0000:8a2e:0370]:7334'
12:43:47 holser_ sure I will create it now
12:43:54 lyarwood holser_: thanks
12:44:22 lyarwood holser_: in terms of what is being passed in or expected?
12:44:42 holser_ what being passed
12:45:01 holser_ we have comment from sahid which is valid
12:45:15 holser_ also def _migrate_uri(dest): should be covered also
12:45:21 holser_ uri = 'tcp://%s' % dest
12:46:01 lyarwood holser_: kk, I'll leave that to you then
12:46:14 holser_ sure...
13:06:09 mdbooth lyarwood: test_instance_action passes :)
13:06:25 mdbooth lyarwood: Question is: what else did I break in the process? ;)
13:07:23 lyarwood mdbooth: welcome to the house of cards that is that fixture :)
13:15:32 mdbooth lyarwood: Ok, everything in test_instance passes
13:15:45 mdbooth lyarwood: Mind if I push it to let CI see what explodes?
13:16:11 mdbooth Actually, lemme just run the python2 tests first.
13:16:29 mdbooth Has to be a first for me
13:16:31 lyarwood mdbooth: yeah go for it when you're ready
13:17:01 lyarwood mdbooth: add yourself as a co-author while you're at it as well, assuming you've had a change things
13:17:27 lyarwood to change things* even
13:23:06 openstackgerrit Matthew Booth proposed openstack/nova master: fixtures: Track volume attachments within CinderFixtureNewAttachFlow https://review.openstack.org/587013
13:23:12 mdbooth lyarwood: ^^^
13:23:52 mdbooth Ended up having to go a bit back to basics.
13:25:18 mdbooth Because the bugs weren't in the fixture code, but rather in the simplifications it was intentionally making. Like using volume_id instead of attachment_id all over the place.
13:25:36 mdbooth As soon as we started asserting things about that, everything broke.
13:26:50 lyarwood mdbooth: yup indeed, thanks for that, I'll rebase the other changes on to this once it passes and sort the other nits out
13:37:38 openstackgerrit Takashi NATSUME proposed openstack/nova master: [placement] api-ref: add description for 1.29 https://review.openstack.org/589407
13:37:52 openstackgerrit Matthew Booth proposed openstack/nova master: fixtures: Fail deleting non-existent attachment in CinderFixtureNewAttachFlow https://review.openstack.org/589900
13:38:24 mdbooth lyarwood: Incidentally ^^^ removes a workaround which was previously required. No idea if it's still required, will find out.
13:38:25 openstackgerrit Takashi NATSUME proposed openstack/nova master: [placement] api-ref: add description for 1.29 https://review.openstack.org/589407
13:42:05 lyarwood mdbooth: kk, we can merge it in if it passes now
13:43:57 mdbooth lyarwood: ack
13:46:47 openstackgerrit Matt Riedemann proposed openstack/nova master: Update the parameter explain when updating a volume attachment https://review.openstack.org/565181
13:48:36 openstackgerrit Matt Riedemann proposed openstack/nova master: Update the parameter explain when updating a volume attachment https://review.openstack.org/565181
13:53:44 openstackgerrit Merged openstack/python-novaclient master: Fix server strings in reboot operation https://review.openstack.org/588981
13:57:10 mdbooth lyarwood: functional failed again. Coffee, then I'll take another look.
14:19:11 mriedem lyarwood: want to address the nit in https://review.openstack.org/#/c/589513/ and then i'll +2
14:20:02 lyarwood mriedem: yeah was about to ask if there was anything else before I hit it
14:22:57 openstackgerrit Lee Yarwood proposed openstack/nova master: libvirt: Reduce calls to qemu-img during update_available_resource https://review.openstack.org/589513
14:24:38 dansmith mriedem: fwiw, this was my idea: https://review.openstack.org/#/c/589567/3
14:24:49 dansmith mriedem: but I agree it would be nicer to do the thing that 95% of people need by default
14:25:09 lyarwood yeah sorry the commit just had it the wrong way around
14:25:26 dansmith mriedem: I just don't think jumping through a bunch of hoops to collect correct info that doesn't matter efficiently is worth it
14:25:28 lyarwood you only need the workaround if you are using DiskFilter etc
14:25:36 dansmith lyarwood: right, that'd be my preference
14:30:58 mriedem +2 on the bottom change
14:31:11 lyarwood thanks I'll sort this other mess out now
14:31:11 mriedem did s10 say if that fixes the majority of the perf regression?
14:32:13 lyarwood mriedem: just brings the time down by 50%, 20 seconds without this, 10 seconds with this, previously 2 seconds.
14:32:41 lyarwood mriedem: so the workaround should get us back to 2 seconds
14:33:38 openstackgerrit Matthew Booth proposed openstack/nova master: fixtures: Track volume attachments within CinderFixtureNewAttachFlow https://review.openstack.org/587013
14:33:39 openstackgerrit Matthew Booth proposed openstack/nova master: fixtures: Fail deleting non-existent attachment in CinderFixtureNewAttachFlow https://review.openstack.org/589900
14:33:47 mdbooth lyarwood: Attempt 2 ^^^
14:34:25 mdbooth Nothing architectural: just a couple of external tests asserting a variable I'd deleted
14:36:24 efried mriedem: o process guru, what do you think about swapping bug numbers here: https://review.openstack.org/#/c/401009/14//COMMIT_MSG@16
14:55:40 mriedem efried: yeah i noticed that as well when i looked at that bug,
14:55:54 mriedem https://review.openstack.org/#/c/447355/ marked it as closing the bug but it didn't really, it should have been partial
14:55:55 efried mriedem: If you agree, I'll flip it and fast approve
14:56:07 efried mriedem: Either that or mark the second bug as a dup
14:56:20 mriedem let me look at what changed since i last +2ed it
14:59:00 melwitt .
15:01:00 mriedem efried: commented
15:01:05 efried ack
15:02:17 openstackgerrit Eric Fried proposed openstack/nova master: Fix host validity check for live-migration https://review.openstack.org/401009
15:02:46 efried mriedem: Done and +A
15:05:57 openstackgerrit Matthew Booth proposed openstack/nova master: fixtures: Track volume attachments within CinderFixtureNewAttachFlow https://review.openstack.org/587013
15:05:58 openstackgerrit Matthew Booth proposed openstack/nova master: fixtures: Fail deleting non-existent attachment in CinderFixtureNewAttachFlow https://review.openstack.org/589900
15:06:07 mdbooth lyarwood: Attempt 3: I feel good about this one
15:11:42 openstack Launchpad bug 1785235 in OpenStack Compute (nova) "metadata retrieval fails when using a global nova-api-metadata service" [Undecided,Incomplete] - Assigned to Liam Young (gnuoy)
15:11:42 melwitt dansmith: there are more updates to this bug from gnuoy, having to do with the redirection api being used https://bugs.launchpad.net/nova/+bug/1785235
15:13:02 dansmith melwitt: okay never heard back from him here that he moved to the non-allinone metadata service
15:13:23 melwitt looking back at the IRC log from 2018-08-03, it looks like you said to run the metadata service with "the regular api server, and just not enable the osapi api if you want to only run metadata"
15:13:56 dansmith right
15:14:14 melwitt it seems like they haven't done that and I don't know how to tell him to do that, like what command and conf to use
15:14:23 gnuoy hi there
15:14:56 gnuoy I have tried moving away from a standalone server and that does work for me.
15:15:21 gnuoy However, I was hoping to get to the bottom of the issue with the standalone service
15:16:00 melwitt I thought we explained why that doesn't work, the database connection and message queue settings are intentionally not sent over RPC
15:16:05 dansmith gnuoy: the standalone service is really just for running on compute nodes in multihost n-net mode, which doesn't make sense for a neutron and multi-cell environment
15:16:21 dansmith which is why I said this is a doc bug if the regular api works for you
15:17:16 gnuoy ok, then thats a misunderstanding on my part. I didn't appreciate the situation was explicit as: not expected to work with neutron and cells
15:17:45 gnuoy in that case I'll update the bug saying that the central api service works for me.
15:17:53 gnuoy sorry for the noise.
15:18:05 dansmith no problem, glad to hear that it works as we expect though :)
15:21:35 openstackgerrit Takashi NATSUME proposed openstack/nova master: [placement] api-ref: add description for 1.29 https://review.openstack.org/589407
15:29:23 openstackgerrit Chris Dent proposed openstack/nova master: [placement] Avoid rp.get_by_uuid in allocation_candidates https://review.openstack.org/589941
15:35:26 openstackgerrit Liam Young proposed openstack/nova master: Remove Neutron MetaAPIProxy from cellsv2-layout https://review.openstack.org/588525
15:35:40 openstackgerrit Sergii Golovatiuk proposed openstack/nova master: libvirt: Always escape IPv6 addresses when used in migration URI https://review.openstack.org/589548
15:36:58 gnuoy dansmith, melwitt, fwiw I've had a go at updating the nova metadata verbage https://review.openstack.org/588525
15:39:21 melwitt cool, thanks
15:42:27 openstackgerrit Jay Pipes proposed openstack/nova master: get provider IDs once when building summaries https://review.openstack.org/589945
15:43:45 kosamara efried: Since yaml seems a better way forward, the spec I would propose would include most of the points of your spec. The larger differences I see so far mostly amount to device discovery and no auto-generated traits (following jaypipes' comment). Also, I would focus on generating the RPs and split the interactions with instances to a follow-up spec.
15:45:42 efried kosamara: Sounds good to me. As you're doing your writeup, it would be nice if you're careful to distinguish the responsibilities of the virt driver versus the compute manager. So for instance, the decision about whether to generate traits could/should (IMO) be left up to the individual driver.
15:46:23 mdbooth Is zuul grumpy about something?
15:46:37 mdbooth "Tenant openstack isn't ready"

Earlier   Later