09:00:20 #startmeeting magnum 09:00:21 Meeting started Wed Jul 1 09:00:20 2020 UTC and is due to finish in 60 minutes. The chair is flwang1. Information about MeetBot at http://wiki.debian.org/MeetBot. 09:00:22 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 09:00:24 The meeting name has been set to 'magnum' 09:00:43 brtknr: seems just you and me 09:00:51 so i assume this could be a short meeting 09:01:43 i saw you added a topic 'build hca in ci', would you like to talk more about it? 09:01:46 flwang1: just saw an email from spyros 09:03:06 yep, just replied 09:05:01 brtknr: let's continue? 09:05:51 sure 09:06:18 flwang1: so hca stable tags are currently manually tagged 09:06:27 yes 09:06:28 it would be good to do this via CI 09:06:38 we discussed this a few weeks earlier 09:06:41 brtknr: do you have any idea to do that? 09:06:56 brtknr: ah, i remember now 09:07:01 https://review.opendev.org/#/c/735100/ 09:07:40 Bharat Kunwar proposed openstack/magnum master: [hca] Only build/push stable images if unpublished https://review.opendev.org/735100 09:10:36 but this one can't resolve the issue that we have to manually tag hca version for stable branch, is it? 09:13:27 brtknr: ^ 09:14:13 flwang1: thats right 09:15:00 but a stable tag doesnt get built if it already exists upstream 09:15:11 thats the key difference 09:15:12 i understand 09:15:25 i'm ok with this change 09:18:28 Cool 09:18:39 Lets talk about your issue 09:18:55 what is allow_cross_minor_version_upgrade? 09:19:03 flwang1: 09:19:33 recently, i'm working on a doc about versions 09:19:35 for k8s 09:19:58 see https://github.com/openstacker/catalystcloud-docs/blob/c246d02362d343b3ae024730b40f59177e17fb4c/source/kubernetes/versions.rst 09:20:31 though i know Magnum does support upgrade and skip minor versions 09:21:26 but we (Catalyst Cloud) would like to ask user do not skip minor version to avoid surprise (may break api compatibility ) 09:21:51 and one of our reviewer was asking if we can enforce it by the service 09:22:28 hence why i'm thinking if we can support it and leave the option to the cloud admin if they would like to enable this enforcement 09:22:35 brtknr: does that make sense? 09:24:27 very informative flwang1 :) 09:24:57 jakeyip: are you talking about the doc? ;) 09:25:08 yeap 09:26:02 jakeyip: thank you 09:26:25 our public cloud users are seriously care the version support policies 09:26:29 hence why we need it 09:27:52 how are you testing upgrades? 09:28:05 we have a pipeline running on gitlab 09:28:36 when we publishing a new template, the pipeline will make sure the new version is upgradeable 09:28:44 from the n-1 version 09:28:52 jakeyip: ^ 09:29:24 flwang1: I see. does the pipeline spins up a new cluster with n-1 and upgrade it? 09:29:41 jakeyip: yes 09:29:51 with our public template 09:30:00 brtknr: any thought? 09:31:20 flwang1: sorry home distractions 09:31:26 just catching up 1 sec 09:33:16 flwang1: sounds good to me, shouldnt be too controvertial if its opt-in 09:33:42 although I slightly prefer the upgradable template proposal 09:33:56 which would also cover this use case 09:34:14 you mean add a new field to cluster template? 09:34:31 e.g. when you create a new template, you can specify a list of templates that you users can upgrade from 09:34:35 flwang1: thats right 09:34:56 this is more similar to other public clouds 09:35:28 brtknr: yep, the only thing is, that one will only work for public templates managed by the cluster admins 09:35:30 it seems inconvenient to force users to upgrade every minor version if there is no reprecussion 09:36:19 flwang1: surely you can only guarentee SLA for public templates managed by cluster admins 09:36:19 brtknr: AKS (and IBM?) doesn't allow skip minor version 09:36:46 flwang1: maybe I am misunderstanding 09:37:01 brtknr: no, you're right 09:37:33 you mean prevent users from going from 1.14.1->1.14.3? or going from 1.14.1->1.16.1 09:37:36 i know the 'allowed_upgrade_from' field can cover this 09:37:47 brtknr: the later 09:38:03 1.14.1 -> v1.14.3 is just patch version 09:38:06 flwang1: i agree that 1.14.1->1.16.1 should not be allowed 09:38:09 we're talking about minor version 09:38:22 major.minor.patch 09:38:29 yep sorry i got confused 09:39:00 i just think a basic check is much easier than a db schema change 09:39:28 i will think about this again and propose a patch later 09:40:10 flwang1: you are right 09:40:29 i dont think we even need a config option 09:40:46 i am happy to make it the magnum default behaviour 09:41:06 brtknr: it's good know you support this idea 09:41:25 TBH, i believe it will make our support life much easier 09:41:27 and lets backport to ussuri 09:41:56 flwang1: btw do you use terraform with magnum? 09:42:16 brtknr: one of our large customer use terraform 09:42:42 and recently i'm going to contribute terraform to support the --merge-labels 09:42:52 flwang1: btw how many clusters do you host? 09:42:57 i just get it done in gophercloud 09:43:05 flwang1: nice! 09:43:23 not sure if i can share the info, sorry 09:43:46 flwang1: ah no worries 09:45:12 brtknr: we do have some serious customers are using our k8s service 09:45:48 that's why we're seriously caring some features like vesions support, upgrade, ca rotate 09:46:04 some of them, private clouds may not really care 09:46:25 flwang1: https://github.com/gophercloud/gophercloud/commit/f4f6cc9e28463ff34984d79089c5985b5b8df85c#diff-6134db4c32ede6a964dced2d7195db8fR36 09:46:45 does that mean merge-labels is true by default 09:47:13 Ah oops that is just a test case 09:48:02 by default it's empty, which means it will be skipped i think 09:48:35 i haven't got time to finish the work in terraform 09:48:47 anything else you guys want to discuss? 09:50:43 brtknr: jakeyip: shall we call this one done? 09:51:00 i'm going to close this meeting, thanks for joining 09:51:02 nothing from me :) 09:51:18 #endmeeting