15:00:25 <lennyb> #startmeeting third-party
15:00:26 <openstack> Meeting started Mon Mar 13 15:00:25 2017 UTC and is due to finish in 60 minutes.  The chair is lennyb. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:28 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:00:30 <lennyb> Hello
15:00:31 <openstack> The meeting name has been set to 'third_party'
15:00:54 <mptacekx> Hello
15:01:08 <lennyb> hi mptacekx
15:01:21 <mptacekx> hi lennyb, how it's going ?
15:01:33 <pots> o/
15:02:03 <lennyb> well, it's already 5pm here, so it's going pretty good :)
15:02:06 <lennyb> hi pots
15:02:17 <mptacekx> I have two topics to discuss today, one is a kind of problem sharing and the other one is a kind of request for help :-)
15:02:24 <lennyb> I have an issue with the jenkins nd nodepool
15:02:35 <lennyb> I see that sometimes Jenkins starts job on the VM, that was already used and currently 'deleting' by the nodepool. I've added quiet-period to the job, but it still annoying.
15:02:36 <asselin_> hi
15:02:43 <lennyb> hello asselin_
15:03:02 <pots> lennyb i have the same issue
15:03:09 <mmedvede> o/
15:03:21 <lennyb> pots, any ideas how to solve it?
15:03:25 <mptacekx> lennyb: nodepool should delete the node after it's finished with the first job run
15:03:27 <pots> does the quiet-period work?
15:03:48 <lennyb> pots, it works fine, I hate spending 30sec in each job
15:04:08 <asselin_> there's supposed to be a trigger that takes the node offline...in zuul
15:04:36 <pots> where's the quiet-period documented?
15:04:54 <asselin_> http://git.openstack.org/cgit/openstack-infra/project-config-example/tree/zuul/openstack_functions.py#n32
15:05:09 <asselin_> http://git.openstack.org/cgit/openstack-infra/project-config-example/tree/zuul/layout.yaml#n58
15:05:19 <lennyb> pots,https://docs.openstack.org/infra/jenkins-job-builder/definition.html
15:05:32 <lennyb> asselin_, thanks
15:05:55 <pots> lennyb: ah, thanks
15:06:14 <lennyb> pots, I will try asselin_ suggestion, seems like the correct one.
15:06:47 <mptacekx> is anyone from you running ci infra (zuul, nodepool) on xenial ?
15:06:56 <lennyb> anything else on this topic, before going to mptacekx?
15:07:20 <mmedvede> mptacekx: I never switched from trusty
15:07:34 <asselin_> me 2
15:07:35 <pots> i am running trusty but about to try xenial
15:07:43 <mmedvede> mptacekx: I'll probably go to xenial with zuul v3 switch
15:07:53 <lennyb> mptaceks, me neither. What is the issue with xenial?
15:08:08 <mptacekx> we're running mitaka cloud on trusty for intel-nfv-ci, but we want to upgrade to newton and see some strange stuff by having newton kolla cloud on trusty
15:08:27 <mptacekx> for ci infra deployment we're using https://github.com/openstack-infra/puppet-openstackci but it looks like there is no xenial support yet
15:09:10 <mptacekx> the tricky part is how to deploy zuul, nodepool, jenkins, all together in some smart way
15:09:19 <asselin_> mptacekx, yeah, I don't think upstream has 'upgraded' yet.
15:09:34 <asselin_> mtreinish, there's a single_node_ci class that does exactly that
15:09:55 <asselin_> http://git.gozer.hpcloud.net/cgit/openstack-infra/puppet-openstackci/tree/manifests/single_node_ci.pp
15:10:02 <asselin_> mptacekx, I mean ^^
15:10:15 <mptacekx> asselin_: correct, we used this one
15:10:50 <mptacekx> but trusty is mentioned in https://github.com/openstack-infra/puppet-openstackci/blob/master/doc/source/third_party_ci.rst
15:11:18 <asselin_> mptacekx, yeah, that was what it was developed/tested on
15:11:41 <mptacekx> and some modules like mysql can't work with systemd
15:11:43 <mptacekx> ESC[1;31mError: /Stage[main]/Mysql::Server::Service/Service[mysqld]: Provider upstart is not functional on this hostESC[0m
15:12:23 <mptacekx> isn't mitaka eol 09/17 ? if not xenial, are all of you using mitaka cloud still ?
15:12:58 <lennyb> mptacekx, we use cloud based on RDO latest
15:13:02 <mmedvede> we are using mitaka. Nothing wrong with using something eol for internal cloud
15:13:08 <asselin_> mptacekx, that is a separate issue....the ci infrastructure runs on it's own images...trusty, centos 7, etc....as long as the cloud can provide the requested OS.
15:13:24 <mmedvede> upgrading to new openstack is an ordeal that can span for 6 month
15:13:25 <mmedvede> s
15:14:15 <lennyb> mmedvede, it took me only few days to bring up new CI + cloud based on RDO
15:14:33 <mptacekx> if I got you properly guys, it's better to stay on mitaka/trusty in coming months and upgrade to newton/xenial in another couple of months ?
15:14:39 <mmedvede> lennyb: bringing up != maintaining/upgrading
15:15:41 <mmedvede> mptacekx: I think you might be mixing up provider cloud with CI infrastructure images
15:15:45 <asselin_> mptacekx, no....these are 2 different things
15:15:48 <lennyb> mmedvede, correct, but you do you must to upgrade?
15:15:48 <mmedvede> mptacekx: they are normally decoupled
15:15:58 <mmedvede> mptacekx: so upgrading one should be independent of the other
15:16:15 <asselin_> mmedvede, +1
15:17:09 <asselin_> you can run your ci infrastruction on any cloud that provides trusty image...doesn't have to be openstack. It can also run on baremetal.
15:17:11 <mptacekx> mmedvede: If I got your point, we should be able to run newton cloud in kolla containers (controller) and having collocated zuul infra on trusty host ?
15:17:34 <mmedvede> mptacekx: e.g. we have Mitaka dev cloud that we use to run our devstack VMs for testing. It would stay Mitaka for awhile. But we can run xenial infrastructure, or newer, on top of that cloud
15:18:37 <mmedvede> mptacekx: maybe with containers it is different. With VMs there is no strong coupling on OpenStack version
15:18:58 <asselin_> I need to leave early today
15:19:03 <mptacekx> mmedvede: I understand that I can run any image inside our ci cloud, my concern is whether I can reliably run newton kolla cloud on trusty if there is no known way how to deploy ci infra on xenial
15:19:12 <lennyb> asselin_, thanks for coming
15:21:11 <mmedvede> mptacekx: so why do you need CI infra on Xenial to be able to run newton kolla cloud? How is CI infra OS (e.g. what os zuul /nodepool runs on) affects your ability to run kolla?
15:21:36 <mmedvede> and where is that newton kolla cloud? I am a bit confused :)
15:21:39 <mptacekx> mmedvede: we're collocating ci infra & ci cloud controller
15:21:59 <mmedvede> oh
15:22:13 <mmedvede> yes, than you are coupled
15:22:28 <mptacekx> I know better word :-)
15:23:04 <lennyb> mptacekx, can you bring new VM for CI only
15:23:44 <mptacekx> we might consider this option if it's broadly used in other setups ....
15:24:03 <lennyb> IMHO it easier to maintain
15:24:23 <lennyb> we have cloud with VMs, when 1 of them is CI
15:24:39 <lennyb> and others are nodepool vms for jenkins
15:25:45 <mptacekx> what about having CI infra (zuul, nodepool, jenkins) running on single VM, and cloud controller on another or even baremetal ...
15:25:53 <mmedvede> mptacekx: building CI infra and your cloud controller inside the same server (without containers/VMs) would bring problems. Usually you want separation
15:26:08 <mmedvede> mptacekx: +1, you can have zuul/nodepool/jenkins in VM
15:26:58 <mptacekx> thanks a lot, I think it's fully answering my question
15:27:21 <lennyb> mptacekx, anything else you wanted to discuss?
15:27:52 <mptacekx> lennyb: just a single topic, we faced some issue last week on devstack
15:27:54 <mptacekx> https://bugs.launchpad.net/devstack/+bug/1671144
15:27:54 <openstack> Launchpad bug 1671144 in devstack "test-requirements are wrongly installed system-wide" [Undecided,In progress] - Assigned to Radek Zetik (radekx-zetik)
15:28:32 <mptacekx> and we're providing fix for that in https://review.openstack.org/#/c/444788/, it's FYI in this round if someone else is also affected
15:28:47 <lennyb> mptacekx, thanks for bringing this up.
15:29:30 <lennyb> btw, asselin_ proposal regarding single-node-use will not work for us, since zuul triggers multijob :(
15:30:16 <lennyb> mptacekx, mmedvede, pots anything else you would like to discuss?
15:30:48 <pots> is there a list of pinned versions somewhere that are recommended with puppet-openstackci?
15:31:03 <pots> e.g. should zuul be pinned?
15:31:22 <mptacekx> lennyb: thanks, nothing from my side
15:32:03 <mmedvede> pots: zuul should be pinned, not sure if it is documented
15:32:32 <pots> or nodepool?  neither of these were pinned when i installed a new CI a few weeks ago
15:32:43 <pots> can you recommend any values?
15:32:47 <lennyb> pots, this is what we use http://paste.openstack.org/show/602521/
15:33:07 <pots> thanks, i'll try those
15:33:31 <mmedvede> pots: https://github.com/openstack-infra/puppet-openstackci/blob/master/contrib/single_node_ci_data.yaml
15:33:43 <pots> that's it for me
15:33:48 <mmedvede> nodepool, zuul, jjb are pinned there
15:34:31 <lennyb> p.s. we succeed to use common CI on internal project ( had to modify few things locally )
15:35:17 <lennyb> If there is nothing more to discuss, I wish you all the best and see you next week.
15:35:24 <mmedvede> thanks lennyb
15:35:25 <pots> thanks everybody!
15:35:45 <lennyb> #endmeeting