| *** markvoelker has quit IRC | 00:01 | |
| *** tyr has joined #openstack-meeting-cp | 00:10 | |
| *** tyr has quit IRC | 00:15 | |
| *** tyr has joined #openstack-meeting-cp | 01:54 | |
| *** markvoelker has joined #openstack-meeting-cp | 01:58 | |
| *** tyr has quit IRC | 01:59 | |
| *** markvoelker has quit IRC | 02:02 | |
| *** coolsvap has joined #openstack-meeting-cp | 03:20 | |
| *** tyr has joined #openstack-meeting-cp | 03:38 | |
| *** tyr has quit IRC | 03:42 | |
| *** markvoelker has joined #openstack-meeting-cp | 03:59 | |
| *** markvoelker has quit IRC | 04:03 | |
| *** tyr has joined #openstack-meeting-cp | 05:26 | |
| *** tyr has quit IRC | 05:31 | |
| *** markvoelker has joined #openstack-meeting-cp | 05:59 | |
| *** markvoelker has quit IRC | 06:04 | |
| *** tyr has joined #openstack-meeting-cp | 07:14 | |
| *** tyr has quit IRC | 07:19 | |
| *** markvoelker has joined #openstack-meeting-cp | 08:00 | |
| *** markvoelker has quit IRC | 08:05 | |
| *** tyr has joined #openstack-meeting-cp | 09:02 | |
| *** tyr has quit IRC | 09:06 | |
| *** tyr_ has joined #openstack-meeting-cp | 10:50 | |
| *** tyr_ has quit IRC | 10:55 | |
| *** sdague has joined #openstack-meeting-cp | 11:49 | |
| *** markvoelker has joined #openstack-meeting-cp | 12:01 | |
| *** markvoelker has quit IRC | 12:06 | |
| *** flaper87 has quit IRC | 12:09 | |
| *** flaper87 has joined #openstack-meeting-cp | 12:15 | |
| *** flaper87 has quit IRC | 12:15 | |
| *** flaper87 has joined #openstack-meeting-cp | 12:15 | |
| *** sdague has quit IRC | 12:16 | |
| *** sdague has joined #openstack-meeting-cp | 12:16 | |
| *** coolsvap has quit IRC | 12:25 | |
| *** markvoelker has joined #openstack-meeting-cp | 12:25 | |
| *** scottda has joined #openstack-meeting-cp | 12:32 | |
| *** tyr_ has joined #openstack-meeting-cp | 12:38 | |
| *** tyr_ has quit IRC | 12:43 | |
| *** xyang1 has joined #openstack-meeting-cp | 12:55 | |
| *** sheel has joined #openstack-meeting-cp | 13:05 | |
| *** xyang1 has quit IRC | 14:22 | |
| *** xyang1 has joined #openstack-meeting-cp | 14:25 | |
| *** tyr_ has joined #openstack-meeting-cp | 14:25 | |
| *** tyr_ has quit IRC | 14:30 | |
| *** sigmavirus24_awa is now known as sigmavirus24 | 14:32 | |
| *** coolsvap has joined #openstack-meeting-cp | 15:05 | |
| *** tyr_ has joined #openstack-meeting-cp | 15:08 | |
| *** tyr_ has quit IRC | 15:12 | |
| *** piet has joined #openstack-meeting-cp | 15:14 | |
| *** piet has quit IRC | 15:52 | |
| *** piet has joined #openstack-meeting-cp | 15:52 | |
| *** tyr_ has joined #openstack-meeting-cp | 15:59 | |
| *** smcginnis has quit IRC | 16:58 | |
| *** smcginnis has joined #openstack-meeting-cp | 16:58 | |
| *** mriedem has joined #openstack-meeting-cp | 16:58 | |
| ildikov | #startmeeting cinder-nova-api-changes | 17:00 |
|---|---|---|
| openstack | Meeting started Mon Jun 13 17:00:04 2016 UTC and is due to finish in 60 minutes. The chair is ildikov. Information about MeetBot at http://wiki.debian.org/MeetBot. | 17:00 |
| openstack | Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. | 17:00 |
| *** openstack changes topic to " (Meeting topic: cinder-nova-api-changes)" | 17:00 | |
| openstack | The meeting name has been set to 'cinder_nova_api_changes' | 17:00 |
| scottda | hi | 17:00 |
| ildikov | scottda ildikov DuncanT ameade cFouts johnthetubaguy jaypipes takashin alaski e0ne jgriffith tbarron andrearosa hemna erlon mriedem gouthamr ebalduf patrickeast smcginnis diablo_rojo gsilvis | 17:00 |
| ildikov | hi :) | 17:00 |
| mriedem | o/ | 17:00 |
| * alaski lurks | 17:00 | |
| * johnthetubaguy lurks | 17:00 | |
| ildikov | etherpad: #link https://etherpad.openstack.org/p/cinder-nova-api-changes | 17:00 |
| patrickeast | hey | 17:01 |
| ildikov | sorry for skipping last week | 17:01 |
| ildikov | I thought two items for today, which are the refactoring and its impacts that jgriffith is working on and the other would be the migration test, I think scottda has updates on that one | 17:03 |
| scottda | I've a migrations-with-BFV patch up: https://review.openstack.org/#/c/326681 | 17:03 |
| scottda | IT needs a devstack change (devstack-gate) which I need to put up and list as a dependency. | 17:04 |
| scottda | But is ready for testing if anyone is interested. | 17:04 |
| scottda | Once that is in, we'd want another test with multi-attach and cinder migrate. | 17:04 |
| scottda | But we said we wouldn't support multi-attach with BFV, so that'd be slightly different. But not more complicated. | 17:05 |
| ildikov | that looks good | 17:05 |
| ildikov | I think we will not support booting from a multi-attach volume | 17:05 |
| ildikov | but we will support attaching a multi-attach volume at boot time | 17:06 |
| * amrith lurks | 17:06 | |
| ildikov | so we have a negative test and a positive as well with the non-bootable one | 17:06 |
| ildikov | scottda: those few jobs are failing due to the missing dependency? | 17:06 |
| hemna | hey guys, I'm working on the grenade failures with os-brick right now | 17:07 |
| scottda | ildikov: Actually, one failure was a requirements issue for something that looked unrelated.... | 17:07 |
| jgriffith | o/ | 17:07 |
| jgriffith | sorry I'm a bit tardy | 17:07 |
| ildikov | hemna: ok | 17:08 |
| scottda | ildikov: The other seemed obscure as well. I did a rebase just to see if it was something fixed in tempest itself. | 17:08 |
| jgriffith | ildikov: scottda wait... not support multi-attach BFV? | 17:08 |
| ildikov | scottda: ok, got it | 17:08 |
| scottda | ildikov: Frankly, I'm learning as I go with regards to getting a Tempest test merged. Anyone with more expertise, let me know if you're available to bother with questions. | 17:09 |
| jgriffith | ildikov: scottda I though one of the primary use cases you wanted multi-attach for was to help live-migration of volume backed instances? | 17:09 |
| ildikov | jgriffith: we agreed to skip boot from multi-attach volume with mriedem and johnthetubaguy I think during Mitaka | 17:10 |
| scottda | jgriffith: At least for the first pass at multi-attach. | 17:10 |
| ildikov | scottda: I'm not a big expert either, but I'll try to take a look and be helpful | 17:11 |
| mriedem | i honestly don't exactly remember the reason why anymore | 17:11 |
| ildikov | mriedem: TBH I think it simply felt odd | 17:11 |
| scottda | I think we just didn't like the complexity, and were unsure of the corner cases. | 17:11 |
| mriedem | nova won't create a multiattach volume when booting, that would require API changes in nova | 17:12 |
| mriedem | i'm not sure what bringing a multiattach volume to attach at boot time would do | 17:12 |
| johnthetubaguy | +1 | 17:12 |
| mriedem | except maybe not landing on a host that could attach that kind of volume, i.e. mitaka node at this point | 17:12 |
| ildikov | I think simply attaching it, but not booting from it should be perfectly fine | 17:13 |
| ildikov | well, placement is an issue in other cases as well | 17:14 |
| johnthetubaguy | so I should check something here | 17:15 |
| johnthetubaguy | is that booting from an already created multi-attach | 17:15 |
| johnthetubaguy | or creating a multi-attach from an image during the boot process | 17:15 |
| johnthetubaguy | or both | 17:15 |
| mriedem | "or creating a multi-attach from an image during the boot process" - we're not doing thta | 17:15 |
| mriedem | you'd have to tell nova in the BDM dict to create multiattach=True which is more proxy | 17:15 |
| ildikov | right | 17:16 |
| mriedem | providing an existing multiattach volume to attach at boot time should be fine | 17:16 |
| mriedem | it's the same as attaching it to an existing instance | 17:16 |
| mriedem | most of this is covered in http://specs.openstack.org/openstack/nova-specs/specs/newton/approved/multi-attach-volume.html#rest-api-impact | 17:16 |
| ildikov | and I think I still have the code for failing when we're trying to boot from an existing multi-attach volume | 17:16 |
| mriedem | i think we can get around the placement issues if we just block in the nova-api that you can't do anything with multiattach volumes until all of the computes in the deployment are new enough to handle it | 17:17 |
| ildikov | source and dest is volume and bootindex is 0 | 17:17 |
| mriedem | ildikov: yup, that's in the spec | 17:17 |
| ildikov | yeap, checking that will help and then placement will only have to check whether we have libvirt or not | 17:18 |
| mriedem | well, | 17:18 |
| mriedem | unless we extend the compute capabilities filter or add a new scheduler filter, i'm not sure that's going to happen | 17:18 |
| mriedem | i have a feeling that non-kvm deployments of openstack will just block this via policy | 17:19 |
| mriedem | johnthetubaguy: is that what rax is/would do? | 17:19 |
| mriedem | probably on the cinder side so you can't create the multiattach volume to begin with | 17:19 |
| ildikov | that would be the easiest | 17:19 |
| ildikov | I think | 17:20 |
| mriedem | anyway, we're kind of getting off in the weeds on things that have already been discussed and are for the most part in the spec | 17:20 |
| ildikov | although we need to get there to have the Cinder changes in place to get to enable multi-attach | 17:20 |
| ildikov | jgriffith: can you give us an update on how the refactoring work goes? | 17:21 |
| jgriffith | ildikov: sure | 17:21 |
| jgriffith | ildikov: was hoping to get feedback on the code | 17:21 |
| ildikov | I added the links to the etherpad for the already uploaded patches | 17:21 |
| jgriffith | ildikov: I'm working on the detach side of things in Nova now | 17:21 |
| jgriffith | attach is working as expected, haven't cleaned out some of the things like calling initialize to get connection info | 17:22 |
| jgriffith | but I think that would be good follow up work | 17:22 |
| jgriffith | want to focus on the base attach/detach cases for now | 17:23 |
| ildikov | you mean you initialize_connection twice on Cinder side with the new code? | 17:23 |
| jgriffith | ildikov: NO | 17:23 |
| jgriffith | ildikov: I mean there's a few places where Nova uses initialize_connection calls to Cinder to re-establish connection info | 17:23 |
| ildikov | yeah, I agree it would be good to have the basic scenarios work before jumping on the more difficult ones | 17:23 |
| jgriffith | ildikov: with no intent of performing an attach | 17:23 |
| ildikov | jgriffith: ah, ok, I wasn't sure what you meant by missing clean up | 17:24 |
| jgriffith | ildikov: ie the volume is already attached, but they want to "refresh" the info | 17:24 |
| mriedem | is there going to be a problem with nova calling os-initialize_connection in non-attach cases? | 17:25 |
| jgriffith | ildikov: FWIW these changes will just support multi-attach for the most part without Nova really having to be involved (except for the libvirt setting) | 17:25 |
| ildikov | jgriffith: right, got it, I think there will be a few cases where we need to figure out what to do based on the instance actions johnthetubaguy described us a few meetings ago | 17:25 |
| jgriffith | mriedem: no... but I'd like to eventually make that go away | 17:25 |
| mriedem | jgriffith: with a new cinder api to get connection info? | 17:25 |
| jgriffith | mriedem: I wrote the code currently in such a way that all of the *old* calls work exactly as they did before | 17:25 |
| ildikov | jgriffith: yeah, I expected that to happen as we want to remove the handling of the volume state machine from Nova | 17:26 |
| jgriffith | mriedem: https://review.openstack.org/#/c/327408/ | 17:26 |
| jgriffith | mriedem: I'm just using wrappers of things for now | 17:26 |
| * johnthetubaguy shakes fist at his internet connection | 17:26 | |
| jgriffith | mriedem: that way it's relatively seemless, and we can continue progressing all the various implementation details over time | 17:27 |
| johnthetubaguy | mriedem: yeah, +1 on both counts, we just block until all compute nodes are upgraded | 17:27 |
| jgriffith | mriedem: rather than a single "big bang" wanted a phased approach that just got us over the API hump | 17:27 |
| jgriffith | mriedem: note the (outdated) gist of the nova changes that go with that patch | 17:28 |
| jgriffith | mriedem: I should have something at least suitable to post as a WIP to Nova shortly and link the cinder, cinderclient patches to it | 17:28 |
| ildikov | jgriffith: you already have the snippets up for attach, right? | 17:31 |
| mriedem | jgriffith: ok, posted a comment in your change | 17:31 |
| jgriffith | ildikov: yes | 17:31 |
| mriedem | jgriffith: would nova-api be calling create_attachment now? | 17:31 |
| jgriffith | mriedem: looking | 17:31 |
| mriedem | rather than the compute? | 17:31 |
| mriedem | it'd have to be compute i guess | 17:31 |
| mriedem | to pass the connector dict | 17:31 |
| ildikov | jgriffith: cool, I planed to play with it a bit in devstack | 17:31 |
| jgriffith | mriedem: yeah... so your points are part of what led me to this wrapper approach | 17:32 |
| jgriffith | mriedem: https://gist.github.com/j-griffith/ded4a08c488f0d90e90fff53afb2d7f1 | 17:32 |
| jgriffith | so for now, I only address things in the base attach case | 17:32 |
| jgriffith | mriedem: this way I don't change anything like swap_volume or the case with Cells etc | 17:33 |
| jgriffith | mriedem: my thought would be start with the base attach/detach cases. Get that merged, then do follow up patches to start hitting all the individual *special* cases | 17:34 |
| jgriffith | mriedem: I mostly want to keep the patch sizes manageable | 17:34 |
| mriedem | jgriffith: with https://gist.github.com/j-griffith/ded4a08c488f0d90e90fff53afb2d7f1 we'd still be reserving the volume in nova-api | 17:34 |
| jgriffith | mriedem: then deprecate the old calls, and mark them for removal next cycle | 17:34 |
| mriedem | would that make create_attachment fail? | 17:34 |
| jgriffith | mriedem: nope, it'll still work because I removed that step :) | 17:35 |
| ildikov | jgriffith: I would assume if we have more info stored on Cinder side along with the way more simpler API then things should work out for those *special* cases as well, but agreed on adressing them one-by-one | 17:35 |
| jgriffith | mriedem: That's something I should clean up a bit thought TBH | 17:35 |
| jgriffith | mriedem: I considered removing it from the nova-api and baking it in here, but instead just left it alone and instead "use" it | 17:36 |
| ildikov | mriedem: I think the aim is to replace reserve, initialize and attach with the single create_attachment call | 17:36 |
| jgriffith | mriedem: I still think there's value in reserve to avoid some possible race conditions | 17:36 |
| ildikov | mriedem: and by having only one call kill the race conditions as well | 17:37 |
| mriedem | yeah i only bring it up because you can have mitaka computes that aren't using create_attachment | 17:37 |
| jgriffith | ildikov: perhaps eventually. for now I'm just replacing initialize_connection and attach | 17:37 |
| jgriffith | mriedem: +1 | 17:37 |
| jgriffith | mriedem: that's why it ended up that way | 17:37 |
| mriedem | but again, nova can check the min compute service version in the deployment and make decisions based on that | 17:37 |
| ildikov | jgriffith: ok, then I'm a step ahead it seems :) | 17:37 |
| jgriffith | mriedem: I suppose ya'll will want something to deal with the nova-newton, cinder-mitaka... I didn't include that in my gist, but will look at making it happen | 17:39 |
| mriedem | yeah, microversion on the cinder side | 17:40 |
| mriedem | nova will need to know if cinder is new enough to call create_attachment | 17:40 |
| jgriffith | mriedem: ok, honestly I wanted to do this without using those, but no reason not to use them | 17:41 |
| jgriffith | mriedem: will just mean I have to finally figure them out and how they really work | 17:41 |
| jgriffith | :) | 17:41 |
| ildikov | jgriffith: are there any items any of us can help with? | 17:42 |
| jgriffith | ildikov: comment on the code that's there, add notes on what you might propose for micro-versions | 17:42 |
| jgriffith | ildikov: once I get the nova WIP posted test | 17:42 |
| ildikov | jgriffith: I'm not saying I'm friends with microversions, but I'm happy to give it a try :) | 17:43 |
| ildikov | jgriffith: ok, cool, will do | 17:43 |
| jgriffith | ildikov: and at that point honestly anybody that wants to joint-contribute on the nova side inparticular that would be great | 17:43 |
| mriedem | there is obviously a cinderclient change that's needed also | 17:43 |
| mriedem | https://gist.github.com/j-griffith/ded4a08c488f0d90e90fff53afb2d7f1#file-gistfile1-txt-L98 | 17:44 |
| scottda | jgriffith: I'll add some notes for microversions. We are friends. | 17:44 |
| ildikov | jgriffith: sounds good, tnx! | 17:44 |
| jgriffith | mriedem: posted https://review.openstack.org/#/c/327409/ | 17:44 |
| ildikov | scottda: :) | 17:44 |
| jgriffith | mriedem: FWIW those two reviews and the gist applied to Nova give a working attach change | 17:45 |
| jgriffith | mriedem: it doesn't include error-recovery, but I've got that and it will be in the Nova WIP when I get it posted | 17:45 |
| jgriffith | and of course no microvresioning etc | 17:45 |
| jgriffith | geeshh... *micro versioning* | 17:45 |
| mriedem | i'm trying to think through the nova-api changes when we get a multiattach volume | 17:48 |
| mriedem | because nova-api is still calling reserve, and the patch in mitaka needed to check if it's a multiattach volume | 17:48 |
| ildikov | I would assume check_attach is out of the game as well | 17:49 |
| ildikov | or at least it should be | 17:49 |
| ildikov | and that checked mutliattach | 17:49 |
| *** piet has quit IRC | 17:49 | |
| mriedem | b/c hemna is removing the calls to os-reserve? | 17:49 |
| ildikov | other than that there is the code for the BFV case | 17:49 |
| ildikov | mriedem: hemna has a patch up for removing check_attach as those checks are in os-reserve as well | 17:50 |
| mriedem | yeah https://review.openstack.org/#/c/315789/ | 17:50 |
| hemna | I've been absent from this meeting...sorry guys | 17:51 |
| ildikov | mriedem: so both Nova and Cinder performs the same checks | 17:51 |
| hemna | grenade issues. | 17:51 |
| ildikov | hemna: no worries, if you can take a look at jgriffith's patches that would be good, we talked about those mostly | 17:51 |
| hemna | ok I'll re-read the meeting after I'm done | 17:52 |
| ildikov | mriedem: so if jgriffith did not remove reserve for now it still does its job for checking the volume status | 17:52 |
| ildikov | I mean volume state, like 'attached', etc. and it's prepared for multiattach as well | 17:53 |
| ildikov | hemna: tnx | 17:53 |
| mriedem | ok 5 minutes left | 17:54 |
| ildikov | right, so as a summary we need reviews for jgriffith's patches, and also comment on the microversioning to have the simpler API calls for attach and then detach, we can move on to the more difficult cases after figuring out the basics | 17:55 |
| *** coolsvap has quit IRC | 17:56 | |
| ildikov | we also need an eye on hemna's remove check_attach patch as that would be another great cleanup item | 17:56 |
| mriedem | there are some comments in that one already | 17:56 |
| ildikov | we also talked about the Cinder migrate tests scottda is working on | 17:56 |
| *** coolsvap has joined #openstack-meeting-cp | 17:56 | |
| ildikov | mriedem: yeap I saw, I'll check on the details later | 17:57 |
| ildikov | is there anything for the today's meeting we would need to discuss? | 17:57 |
| mriedem | end it | 17:59 |
| ildikov | ok, I'll take it as a no :) | 17:59 |
| scottda | bye | 17:59 |
| ildikov | thanks everyone!!! | 17:59 |
| ildikov | #endmeeting | 17:59 |
| *** openstack changes topic to "OpenStack Meetings || https://wiki.openstack.org/wiki/Meetings" | 17:59 | |
| openstack | Meeting ended Mon Jun 13 17:59:45 2016 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4) | 17:59 |
| openstack | Minutes: http://eavesdrop.openstack.org/meetings/cinder_nova_api_changes/2016/cinder_nova_api_changes.2016-06-13-17.00.html | 17:59 |
| openstack | Minutes (text): http://eavesdrop.openstack.org/meetings/cinder_nova_api_changes/2016/cinder_nova_api_changes.2016-06-13-17.00.txt | 17:59 |
| openstack | Log: http://eavesdrop.openstack.org/meetings/cinder_nova_api_changes/2016/cinder_nova_api_changes.2016-06-13-17.00.log.html | 17:59 |
| *** mriedem has left #openstack-meeting-cp | 18:00 | |
| *** tyr_ has quit IRC | 18:00 | |
| *** ebalduf has quit IRC | 18:06 | |
| *** coolsvap has quit IRC | 18:13 | |
| *** sdake_ has joined #openstack-meeting-cp | 18:23 | |
| *** sheel has quit IRC | 19:35 | |
| *** cebreidian has quit IRC | 19:59 | |
| *** sdake_ is now known as sdake | 20:06 | |
| *** ericksonsantos has quit IRC | 20:59 | |
| *** rockyg has joined #openstack-meeting-cp | 21:04 | |
| *** rockyg has quit IRC | 21:47 | |
| *** beekhof has quit IRC | 22:59 | |
| *** sheeprine has quit IRC | 23:03 | |
| *** hemna is now known as hemnafk | 23:04 | |
| *** xyang1 has quit IRC | 23:05 | |
| *** sheeprine has joined #openstack-meeting-cp | 23:06 | |
| *** beekhof has joined #openstack-meeting-cp | 23:07 | |
| *** sdague has quit IRC | 23:38 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!