09:01:06 #startmeeting blazar 09:01:07 Meeting started Tue Apr 24 09:01:06 2018 UTC and is due to finish in 60 minutes. The chair is masahito. Information about MeetBot at http://wiki.debian.org/MeetBot. 09:01:08 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 09:01:10 The meeting name has been set to 'blazar' 09:01:18 hi all, time to blazar meeting 09:01:23 #topic RollCall 09:02:02 o/ 09:02:09 Todya's agenda is 09:02:18 1. September PTG 09:02:23 2. PlacementAPI support 09:02:30 3. Reservation re-allocation 09:02:32 4. AOB 09:02:35 anything else? 09:02:37 o/ 09:02:43 hiro-kobayashi, priteau: hi 09:02:43 Good morning masahito and hiro-kobayashi 09:02:55 priteau: good morning :-) 09:03:09 hi priteau 09:03:23 bertys and GeraldK told me they can't join today's meeting. 09:03:55 #topic September PTG survey 09:05:04 The official announced the next PTG for Stein will be held in Denver on 10th-14th Sep. 09:05:16 http://lists.openstack.org/pipermail/openstack-dev/2018-April/129564.html 09:06:17 They asked me that we plan to have PTG during this week. 09:07:11 Do you both have a plan to join the PTG? 09:07:33 Yes, I plan to join it. 09:07:47 I can't say yet, it will depend if my employer is happy for me to travel there 09:08:35 They also requested me replaying whether to have or not to have until 2nd May. 09:09:00 I plan to join it, too. 09:09:44 If it's just hiro-kobayashi and you, would you still hold a Blazar meeting there? 09:10:40 I hope new people will join it. 09:10:58 That would be great. 09:11:38 priteau: If it's just us, we can implement new features there instead of discussion. 09:11:49 A hackathon? :-) 09:11:56 yes :-) 09:12:05 Sounds good :-) 09:12:43 I will try to let you know if I can attend as soon as possible, but I might know only weeks from now 09:12:43 I plan to join other project meeting. So I'll be there in either case. 09:13:59 priteau: got it. The early bird ticket will end by 10th May. 09:14:19 I will get a ticket anyway, since refund is possible until end of August 09:14:54 sounds nice. 09:15:07 anything else for this topic? 09:16:33 #topic PlacementAPI support 09:17:13 I heard hiro-kobayashi has something to discuss for the topic. 09:17:22 yes 09:17:41 I created an etherpad for discussing placement support: https://etherpad.openstack.org/p/blazar-placement-design 09:18:12 I wrote requirements and basic approaches down there. 09:18:57 The approach 1, 3 and 4 were basically copied from the Dublin PTG etherpad. 09:19:19 I don't understand much about approach 4, though. 09:19:45 Thank you hiro-kobayashi for creating the document 09:20:08 I think there is also the concept of aggregates in placement API 09:20:35 hiro-kobayashi: Nice works. 09:20:46 priteau: Yes it would be. I appreciate it if you could write your idea down there 09:21:09 And, feel free to edit existing ideas 09:21:47 I think it's better to make a consensus on the basic approach in the Blazar team before Vancouver summit. 09:23:00 I will review and add more details as soon as possible, hopefully this week 09:23:15 I'd like to add some items for each approach. The way of user interaction with Nova will be changed, etc. 09:23:15 Thanks priteau! 09:23:31 Thanks! 09:23:39 Me too. I'll add my comments, too. 09:24:13 After one or two weeks later, we should share this to Placement team. 09:24:44 As hiro-kobayashi said, that should be done before the Vancouver summit. 09:25:12 masahito: Agree. 09:26:49 anything else? 09:27:00 Nothing from me. 09:27:25 #topic Reservation re-allocation 09:28:02 I'm working on this spec now. https://docs.openstack.org/blazar/latest/specs/queens/flavors-extra-specs.html 09:28:59 I want talk about the third work items, Add re-allocation logic to ComputeHostExtraCapability management. 09:30:14 Based on the discussion on gerrit, Blazar needs to re-allocate reservation once the extra_capability is updated, changed or deleted. 09:31:13 However, Host delete API doesn't allow cloud admins to remove hosts if there are reservations to use the hosts. 09:32:29 IMHO, if the Update Host API follows the way of the Delete Host API, the Blazar shouldn't accept the update itself. 09:32:48 One way to resolve the problem would be to prevent extra capability update if the node is reserved 09:33:07 priteau: +1 09:33:28 However, that is quite a strong requirement. A better way to handle it would be to prevent extra capability update only if the node has reservations that are using resource_properties 09:34:09 make sense. 09:34:09 But we could start with the first approach, and then make it smarter 09:34:26 Good approach 09:35:14 Alternative is disabling the host and resource-monitoring re-allocate the reservations. 09:36:18 Will it re-allocate a running reservation? I don't remember the behavior 09:37:11 When the reservation is active, resource_changed flag becomes true at least. 09:37:13 Currently It is not allowed to reallocate if it's active. 09:38:16 masahito: I think blazar should not allow reallocate if it's active except a failure recovery situation. 09:38:47 hiro-kobayashi: IIRC, in addition to resource_changed flag, the missing_resource flag becomes true if there is no alternative resource. 09:39:31 yes 09:40:40 hiro-kobayashi: because it raises complexity of Blazar? 09:41:38 masahito: yes. 09:42:36 got it. 09:42:57 That's from my side. any comment for the topic? 09:43:14 If blazar allow reallocation of active reservation, blazar has to handle running instances on removed hosts. 09:44:13 masahito: I would prefer starting with a simple approach first and increasing complexity later 09:45:04 How about alining with host reservation update? 09:45:07 https://github.com/openstack/blazar/blob/master/blazar/plugins/oshosts/host_plugin.py#L527 09:45:53 I agreed priteau's idea. I plan to start with your suggestion first. 09:46:22 I'd like to talk the goal of the BP. 09:48:01 If the reservation is in pending status, Blazar try to re-allocate. If it's in active, it fails. right? 09:48:51 Even if the status is in pending, the update is not allowed in case of missing_resource is true. 09:49:12 I think that's right. 09:51:01 Make sense. Anyway I'll start with the simple way to do this. 09:51:10 #topic AOB 09:51:27 Does someone have something to share/discuss? 09:51:35 last 10 mins 09:52:31 Nothing from me 09:53:24 I have completed the support for Ironic using Jay's unmerged Nova aggregates patch. The changes to Blazar were quite small. 09:53:49 priteau: Great work! 09:54:30 Good news! 09:55:54 Moving to placement API will probably be a lot more work ;-) 09:56:41 Yes. But that's a really good news. We can move things forward. 09:58:01 priteau: btw, do you have some specific topics in next meeting now? The day of next meeting, 1st May., is holiday in Japan. 09:58:17 Nothing specific from me. 09:58:28 Would you like to skip the meeting? 09:58:43 If nothing and bertys and GeraldK also doesn't have nothing, I want to skip the meeting. 09:58:52 priteau: yes. 09:59:02 No problem 09:59:39 I'll ask them later. If they doesn't have no topics, I announce the skipping in openstack-dev ML. 10:01:17 Running out of time. Nice discussion and greate updates! 10:01:28 Thanks all. bye. 10:01:38 Bye everyone, thank you! 10:01:46 #endmeeting