Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-12-07
15:42:20 mriedem ildikov: doesn't have anything to do with multiattach
15:42:30 stephenfin jianghuaw_: +2. Thanks for the follow-up
15:42:41 mriedem ildikov: i added todos like this https://review.openstack.org/#/c/525787/4/nova/virt/block_device.py@589
15:42:44 mriedem does that make it clear?
15:42:55 jianghuaw_ stephenfin, thanks.
15:43:18 ildikov mriedem: no :)
15:43:20 mriedem stephenfin: are you still ok with this https://review.openstack.org/#/c/507473/ ?
15:43:20 edleafe mriedem: I got the "AssertionError: The migration table left empty." error in one of my series too
15:43:35 edleafe mriedem: passed locally
15:43:38 ildikov mriedem: we reserve the volume now for BFV
15:43:52 ildikov mriedem: or do we not do that still for the case when Nova creates the volume?
15:44:04 stephenfin mriedem: Eek, seems like forever since I reviewed that. I'll take another look shortly
15:44:05 mriedem ildikov: that ^
15:44:24 mriedem ildikov: when you pass a volume to bfv, we reserve it in the api by creating an attachment (with the new patch)
15:44:33 mriedem when nova creates the volume, we don't create an attachment (on the compute)
15:44:36 mriedem so we attach using the old flow
15:44:46 mriedem because self['attachment_id'] isn't set in the DriverBlockDevice
15:44:52 mriedem hence the todo i left in that code
15:45:11 ildikov mriedem: isn't that refresh used for live_migrate?
15:45:22 stephenfin jaypipes, efried: Am I correct in saying libvirt doesn't currently report traits to placement?
15:45:30 mriedem ildikov: refresh?
15:45:34 jaypipes stephenfin: correct.
15:45:59 stephenfin jaypipes: Phew. And that's something not targeted for Queens
15:46:00 stephenfin ?
15:46:02 jianghuaw_ mriedem, are you happy to have a quick look on https://review.openstack.org/#/c/523360/ which patch you ever +2; now only changed the picture's background back to transparent.
15:46:08 jianghuaw_ so hope a simple one:-)
15:46:10 ildikov mriedem: sorry, looked at the wrong line, but you added a note into that one as well
15:46:17 ildikov mriedem: which I didn't really get
15:46:24 jaypipes stephenfin: no, not targeted
15:46:29 mriedem ildikov: link to the line please
15:46:35 stephenfin Ideal. Thanks, jaypipes
15:47:03 ildikov mriedem: what you pasted above: https://review.openstack.org/#/c/525787/4/nova/virt/block_device.py@589
15:47:27 mriedem ildikov: what don't you get about ^ ?
15:47:43 mriedem ildikov: in that code, nova is creating the volume
15:47:44 mriedem agree?
15:47:57 ildikov mriedem: that why we need to create a new attachment here too: refresh_connection_info
15:48:21 mriedem ildikov: do you mean https://review.openstack.org/#/c/525787/4/nova/virt/block_device.py@473 ?
15:48:24 jaypipes mdbooth: you working on fixing those unit test failures on the BDM uuid patch?
15:48:42 ildikov yeah, I agree with that, I just thought we covered everything for that case of BFV as well as we removed check attach, so things needed to be ok on the Cinder side as well regarding volume state, etc.
15:49:29 ildikov mriedem: no, I meant line 589, I don't get why we need to add that one
15:50:36 mriedem ildikov: if we create the volume but don't create the attachment,
15:50:43 mriedem when we call super(...).attach()
15:50:54 mriedem https://review.openstack.org/#/c/525787/4/nova/virt/block_device.py@511 will be True
15:51:02 mriedem and we will go down _legacy_volume_attach
15:51:16 ildikov sorry, my bad, mixed it up
15:51:19 ildikov forget that one
15:51:24 openstackgerrit Ed Leafe proposed openstack/nova master: Return Selection objects from the scheduler driver https://review.openstack.org/495854
15:51:24 openstackgerrit Ed Leafe proposed openstack/nova master: Refactor the code to check for sufficient hosts https://review.openstack.org/520242
15:51:25 openstackgerrit Ed Leafe proposed openstack/nova master: Modify select_destinations() to return objects and alts https://review.openstack.org/510159
15:51:25 openstackgerrit Ed Leafe proposed openstack/nova master: Move the to_dict() method to the Selection object https://review.openstack.org/523492
15:51:26 openstackgerrit Ed Leafe proposed openstack/nova master: Move the claim_resources method to scheduler utils https://review.openstack.org/511357
15:51:26 openstackgerrit Ed Leafe proposed openstack/nova master: Change RPC for select_destinations() https://review.openstack.org/516707
15:51:27 openstackgerrit Ed Leafe proposed openstack/nova master: Change compute RPC to use alternates for resize https://review.openstack.org/526436
15:51:27 openstackgerrit Ed Leafe proposed openstack/nova master: Make conductor pass and use host_lists https://review.openstack.org/511358
15:51:32 edleafe mriedem: jaypipes: ^^ rebased
15:51:46 edleafe mriedem: also added the alternates for resize patch
15:52:28 mriedem ildikov: so we're all good now?
15:52:55 mriedem edleafe: ok reapproved the bottom 4
15:53:09 ildikov I guess, I need to understand that part of the code as I thought we managed to cover everything regarding reserve back at the time of the old flow and check_attach removal
15:53:20 edleafe mriedem: thx
15:53:38 ildikov my bad, this code is super confusing... :/
15:53:40 mriedem ildikov: well, calling check_attach in the old flow on a volume that we *just* created was always kind of pointless
15:54:12 mriedem something would have had to slip in between the microseconds that the volume went to 'available' and we started using it
15:54:30 ildikov mriedem: I've never questioned that, I just said that we added extra an extra reserve call to have everything covered, but then it existed only in my head
15:54:50 mriedem i think we only ever reserved from the api
15:55:32 mriedem yeah https://github.com/openstack/nova/blob/stable/ocata/nova/virt/block_device.py#L250
15:55:39 mriedem check_attach wouldn't reserve
15:55:39 ildikov yeah, I wasn't sure where exactly we create that volume I guess
15:55:54 mriedem yup, for blank/image/snapshot it's created on the compute
15:55:57 mriedem right before we attach it
15:56:16 mriedem we should create the volumes way earlier, like in conductor, but that's not happening today
15:56:46 ildikov ok, so I'm just partially stupid, that's something :)
15:56:57 mriedem this is the attach code in ocata https://github.com/openstack/nova/blob/stable/ocata/nova/compute/api.py#L3666
15:57:08 mriedem https://github.com/openstack/nova/blob/stable/ocata/nova/compute/api.py#L3653
15:57:17 mriedem so that's the only time (aside from bfv with an existing volume) that we'd reserve
15:57:53 ildikov ok
15:58:24 mriedem i believe i have a todo sitting here to document how boot from volume actually works
15:58:29 mriedem because i explained it once in the mailing list
15:58:48 mdbooth jaypipes: Yeah. Working on something further down the stack atm.
15:59:08 mdbooth jaypipes: Wait, are they at the beginning of the stack, too?
15:59:13 openstackgerrit Balazs Gibizer proposed openstack/nova master: Add regression test for bug 1735407 https://review.openstack.org/526095
15:59:14 openstack bug 1735407 in OpenStack Compute (nova) "[Nova] Evacuation doesn't respect anti-affinity rules" [Medium,In progress] https://launchpad.net/bugs/1735407 - Assigned to Balazs Gibizer (balazs-gibizer)
15:59:20 jaypipes mdbooth: no, second patch in series
15:59:20 ildikov mriedem: if you have a link to that I would be happy to read it :)
15:59:47 mdbooth Hmm, that was careless.
15:59:51 mdbooth jaypipes: fixing it now.
16:00:26 ildikov mriedem: we can have a quick meeting and see what's next
16:01:11 mdbooth Ah, fixed that locally, didn't realise I'd pushed the old one, sorry.
16:07:26 openstackgerrit Matthew Booth proposed openstack/nova master: Make BlockDeviceMapping object support uuid https://review.openstack.org/242603
16:09:02 maciejjozefczyk Hello Guys, im trying to debug why I've faced a problem with scheduling. My instances stucks in 'scheduling' state. After enabling debug on nova-scheduler I've found that a host was selected. On nova-compute side there were 0 info about building an instance. It could be possible that the failure point is nova-conductor?
16:16:57 tasker morning! couple of questions about nova metadata service: why ( and what for ) are there different dates in the URI? does "latest" coalesce the various other date entries and get all of the latest date from them? does "latest" count as a different date than the others? where is the metadata physically stored ( on disk or in a database )?
16:23:55 mriedem sdague: efried: i'm surprised our docs don't say anything about configuring [neutron] with a user that has the admin role
16:24:01 mriedem because we definitely require that
16:24:10 mriedem for doing thigns with the port binding profile
16:24:29 mriedem maybe the install guide does, but the config reference doesn't
16:26:36 mriedem hmm, not here https://docs.openstack.org/nova/latest/install/compute-install-ubuntu.html
16:27:19 efried mriedem This coming from https://review.openstack.org/#/c/509892/ ?
16:28:20 mriedem no
16:28:23 mriedem https://bugs.launchpad.net/cinder/+bug/1736773
16:28:25 openstack Launchpad bug 1736773 in Cinder "attachment-show is including `connection_info` for non-admin callers, it shouldn't" [High,Triaged] - Assigned to John Griffith (john-griffith)

Earlier   Later