15:00:42 #startmeeting openstack-helm 15:00:43 Meeting started Tue May 5 15:00:42 2020 UTC and is due to finish in 60 minutes. The chair is gagehugo. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:44 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:00:46 The meeting name has been set to 'openstack_helm' 15:00:56 o/ 15:00:58 #link https://etherpad.opendev.org/p/openstack-helm-weekly-meeting agenda 15:00:58 o/ 15:01:11 \o 15:01:55 Hi all, where do I add a question I wanted to ask? In the agenda there is only 2020-04-28 15:02:16 sorry I failed to update for today 15:02:30 I can ask here, no problem 15:02:45 updated 15:02:51 and here is good too 15:03:39 So the question I have is in regard to the integration of Tungsten Fabric with OpenStack Helm. Namely, the review https://review.opendev.org/#/c/663390/ 15:04:24 Basically, what is required to get integrated and approved exactly? We have our own all-in-one single node deployment which we could easily add to any CI. 15:05:51 let me check portdirect's comments 15:08:35 I believe he just wants some form of testing done on the CI to verify that it works 15:08:45 in zuul I mean 15:08:55 I agree with that. The question is what exactly is required from us 15:09:22 What kind of testing would suffice and how to integrate our stuff? 15:10:45 Adding an experimental check in zuul would be a good first step imo 15:11:29 Yes, but what would that check? And how to add it? 15:13:25 It would verify that the changes in that ps work as intended from an upstream perspective, and to add a check job involves modifying the files under zuul.d/ to add it, where the other checks/jobs live 15:14:14 I believe portdirect also mentioned some documentation on the TF backend as well would be good 15:14:27 Do we have to deploy Tungsten Fabric to check that? In that case it would take some time and a machine (VM) with 32GB memory. 15:15:31 In some way preferably to verify that it works as intended 15:16:01 There is different documentation but I don't think there is any specific for Helm deployment so the question is exactly what kind of documentation is required? For what audience? With what kind of details? About what? Deployment? Production? Development? What is the acceptable bare minimum? 15:16:45 This I understand but as I said, is there a 32GB VM available with Centos 7.x to run the deployment? 15:17:11 And it deploys Helm OpenStack as well so it takes quite a time. What is the acceptable time? 15:19:00 OSH has documentation, if there is existing docs somewhere at the very least linking to that would be acceptable. 15:19:25 Ok. Got it about the documentation. What about the rest of the questions? 15:19:50 We would need to check with infra about that 15:20:03 TF has a hard requirement for centos and 32gb ram? 15:20:32 I see in zuul.d jobs for deployment of various charts but Tungsten Fabric is a neutron/nova/heat networking backend so we'll have to deploy everything, I guess? And also we'd have to pull external git repos for TF and devtools for deployment and the product 15:20:55 Ubuntu is also ok, but less than 24GB is risky 15:21:32 But Ubuntu will work longer because kernel driver gets built on this platform which takes time 15:22:22 hmm ok 15:22:32 So as you see I have whole lot of questions. I tried to contact portdirect in slack but he didn't respond. In order to resolve this quick enough is there any other way to communicate with some of you guys who would help resolving all of these as we go? 15:23:09 Ideally I show what we have and you tell us how to best integrate it 15:23:12 I don't think a lot of us are active in the k8s slack, usually we respond to IRC pings in the openstack-helm channel 15:23:40 I came to openstack-helm 40 minutes ago - it's office hours I guess - nobody responded 15:24:03 office hours aren't really a thing anymore afaik 15:24:11 they might be stuck in a meeting right now 15:24:14 Ok when can I ry?t 15:24:18 "try" 15:24:21 I can ping him later 15:24:42 And what to do about "We would need to check with infra about that"? 15:25:00 I'm not sure what image flavors are available atm 15:25:46 Is there any concept of "unit tests" for helm charts? We've got several small additions to some of them 15:25:49 https://zuul.opendev.org/t/openstack/labels are all the nodes infra has 15:26:02 ^ 15:26:22 That'll do. 15:26:36 So what about "unit" tests? 15:27:08 Most charts have either rally testing scripts or manual verification scripts to check if things are working as intended 15:27:12 lemme find an example 15:28:16 sure 15:28:54 https://github.com/openstack/openstack-helm/blob/master/tools/deployment/developer/common/900-use-it.sh is what the compute gates run to test 15:28:55 And how would it work for pulling external gits for that? 15:30:07 The above link has smoke test for checking openstack deployment. That returns us to the questions - should we deploy whole TF, how do we pull external gits, what about long time for checking, etc 15:30:16 A 15:30:21 you can pull in dockerhub images in the check 15:30:42 Any test which requires correctly deployed openstack is a long story 15:30:48 dockerhub images of what? 15:31:00 Of openstack containers? 15:31:43 That's exactly what we do, yes, but we have tf-devstack in github which does that and that takes quite the time. Up to 1 hour 15:31:57 1 hour to deploy? 15:32:13 devstack takes quite a long time to run in general though 15:32:21 Maybe less. But how much time does Openstack helm deploys as a whole? More than 30 mins 15:32:48 Add TF to it and you'll get somewhere between 30-60 mins. 15:33:15 You can specify the timeout to accommodate that in zuul 15:33:26 the compute check jobs take ~1 hour 15:33:49 and that's all the openstack services sans cinder 15:34:19 So are we ok to add 1 hour more to it? 15:34:33 Or rather in parallel 15:34:58 I would add it and see what it runs 15:35:03 time* 15:36:44 alevineprogmatic: there are 32GB nodes avaialble for tests in OpenDev Zuul CI. However, if possible, try to use several smaller VMs, not all-in-one installation. 15:36:54 So let me sum it up: you suggest to add to zuul.d a job which would git clone our tf-devstack, then invoke it to run deployment of Openstack Helm with TF and then to somehow verify that it's deployed, am I right? 15:37:11 no 15:37:16 not devstack 15:37:17 roman_g, sorry, won't help much. 15:38:26 Ok, then simply helm chart install? We'll have to add some overrides so we'll copy our code. tf-devstack is not Openstack's devstack - it's a part of our test suite. 15:39:25 ah ok 15:39:46 I see now, looking at the repo 15:40:13 Ok. I got you. Now we have something to start with. I guess next review should be for zuul.d 15:40:46 Thank you. 15:41:01 yes, add the appropriate overrides 15:41:41 alevineprogmatic: anything else you want to discuss? 15:41:50 That's all for now. Thanks 15:43:32 #topic open discussion 15:43:48 Does anyone else have anything they want to discuss? 15:47:24 thanks everyone, have a good rest of the week 15:47:26 #endmeeting