14:00:00 #startmeeting Nova Live Migration 14:00:01 Meeting started Tue Apr 19 14:00:00 2016 UTC and is due to finish in 60 minutes. The chair is PaulMurray. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:00:03 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:00:06 The meeting name has been set to 'nova_live_migration' 14:00:14 o/ 14:00:15 o/ 14:00:17 Hi all 14:00:21 o/ 14:00:27 hi 14:00:30 hi 14:00:51 I managed to start exactly on time today ! 14:01:00 lets wait on minute for later comers 14:01:01 (y) 14:01:18 in the meantime - agenda here: https://wiki.openstack.org/wiki/Meetings/NovaLiveMigration 14:01:19 hi 14:01:58 #topic Summit sessions 14:02:20 #link Nova Live Migration session: 14:02:20 https://www.openstack.org/summit/austin-2016/summit-schedule/events/9095?goback=1 14:02:42 I don't know why that came out on two lines 14:03:22 can't remember how to undo 14:03:25 #undo 14:03:26 Removing item from minutes: 14:03:44 #link Nova Live Migration session: https://www.openstack.org/summit/austin-2016/summit-schedule/events/9095?goback=1 14:03:54 anyway - we have one session 14:04:09 there are also two unconference sessions 14:04:11 o/ 14:04:18 and the contributor meetup on friday 14:04:39 So we can spread anything we want to talk about over those if needed 14:05:05 What should we give priority to in the Live Migration session ? 14:05:29 PaulMurray: I think it's worth to add storage pool specs to etherpad too? Or there will be no changes with it? 14:06:14 It will change, I'll get to it soon 14:06:48 We should focus on controversy. 14:07:03 e.g. post-copy 14:07:05 are you going to discuss about how to include post-copy? 14:07:29 we could do that 14:07:31 I plan to update both storage pools spec and use storage pools to do cold migrate to build on what mdbooth is doing to tidy up imagebackend etc 14:07:45 I will not attend the summit, but pkoniszenwski will and he will try to show the pros and cons of the approaches we've been discussing 14:09:04 I didn't see any more dicussion on the ML 14:09:16 So are we still with the options posted there originally 14:09:17 ? 14:09:20 I think it's important to decide on summit will we expose all new features over api or not 14:09:41 Post-Copy should be enabled via config setting, if enabled then it should be set in manual mode so monitor thread can activate it using API call as deemed appropriate 14:09:47 it will affect ops and it would be good to ask them 14:10:13 Lets do that in advance - has anyone put a post on the ops ML? 14:10:20 * kashyap joins late 14:10:42 I think the possition in nova is we cannot expose virt features in API 14:10:44 +1 for ops ml, to get feedback before summit 14:11:15 We can only expose "properties" of VMs 14:11:18 +1 for not exposing driver specific flags via API 14:11:21 at an abstract level 14:11:45 we did remove the disk_over_commit even because it was a driver specific thing 14:11:47 andrearosa: Not necessarily driver specific 14:12:24 well, here is spec from paul-carlton1 which tries to expose this things wrapping them in abstractions 14:12:34 so, there are options imho 14:12:54 I'd ask ops anyway 14:13:10 who wants to do an ops ML post ? 14:13:19 I could 14:13:49 #action tdurakov to post question to ops ML on live migration API for post-copy etc. 14:13:57 thanks 14:14:14 If we do this topic in the session it will eat up time, but its a good one 14:14:22 anything else that should be there ? 14:15:12 will the post-copy topic also cover auto-converge 14:15:18 i.e. will it be general ? 14:15:28 I'd say yes 14:15:33 not only post-copy 14:16:14 Again it should be able to be turned on or of in config 14:16:53 paul-carlton1, there was some debate about adding config options - so I guess this all comes together 14:17:16 if on then the live migration monitir thread can set the initial throttling and increment to control when it gets used and how aggresively 14:17:51 paul-carlton1: your spec claims to expose it tbh 14:18:02 with abstraction layer, but to expose 14:18:28 PaulMurray, agreed too many config options are bad but need to give operators option to turn these things off if they want 14:18:42 yep 14:19:04 so we'll bring this all together then 14:19:17 tdurakov, my thinking is that live mig should be told how are to try and work out what it needs to do based on what is availble 14:19:45 * mdbooth would like to discuss some less fully formed ideas, such as improving shared storage detection and a move to Tasks 14:20:18 paul-carlton1: yes, but where to store such conf? on computes on controllers? There still room for options 14:20:23 mdbooth: +1 14:20:25 tdurakov: You've been doing work relating to both of the above, right? 14:20:38 yes 14:21:04 there are several patches on review with poc and spec, there 2 more to be done 14:21:08 That's the kind of road-mappy type stuff which we want people to generally understand and be on board with 14:21:44 * mriedem was totally in the wrong channel 14:21:57 mdbooth, tdurakov how much time do you think that would need ? 14:22:16 mdbooth: true, I will not be on Austin Summit, but alaski will be, so he could provide some details i believe 14:22:41 tdurakov: Ah, that sucks. I've noted that I'd really like to do that, but not in great detail. 14:22:53 PaulMurray: might want a different etherpad for the summit session since i currently have https://etherpad.openstack.org/p/newton-nova-live-migration linked into it 14:23:05 which is more detailed than you probably want for a session 14:23:33 mriedem, yes, and the notes on the sessions details page are likely to be inaccurate 14:23:36 tdurakov: If you email me a canonical list of patches I could potentially bone up and advocate for you if you'd like. 14:23:38 by this discussion 14:23:46 mdbooth: sure 14:23:57 mriedem, I can reorganise the etherpad 14:24:05 well, I could start ml for that 14:24:12 mriedem, Its a place holder for now 14:24:12 tdurakov: Sure thing 14:24:30 tdurakov: Pretty sure I'll have read most of them at some point anyway 14:25:11 PaulMurray: maybe off topic, but is the subteam tracking this? https://blueprints.launchpad.net/nova/+spec/migration-state-machine 14:25:26 the owner was asking for it to be re-approved for newton, it'll probably come up in the thursday nova meeting 14:26:04 mriedem, we were tracking that in mitaka 14:26:25 mriedem, its not on the list in thenewton etherpad 14:26:41 I recall having looked at it a while back and approving of the concept at least 14:27:50 mriedem, it was not approved for mitaka, only discussed - so needs proper review 14:27:59 bac to topic 14:28:16 We have two items for the main session: 14:28:32 how to expose post-copy etc 14:28:51 and shared storage detection 14:29:06 That is plenty for one session 14:29:15 And may over spill 14:29:36 there are topics in the neutron/nova session that cover live migration 14:29:51 - possibly cinder too (mutli attach) 14:30:17 mdbooth, do you need to do anything on storage pools ? 14:30:29 or is that just spec review now 14:30:43 PaulMurray: I'm hoping there's nothing controversial in there other than the churn in that specific code. 14:30:56 ok 14:30:58 So I'll likely advocate for it generally 14:31:34 We also need a plan for CI - but that can be covered on the friday session 14:31:51 anything else should go into the unconference sessions 14:31:57 PaulMurray: will it be possible to follow the session online? or get access to some session minutes if not? 14:32:17 luis5tb_, I think the best hope will be etherpads 14:32:24 I don't think there is anything live 14:32:29 ok, thanks! 14:33:37 Note: I will be presenting (along with pkoniszewski and andrearosa at the same time as the weds unconference, so I will see if we can get preference for thrus if there 14:33:46 are topics on the unconference list 14:34:03 our talk will be live migration in mitaka 14:34:07 BTW 14:35:00 #action PaulMurray sort out agenda 14:35:36 #undo 14:35:37 Removing item from minutes: 14:35:46 #action PaulMurray sort out agenda for summit session 14:36:08 #topic Specs 14:36:21 Very little review on the specs at the moment 14:36:31 anything anyone wants to bring up ? 14:37:08 about specs? 14:37:41 tdurakov, yep - there are about 10 on the etherpad with very little review 14:38:12 action- review specs? 14:38:28 yes please - especially with the summit next week 14:38:53 #action sub-team to review specs on https://etherpad.openstack.org/p/newton-nova-priorities-tracking 14:39:16 PaulMurray: what the difference between Unapproved Specs and Work Items in etherpad? 14:40:16 tdurakov, the unapproved specs should really be what is up for review in the priorities-tracking page 14:40:36 acked 14:40:41 in mitaka I ended up with a list of specs approaved and specs not approved at the bottom of the sub-team etherpad 14:40:57 just to remember what should be done and what is not going to be done in this cycle 14:41:18 the list of work items built up with notes all through the cycle 14:41:32 I reorganised it at the summit to reflect the priority for the cycle 14:41:42 decided at the time 14:42:32 For now we should go by the priorities tracking link above 14:43:26 anything else to discuss ? 14:44:10 No meeting next week due to summit 14:44:30 If you are there, say hello - especially if I haven't met you before 14:44:34 :) 14:44:43 thanks for coming 14:44:52 #endmeeting