13:00:36 #startmeeting monasca 13:00:37 Meeting started Tue Jul 21 13:00:36 2020 UTC and is due to finish in 60 minutes. The chair is chaconpiza. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:00:38 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:00:40 The meeting name has been set to 'monasca' 13:00:56 Hello 13:01:02 hi 13:01:06 Hello 13:01:25 hi 13:01:28 Hi Witek 13:01:32 Hi Matthias 13:01:36 Hi Adrian 13:01:39 The agenda as usual https://etherpad.opendev.org/p/monasca-team-meeting-agenda 13:02:12 Before to start with the agenda let's go for the previous action items 13:03:16 Thanks for the changes for "Set legacy_kafka_client_enabled = False on default" 13:03:48 they are being merged 13:04:37 I guess I have to recheck https://review.opendev.org/740954 13:06:08 Zuul didn't start the gate job? 13:06:14 no 13:07:08 now, rerunning check job 13:07:22 Ok, 13:07:55 the next action item is related to https://review.opendev.org/#/c/722515/5 13:08:25 It fails because in Victoria py38 is voting 13:09:03 So, I removed the testcase that brakes it https://review.opendev.org/#/c/742167 13:09:12 and rebased on top of it 13:09:54 fine with me if that works 13:09:55 Nevertheless it is taking too long the py38 check. Something is wrong there. 13:10:05 I will check it 13:11:01 The next action from last week came from Luigi: https://github.com/openstack/monasca-transform/blob/master/.zuul.yaml#L10 13:11:59 it's mine, no progress yet 13:12:31 witek_, I am not sure how to removed it, whether all file or only lines 10-14 13:12:51 I'll take care 13:13:04 Great, thanks 13:14:04 The next action from two week ago is about the migration of CI/CD jobs to new Ubuntu LTS Focal. I will focus on this task this week. 13:15:04 Then let's start with today topics... 13:15:23 OK 13:15:26 #topic Issues with monasca-agent (Ussuri) 13:15:34 Thanks Matthias 13:15:40 As described in agenda, just short overview here: 13:16:04 I started to test with monasca-agent for Ussuri and came across a few issues. 13:16:40 When testing this with devstack (thanks, Martin and Witek), http_check was not running at all 13:17:05 The additional error in devstack (wrong config) has been fixed already. 13:17:28 However, finally my conclusion was: agent testing is not sufficient. 13:18:22 See 1.3 for some initial proposals. 13:18:42 *it is merged in Master ^^ and it is being merged the cherry-pick on Ussuri, Train is ok. 13:19:02 I forgot: Thanks, Martin, for providing the fix in devstack 13:19:59 from my point of view, we could add two kinds of tests 13:20:31 * unit tests - they should catch regexp issues with Python 3 13:21:35 * functional tests - checking that provided configuration provides expected metrics on a given OpenStack deployment 13:22:23 Yes, agreed. Currently, the only Openstack service installed is keystone, I think 13:23:01 in our CI, yes 13:23:04 ^^The only one OpenStack service using the vagrant file from monasca-api repo 13:23:06 yes 13:23:57 it should be enough for the test 13:24:14 Maybe, I don't know, should be checked 13:24:33 we could check if running any tests with Kolla would be possible 13:25:36 chaconpiza: optimally all core services should be tested 13:26:33 Yes, I see. 13:27:25 Other issue is causing troubles in our new deploy is reading from /proc 13:27:44 Matthias do you want me to introduce the topic? 13:28:28 Let's quickly finalize the testing topic. How to continue? Action item? 13:28:37 alright 13:29:08 My proposal: 13:29:40 I could work on the specification (as a first step). But not in the next 2-4 weeks. 13:30:09 Witek, could you explain your idea with "tests with Kolla"? 13:30:16 I don't understand this 13:30:23 Not necessarily now 13:30:45 I was wondering if there's any Kolla CI job installing OpenStack including Monasca 13:31:05 Ah, ok. Could you check this part? 13:31:23 if yes, we could write simple tests checking if API can return expected metrics at the end of the run of such CI job 13:31:53 we could do the same with DevStack, but it's more far away from real deployment 13:32:07 Yes, I agree 13:33:20 the first to start with, should be just adding unit tests 13:35:10 Yes. As said, I can work on this, but not in the next cpl. of weeks. 13:36:00 do we have a bug report for this? 13:36:23 #action I will create the story about it and add it to the board 13:36:47 affected branches are master and ussuri 13:39:12 Of course, I will fix the bugs I detect. I mainly brought it up here to discuss the lack of testing 13:40:53 reporting the bugs helps to share the effort, someone can pick it up 13:42:58 Alright lets move to the next part of the monasca-agent, the time is running away... ;) 13:43:05 ok 13:43:13 We noticed that in an OpenStack deployment where the services run bare-metal we end-up with users and groups created on the operating system, like cinder, keystone, etc. 13:43:36 those users are the owners of some directories in /proc 13:43:58 Then the mon-agent user (with the sudoers permissions) can read the info from /proc and generate the associated metrics. 13:44:17 But in a deploy where the OpenStack services run on containers 13:44:34 we found in /proc that there are directories without user/group but only with user ID. 13:45:16 The mon-agent can't read them even with sudoers permissions. 13:46:11 Is inside of each container where the users/groups are created, like cinder for example. 13:46:44 We are wondering whether in the deploy with kolla-ansible they have a similar situation 13:46:45 Some additional info: the numeric user-id that can be seen on the host is the user-id of the related user in the container (e.g. cinder). But this user doesn't exist on the host 13:46:58 OK, sorry, you wrote it alread :-) 13:47:06 ;-) 13:47:49 is it the user ID mapping issue, or general lack of permissions inside the container? 13:48:14 No, the issue is that the user doesn't exist on the host. 13:48:49 right, in the /proc directory inside the containers it is ok 13:50:13 The agent is reading from /proc, and I wouldn't like to change this... 13:50:59 (reading from the /proc in the bare-metal host) 13:53:23 One step forward maybe is to deploy at least once using Kolla-Ansible to get familiar with that. 13:53:51 Is not the same like TripleO, but both share the images 13:54:33 that would be good in any case 13:55:22 #topic AOB 13:55:54 we still have few minutes for any other topic 13:56:07 i have one small topic 13:56:37 i created stable/pike branch in monasca/monasca-agents-installer repository 13:57:59 ^^ Because of specific changes using only master is not enough to create agents for the different branch releases 13:58:17 i created this branch because we still need to suport old version of monasca-log-agent installer base on old logstash version 13:58:43 makes sense 13:59:23 Thanks for bringing the topic Adrian 13:59:32 Lets close the meeting 13:59:48 Thanks for coming today 13:59:55 Bye 14:00:08 bye 14:00:10 thank bye : ) 14:00:14 thanks, bye 14:00:17 #endmeeting