14:02:05 #startmeeting nova 14:02:06 Meeting started Thu May 28 14:02:05 2015 UTC and is due to finish in 60 minutes. The chair is johnthetubaguy. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:02:07 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:02:09 The meeting name has been set to 'nova' 14:02:18 \o 14:02:21 o/ 14:02:23 o/ 14:02:24 o/ 14:02:29 o/ 14:02:30 o/ 14:02:30 o/ 14:02:32 o/ 14:02:32 o/ 14:02:32 o/ 14:02:34 o/ 14:02:35 o/ 14:02:35 o/ 14:02:36 #link https://wiki.openstack.org/wiki/Meetings/Nova#Agenda_for_next_meeting 14:02:40 o\ 14:02:42 hello all 14:02:42 o/ 14:02:52 #topic Release Status 14:02:52 o/ 14:03:11 so there are lots of proposed deadline we discussed at the summit in the process session 14:03:15 * mriedem joins late 14:03:20 basically same as kilo, is the rough summary of that 14:03:30 they are in the agenda I liked above 14:03:47 liberty-1 being spec freeze, but we keep the backlog open 14:03:54 thats the one thats coming close 14:04:07 rather than wasting time, I will get a proposal sent to the ML 14:04:20 johnthetubaguy: cool, that answer my question 14:04:23 but does anything in the list in the agenda wiki look totally off to people? 14:04:33 garyk: :) 14:04:45 so moving on then 14:04:46 #link https://etherpad.openstack.org/p/liberty-nova-priorities-tracking 14:04:57 we have an etherpad of doom for liberty 14:05:13 we are trying out the idea of subteams suggesting their top 3 or top 5 patches 14:05:26 coolness 14:05:31 again, there will be more details on whats happening on the ML soon, but thats the net net 14:05:41 its mostly all written down in the above etherpad 14:06:07 in future meetings, we can raise any stuck items listed in the etherpad I guess, but lets keep moving for now 14:06:36 there is a stuck spec review, but its not actually got a -1, so I am ignoring that: https://review.openstack.org/135387 14:06:44 there is one non-spec blueprint to discuss 14:06:49 #link https://blueprints.launchpad.net/nova/+spec/config-drive-support-for-virtuozzo 14:07:13 basically add config drive support to parallels, its almost a bug fix 14:07:22 any objections to approving that one? 14:07:36 thanks for bringing this up 14:07:53 it's not a bug fix, let's be clear :) 14:07:58 mnestratov: no worries, it got on the agenda, I think mikal put it there during the summit 14:08:01 but yes seems like an obvious thing to do 14:08:26 dansmith: yes, very true, its a "me too" feature that got left out in the initial run of parallels support patches 14:08:34 yes 14:08:42 are you wondering if we can do it without a spec? 14:08:52 oh, sorry, yes, thats the thing here 14:09:09 do we approve this without a spec, as proposed in the meeting agenda by mikal 14:09:15 if it doesn't impact anything other than just driver support then I think that's fine 14:09:29 dansmith: cool, I am thinking the same thing, it looks that way 14:09:38 there was another proposal for a feature parity thing affecting ceph that was similar, but it affected glance usage and ceph setup, 14:09:40 but we had a general spec for the previous release cycle 14:09:53 and thus I opined in favor of a spec because of the cross-service impact 14:09:55 mnestratov: you mean for the parallels support in general? 14:10:05 dansmith: good call out, +1 on that 14:10:12 johnthetubaguy: yes 14:10:14 so if this is all driver-contained, then it seems reasonable to just use a BP for tracking to me 14:10:19 +1 14:10:23 any objections? 14:10:47 +1 14:10:56 #info https://blueprints.launchpad.net/nova/+spec/config-drive-support-for-virtuozzo is approved for liberty without the need for a spec 14:11:05 I think thats all updated now 14:11:06 thanks 14:11:21 mnestratov: please mark it as "NeedsCodeReview" once all the code is up for review 14:11:29 mnestratov: I guess you can do that right now? 14:11:36 johnthetubaguy: sure 14:11:43 thanks 14:11:46 so moving on 14:11:56 #topic Events: design summit and midcycle 14:12:10 so I am going through all the etherpads looking for actions 14:12:16 #link https://etherpad.openstack.org/p/YVR-nova-liberty-summit-action-items 14:12:23 one action item should be a mid cycle in europe 14:12:37 * bauzas should think about renaming himself :) 14:12:53 +1 14:12:54 mid cycle in europe +1 14:12:59 but its been delayed as we are attempting to deploy new code in our SYD region, and that being quite rough going… 14:13:02 honestly +1 to it 14:13:28 +1 14:13:36 for europe 14:13:43 so for a european midcycle, I worry about diluting the US based one we already agreed would happen now 14:13:53 yep 14:13:55 what about doing it during the M release? 14:14:04 well, pencilling one in for then 14:14:12 sounds good to me 14:14:20 fair point 14:14:20 makes sense to ask about a blueprint approval here? 14:14:26 if someone has office space for that, do get in touch, so we can try sort that sound 14:14:27 sahid: open discussion 14:14:50 sahid: we did that bit already, stuff needs to go on here: https://wiki.openstack.org/wiki/Meetings/Nova#Agenda_for_next_meeting but yeah, we can do that in Open now 14:15:12 mriedem, johnthetubaguy ah ok thanks i was litle bit late 14:15:15 #action johnthetubaguy to document the idea of a european midcycle next time 14:15:18 https://blueprints.launchpad.net/nova/+spec/libvirt-set-admin-password 14:15:25 a really small one 14:15:28 sahid: no worries, lets get back to that in a bit 14:15:32 ok 14:15:46 Ok, so lets keep moving 14:15:48 warm Europe for M midcycle, please :) 14:16:08 not northern finland in winter? 14:16:09 edleafe: depends who gives us office space 14:16:12 but Bristol is beautiful on June 2nd-3rd! 14:16:27 anyways, lets move on 14:16:28 Grenoble has quite lots of snow on the slopes by January... :D 14:16:43 feel free to start an ML thread on that, ideally with real offers of space 14:16:58 #topic Community Building 14:17:15 so this is partly a place holder to track these efforts 14:17:28 mentoring being a big part of what was discussed around scaling out the nova team 14:17:35 #link https://etherpad.openstack.org/p/YVR-nova-liberty-process 14:17:47 but little updates for this week 14:18:02 except, I wanted to do a survey to see if people are still happy with these meeting times 14:18:20 the survey will close the next time we have a meeting at this time 14:18:23 #link http://doodle.com/eyzvnawzv86ubtaw 14:18:49 OK, so moving on 14:18:53 #topic Bugs 14:19:10 I notice our bug tag owners is getting stale : https://wiki.openstack.org/wiki/Nova/BugTriage 14:19:22 it would be good to fix that, get folks to step up 14:19:27 agreed 14:19:29 * johnthetubaguy wink wink, nudge nudge 14:19:38 mriedem: hows the gate these days? 14:19:47 fine for nova on master 14:20:08 https://bugs.launchpad.net/nova/+bug/1457527 14:20:08 Launchpad bug 1457527 in OpenStack Compute (nova) "Image-cache deleting active swap backing images" [Critical,In progress] - Assigned to Michael Still (mikalstill) 14:20:19 https://bugs.launchpad.net/nova/+bug/1447380 14:20:19 Launchpad bug 1447380 in OpenStack Compute (nova) "wrong cinder.conf.sample generation: missing directives for keystone_authtoken (at least)" [Critical,In progress] - Assigned to John Griffith (john-griffith) 14:20:34 so they are both critial bugs right now 14:20:46 how is that second one critical? 14:20:49 ndipanov was talking about the first yesterday, there is a patch https://review.openstack.org/#/c/83516/ 14:21:01 the second is not crticial 14:21:02 dansmith: I don't think it is really, its just marked that way 14:21:20 i moved it to medium 14:21:29 cool 14:21:32 mriedem, but that will not fix it for existing instances 14:21:42 I started looking into that but it's not sooo easy 14:21:54 if we want to cover all cases (migration etc) 14:21:55 can we fix it in the list object? 14:21:56 Is the openstack bootstrapping hour still a thing? I liked those. Could they do a session about bug triaging? 14:22:04 mriedem: oh me too, but thats cool 14:22:25 markus_z: feel free to help, we have 40 open bugs even when triaging once per week 14:22:25 markus_z: need topics and presenters 14:22:44 Nova is generating app. 20 new bugs per week 14:22:46 markus_z: sdague generally runs those so try to find him later 14:23:05 cools, I think sdague will be back online next week 14:23:07 half of them are invalid from my perspective 14:23:30 bauzas: It like to, but I had issues to understand the process the last time I read the wiki 14:23:33 #action markus_z and mriedem to talk to sdague about bug triage and the bootstrapping hour 14:23:42 oh i see what you did there 14:23:47 markus_z: ping me and I'll be glad to help 14:23:54 bauzas: thanks 14:24:02 johnthetubaguy: thanks, I guess ;) 14:24:18 anything more on that critical bug? 14:24:39 probably best to take that into #openstack-nova after we are done at this point 14:25:02 #topic Open Discussion 14:25:12 so we have a stuck review listed, but its not got a -1 on it 14:25:18 just about bugs, the trivial bugs list is up-to-date 14:25:21 feel free to read the agenda for the link 14:25:31 we need to define stuck reviews on that page, apparently 14:25:52 #action johnthetubaguy to define "stuck reviews" on the agenda page 14:25:54 i.e. "stuck != unreviewed for N days" 14:25:56 yeah 14:26:10 unless N > 180 or something 14:26:14 but anyways 14:26:26 mriedem: min libvirt version 14:26:29 could we rename that to 'conflicting reviews' ? 14:26:30 that's different 14:26:32 yeah 14:26:38 johnthetubaguy: the min libvirt version thing was old, removed it 14:26:39 it was supposed to be to break ties 14:26:56 "Reviews requiring mediation" 14:26:57 mriedem: gotcha, probably my bad, I thought we sorted that already 14:27:18 i did add this https://blueprints.launchpad.net/nova/+spec/aggregate-network-filter 14:27:22 lxsli: I quite like "mediation" actually 14:27:24 does that require a spec? it's a new scheduler filter 14:27:33 bauzas: ^ 14:27:56 mriedem: the problem is that it needs to provide more to the scheduler than what's already in HostState 14:28:30 mriedem: here the proposal is to use filter_props, but since I'm working on RequestSpec, I'm a little worried by any impact it would cause 14:28:36 bauzas: oh, like it sends not just a list on instance uuids, but also networks? 14:29:01 johnthetubaguy: yup, it adds the networks list as part of the big filt_props blob 14:29:43 johnthetubaguy: so that's not technically bounded to what the Scheduler proposes 14:29:58 and that opens room for discussion 14:30:05 bauzas: are you happy to write up a dev ref on how this is happening (or spec) in general? 14:30:19 bauzas: and maybe describe what sorts of features are blocked because of that work? 14:30:21 johnthetubaguy: yeah, that's an action I have to do (and you marked it as well :) 14:30:59 johnthetubaguy: yeah, put my name on that 14:31:06 bauzas: I think we need to just do this one first, explain why we can't do this right away, and what effort to help with to get there quicker 14:31:16 johnthetubaguy: totally agreed 14:31:28 johnthetubaguy: or find a mitigation plan 14:31:30 #action bauzas to describe what is blocking https://blueprints.launchpad.net/nova/+spec/aggregate-network-filter and how to help 14:31:45 bauzas: do we want a spec for the above blueprint, because of this? 14:32:16 johnthetubaguy: that's harsh to say, but I would at least be able to discuss on the format of that, since it needs to be backported to the RequestSpec object 14:32:44 johnthetubaguy: so I really would like to avoid non-typed, unformal dicts added to the already non-typed, unversioned dict that we have 14:33:01 sounds like a spec to me 14:33:16 +1 on a spec 14:33:46 OK, needs a spec as it needs new info 14:33:54 well, at least we need to make sure that any change to filter_properties or request_spec is enough documented 14:34:02 so yeah for a specc 14:34:04 spec 14:34:05 #info https://blueprints.launchpad.net/nova/+spec/aggregate-network-filter needs a spec because of the extra information that needs to be passed around 14:34:17 sahid: you had a spec as well right? 14:34:22 I mean, blueprint 14:34:47 johnthetubaguy: you mean the one i told before ? 14:34:51 yep 14:35:06 johnthetubaguy: https://blueprints.launchpad.net/nova/+spec/libvirt-set-admin-password 14:35:28 the last relase of libvirt provide the super to configure user password 14:35:45 sahid: so this is libvirt implementing the API xenapi already has? 14:35:50 and so the aim of the blueprint is to use it to implement the set_admin_password PAI 14:36:02 I guess this will require an in guest agent? 14:36:48 i see danpb has +2ed all of the changes w/o the blueprint being approved 14:36:56 hum... 14:37:00 what do you mean ? 14:37:22 https://review.openstack.org/#/c/185910/ 14:37:25 johnthetubaguy: yeah, I would think it would, if it's going to work online like the xenapi one does (right?) 14:37:55 sahid: the two patches listed in the whiteboard already have +2 reviews from Daniel Berrange 14:38:11 so… sahid its not clear to me from the blueprint how it works 14:38:13 cfriesen: yes but we need to bp approval no? 14:38:20 which seems problematic 14:38:43 ok so you need specific detail about how it works internally on libvirt? 14:39:03 sahid: I need to be able to tell if it will appear the same to users as the xenapi support 14:39:03 sahid: are the dependent patches in that series for the guest object tied to that blueprint as well? 14:39:15 sahid: is this supposed to work live, or only if the guest is off (or rebooted as part of this process) ? 14:40:32 OK… so given there are questions 14:40:35 we need a spec for this 14:41:09 at least its the only other way we have of agreeing 14:41:18 johnthetubaguy: hum so the support is only for qemu 14:41:22 I have -2ed the patches since the blueprint is not agreed 14:41:47 sahid: this is using the existing set_admin_password() call? 14:41:51 sahid: yeah, we need a spec to talk about all the details, it seems very unclear about what is going on I am afraid, like what image properties you need to make sure the API works, etc, etc 14:41:54 mriedem: well i have considered it was yes 14:42:44 cfriesen: yes 14:42:57 #action sahid to submit a spec for the blueprint https://blueprints.launchpad.net/nova/+spec/libvirt-set-admin-password as its not clear how a user can use the feature from the spec, and it not clear its the same as the existing xenapi feature 14:43:15 johnthetubaguy: there is no need of image property actually this use a nova API already present 14:43:21 johnthetubaguy: ack 14:43:32 Is there a libvirt subteam? I see there are a lot of refactorings ongoing. It's not listed in the etherpad "liberty-nova-priorities-tracking" 14:43:36 sahid: so does this API work with all images that currently exist? 14:43:57 markus_z: there used to be and had meetings but attendance dropped so danpb disbanded 14:44:05 the API (nova)is actually not really useful since we do not pass the user 14:44:11 we consider we are using linux 14:44:13 so root 14:44:26 but we can have image with Windows 14:44:31 well, it is setting the admin password, 14:44:32 mriedem: OK 14:44:37 so that is why i have added the thing with image property 14:44:40 which would be root on linux or administrator on windows 14:44:48 sahid: the API resets root on linux, and administrator on linux, using an in guest agent, right now, anyways 14:44:58 oops 14:45:02 administrator on windows 14:45:05 johnthetubaguy: and live without a reboot, right? 14:45:10 dansmith: yes 14:45:14 dansmith: yes 14:45:36 sahid: so this will require a qemu agent to be running in the guest to provide the same functionality, right? 14:45:36 sahid: so how does it do it live? using an in guest agent right? 14:45:46 the image property things was for image with an other user than Administrator 14:45:53 johnthetubaguy: yes 14:46:00 i'm looking at the patch rieh now 14:46:06 sahid: so the API should return not implemented, or similar if there is no guest agent installed, like the xenapi one does? 14:46:33 johnthetubaguy: yep i did not add the condition, this change is only for semu 14:46:36 qemu 14:46:59 sahid: so thats inconsistent, feels like we need these details discussed in the spec 14:47:18 johnthetubaguy: do you think i can just try to enhance the blueprint? 14:47:27 no 14:47:29 this needs a spec 14:47:33 then if that looks not good i will do the spec 14:47:41 right, there is too much discussion here for the blueprint to work 14:47:42 +1 for spec 14:47:52 ok so go for the spec 14:47:57 thank you guys 14:47:58 i also want to know about testing since we have a newer libvirt in the fc21 job but i don't think it's this new 14:48:02 #info agreed this really needs a spec https://blueprints.launchpad.net/nova/+spec/libvirt-set-admin-password 14:48:06 mriedem: +1 14:48:29 mriedem: yeah 14:48:37 OK, cools 14:48:41 any more for any more? 14:48:42 mriedem: this seems to add another moving part, so would really like to see it *actually* tested for sure 14:48:45 we would like to make our pcs6 CI system comment on kilo/stable https://wiki.openstack.org/wiki/ThirdPartySystems/Parallels_PCS6_CI 14:48:46 johnthetubaguy and community members please provide your valuable feedback on nova-specs, https://review.openstack.org/135387 14:49:20 abhishekk: its in the review queue, its going to take a little while to get through the backlog post summit 14:49:26 mnestratov: why stable? 14:49:29 why not master? 14:49:45 johnthetubaguy: thank you for the update 14:50:11 let's just agree that we'll never do anything about the performance of the unshelve API :) 14:50:28 early mark? 14:50:38 actually we are in the process to building next version of our product for master 14:50:59 mnestratov: let's talk about it in -nova 14:51:00 so master a bit later with a next version 14:51:05 ok 14:51:14 cool, so we are done then 14:51:15 thanks all 14:51:25 #endmeeting