16:01:29 #startmeeting containers 16:01:30 Meeting started Tue May 30 16:01:29 2017 UTC and is due to finish in 60 minutes. The chair is adrian_otto. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:01:31 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:01:33 The meeting name has been set to 'containers' 16:01:54 #link https://wiki.openstack.org/wiki/Meetings/Containers#Agenda_for_2017-05-30_1600_UTC Our Agenda 16:01:58 #topic Roll Call 16:02:04 Adrian Otto 16:02:10 Spyros Trigazis 16:02:21 Kevin Lefevre 16:02:25 Ton Ngo 16:02:58 hello strigazi ArchiFleKs and tonanhngo 16:04:32 #topic Announcements 16:04:40 1) strigazi , you have one 16:04:59 Magnum is accepted to run jobs on CentOS CI https://bugs.centos.org/view.php?id=13192 16:05:41 I stated working on it. The centos folks are very helpful 16:05:55 excellent! 16:06:02 We will have jobs on physical machines soon I think 16:06:59 So we should see fewer false negative on the job? 16:07:24 tonanhngo Yes, almost none 16:07:50 We should be able to add more complex tests 16:08:06 great, so now we should take the failed job more seriously 16:08:23 Yeap 16:08:55 any other announcements from team members? 16:09:21 #topic Review Action Items 16:09:27 1) adrian_otto to target https://blueprints.launchpad.net/magnum/+spec/integrate-kuryr-kubernetes for Pike, with a comment that it's okay for this to slip to Queens if needed. 16:09:42 I did not complete this because I need to set up a new series in Launchpad. 16:09:58 I don't expect to have time for this until late this week. 16:10:04 I will carry it forward 16:10:08 #action adrian_otto to target https://blueprints.launchpad.net/magnum/+spec/integrate-kuryr-kubernetes for Pike, with a comment that it's okay for this to slip to Queens if needed. 16:10:38 #topic Blueprints/Bugs/Reviews/Ideas 16:10:51 In this section we may discuss any work item 16:11:27 I did some work on upgrades and the use of the heat container agent 16:11:57 On upgrades, I have two work items 16:12:30 One is to to have the etcd data in a cinder volume to be able to upgrade single master clusters, (k8s clusters) 16:13:36 that works now? 16:13:38 And the other is to to upgrade in two steps, Master first and then node. With two different api calls. Very similar to GKE 16:14:38 adrian_otto You can rebuild the vm with a new image or even a new vm and keep the etcd data. 16:15:08 good 16:15:37 GKE has max 5 mins k8s api downtime. We can have something similar. 16:15:55 The application running in the cluster should not have any downtime 16:15:58 yes, that should be fine 16:17:46 any other discussion on work items? 16:18:44 For the heat-agent 16:19:12 It worked, even on the CI. WIP https://review.openstack.org/#/c/468816/ 16:19:48 With this feature we can relief the user_data passed with cloud-init to the vms 16:20:27 That's it from me 16:20:40 what do you mean by "relief"? 16:21:07 adrian_otto We are reaching the limit of 64KB of user_data 16:21:49 We need to reduce the size because we may need to add more custom scripts 16:21:49 do we still pass some data through user_data? 16:21:56 ok, so using the heat agent reduces the size of data we need to put in user data 16:22:14 And we reduce the load on the nova db 16:22:25 adrian_otto yes. 16:22:30 but that load shifts to heat 16:23:07 tonanhngo yes, not all data can be passed with the heat-agent 16:23:16 eg, start the heat agent 16:23:17 ok 16:23:21 or start docker 16:23:37 adrian_otto yes we will increase the load in heat, but 16:23:51 it will be only during creation 16:24:26 if we don't expect to update that data, the heat agent won't poll heat. AFAIK 16:25:16 Note the action, CREATE https://review.openstack.org/#/c/468816/1/magnum/drivers/k8s_fedora_atomic_v1/templates/kubemaster.yaml@478 16:26:04 we will need it to poll in case of conducting upgrades 16:27:27 I think if it has action CREATE, we don't need to. I will check 16:28:00 I think it's still okay even if it does poll heat for upgrades 16:28:16 btw, if a deployment has zaqar available. The can use the push model. 16:28:26 because you can change your heat transport to Zaqar to eliminate that if you need to support a huge number of clusters. 16:28:48 we should put that in our docs though as an anticipated scalability issue 16:28:52 Anyway the polling time can be large in the case of upgrades 16:28:58 yes 16:29:23 we can make the interval configurable as well 16:30:00 it configurable in heat I think but we can have a value in magnum as well 16:33:59 #topic Open Discussion 16:34:27 feel free to raise any topic here for discussion 16:34:52 I'm good 16:35:05 me too 16:35:16 ok, any objections to wrapping up now? 16:35:29 sure 16:35:46 Thanks Spyros for great update 16:36:04 Our next team meeting will be on 2017-06-13 at 1600UTC in #openstack-meeting-alt 16:36:04 See you then!! 16:36:05 thanks for attending everyone 16:36:12 #endmeeting