17:01:47 <mattclay> #startmeeting Ansible Testing Working Group
17:01:47 <zodbot> Meeting started Thu Jan 11 17:01:47 2018 UTC.  The chair is mattclay. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:01:47 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
17:01:47 <zodbot> The meeting name has been set to 'ansible_testing_working_group'
17:02:05 <mattclay> #chair gundalow
17:02:05 <zodbot> Current chairs: gundalow mattclay
17:02:13 <gundalow> Hi
17:02:32 <mattclay> #info Agenda https://github.com/ansible/community/issues/248
17:02:41 <mattclay> It looks like there's nothing on the agenda today.
17:02:46 <mattclay> gundalow: Do you have anything to discuss?
17:02:55 <gundalow> [16:26] < vachon> gundalow, apparently the subscription in use for the shippable azure is out of quota
17:03:02 <gundalow> from #ansible-devel
17:03:07 <gundalow> #topic Azure quota
17:03:45 <mattclay> I'll take a look at that after the meeting. We've had an increase in Azure tests lately and may be hitting the quota due to that. Tests that don't clean up after themselves properly could also be responsible.
17:04:13 <gundalow> mattclay: ah, so unlike AWS things can be left over between runs?
17:04:18 <mattclay> If the tests don't clean up when finished it can take longer for resources to be removed, particularly if jobs are canceled before finishing.
17:04:39 <mattclay> Nothing gets left over, it just takes longer (hits timeout) instead of immediate cleanup. Same as AWS tests.
17:04:52 * Pilou waves
17:04:56 <gundalow> #chair Pilou
17:04:56 <zodbot> Current chairs: Pilou gundalow mattclay
17:05:06 <gundalow> mattclay: ah, thanks for the detail
17:05:23 <mattclay> In the case of Azure though it usually happens only if the Shippable job is canceled.
17:05:26 <gundalow> #info We've had an increase in Azure tests lately and may be hitting the quota due to that. Tests that don't clean up after themselves properly could also be responsible. If the tests don't clean up when finished it can take longer for resources to be removed, particularly if jobs are canceled before finishing.
17:05:34 <gundalow> Just so I can find it in the logs again after
17:05:56 <gundalow> #topic POWER/AIX and POWER/LINUX testing
17:06:41 <gundalow> kairo misc and Pilou and myself were talking a little bit about testing on IBM POWER earlier in the week
17:07:10 <Pilou> I have created an account there http://openpower.ic.unicamp.br/minicloud/ but I didn't test the related PR yet
17:07:22 <gundalow> kairo works for IBM
17:07:56 <gundalow> Pilou: is the process fairly easy?
17:08:07 <gundalow> http://openpower.ic.unicamp.br/minicloud/ <--- gives you POWER/LINXU
17:08:23 <mattclay> Pilou: I assume those accounts are only suitable for manual testing, not CI?
17:08:39 <Pilou> yes & yes
17:09:37 <misc> yeah, getting some aix is quite expensive
17:09:39 <gundalow> Though there is a contact so we *may* be able to find if we can get something that's suitable for CI
17:10:02 <misc> I did look inside Rh, and we have some aix box for support and QE, but internal only
17:10:20 <gundalow> though I expect it would be low capacity, like macstadium
17:10:33 <mattclay> Probably not fast to provision instances either.
17:10:57 <misc> (also, we did figure that 1.456$ per period wasn't the same as 1,456$ per period)
17:11:08 <Pilou> should we propose to kairo to run a Zuul instance :) ?
17:12:18 <misc> I wonder if that's somewhere where dci could help too
17:14:29 <gundalow> #chair musc kairo
17:14:29 <zodbot> Current chairs: Pilou gundalow kairo mattclay musc
17:14:36 <gundalow> #unchair musc
17:14:36 <zodbot> Current chairs: Pilou gundalow kairo mattclay
17:14:42 <gundalow> #chair misc
17:14:42 <zodbot> Current chairs: Pilou gundalow kairo mattclay misc
17:15:05 <gundalow> kairo: Just chatting about POWER/Linux & AIX
17:15:37 <kairo> Hi all, regarding Power/Linux I spoke with the guy that is responsible by UNICAMP Power minicloud and is possible
17:15:38 <gundalow> misc: You make a really good point, now we have Distributed CI (DCI) We *could* run post-merge tests on POWER/AIX that's behind the RH VPN
17:15:45 <kairo> but not for AIX
17:15:56 <gundalow> kairo: That's great, thanks for contacting them
17:16:13 <gundalow> What exactly do you think they'd be able to give us access to?
17:16:40 <misc> gundalow: or get a partnership with ibm and have them throw some aix server to us ?
17:17:00 <gundalow> catching that would break your arms
17:17:11 <kairo> gundalow: I will work on that tomorrow, but I think next week is possible. I do it on my free time, that I will do my best.
17:17:29 <gundalow> kairo: feel free to cc me in and let me know how I can help
17:17:45 <kairo> I have my personal power, but I need to buy another power supply. :)
17:18:15 <kairo> I have a POWER7 running PowerVM and AIX
17:18:43 <kairo> but, I cannot guarantee when it will be working, sorry.
17:18:56 <kairo> Does Red Hat have Power Machines?
17:18:59 <gundalow> That's cool, you are already doing great things
17:19:53 <kairo> My personal goal is put Ansible better then others automations tools for AIX ;)
17:21:04 <gundalow> kairo: you don't need to have shell set to smitty :)
17:21:10 <kairo> If I could have a place to create DCI tests would be perfect. IBM still investing a lot on Power systems and would be nice if we could have Ansible more supported by AIX
17:21:59 <mattclay> If we have enough capacity for AIX I wouldn't be opposed to adding it to CI.
17:22:21 <gundalow> kairo: I've setup DCI for testing Ansible Networking. So if we can get a small RHEL7/CentOS7 machine running somewhere that can talk to a POWER machine I can do the rest.
17:22:26 <mattclay> Or Linux on POWER.
17:22:46 <gundalow> mattclay: Could ansible-core-ci talk to something behind a VPN?
17:23:00 <kairo> mattclay: a small LPAR running AIX to start would be nice.
17:23:18 <mattclay> gundalow: Yeah, we could probably make that work.
17:24:09 <kairo> gundalow: for Linux on Power it will be easy as the guy for UNICAMP minicloud is my colleague at IBM
17:24:12 <mattclay> We either need VMs that provision quickly (a few minutes) or we'll need to use nodepool to keep instances around for immediate use by CI.
17:26:03 <mattclay> For Linux on POWER, what tests are we expecting to get benefit running vs our existing platforms? Mainly fact gathering?
17:27:57 <kairo> I think facts.,,
17:28:02 <gundalow> Facts was the main thing
17:29:22 <mattclay> That's a small enough set of tests that we may be better off with manual testing, rather than investing in adding another platform to CI.
17:30:17 <gundalow> Would it be worth adding tests for the AIX service work?
17:31:29 <mattclay> Is AIX an option for CI? I though Linux on POWER was the only potential option for CI?
17:31:48 <kairo> gundalow: not only service, if you see I submitted many different modules
17:33:17 <misc> mhh, would facts be different for linux on power than linux on something else ?
17:35:10 <kairo> misc: kernel modules, devices maybe
17:35:57 <misc> kairo: I kinda expect them to the the same code path
17:36:14 <misc> we could run the codecoverage tool and compare
17:37:48 <kairo> misc: to give you it more precisely I will get this Linux box and we can check if it will be a potential option for CI
17:38:18 <gundalow> That would be great
17:38:34 <gundalow> Not sure if it's worth discussing anything else at the moment then
17:38:49 <mattclay> kairo: Thanks
17:39:41 <kairo> about the AIX, any chance to have it as not my power box that is hosted at home :P
17:48:02 <mattclay> A while back I looked at using the IBM Power Development Platform (PDP) for getting access to AIX for testing. Unfortunately the setup there is intended for interactive use rather than CI.
17:49:51 <mattclay> Is there anything else to discuss?
17:50:01 <Pilou> Some test related PR without review: controller host and managed node can be distinct #34498 #33968, broken import #34479 :)
17:50:04 <kairo> I just confirmed and we will have the Linux on Power soon.
17:50:11 <kairo> ;)
17:51:30 <mattclay> #topic controller host and managed node can be distinct #34498 #33968
17:51:56 <mattclay> Pilou: How are you running those tests that your control host and managed host are different?
17:52:04 <gundalow> back
17:52:21 <Pilou> using LXC containers
17:52:41 <Pilou> I try to mimic ansible-test behaviour
17:53:11 <mattclay> Why are you running them on separate hosts? Is it specifically to test that scenario?
17:54:26 <Pilou> because control host and managed node aren't the same in most use case
17:55:38 <mattclay> Were there specific cross-host issues you were trying to test?
17:56:35 <Pilou> there isn't any specific cross-host issues with these 2 PRs
17:57:07 <Pilou> but tests should works anyway
17:59:46 <mattclay> Pilou: I assume you had to run the tests w/o ansible-test, or make some modifications, aside from inventory, to run the tests against lxc containers?
18:00:50 <Pilou> These are similar to #25672, see the ansible-playbook command in the description
18:02:35 <Pilou> extract from my current script: https://paste.fedoraproject.org/paste/4RoYdDmbkYTIc7CS9JDJxw
18:04:22 <mattclay> OK, that makes more sense. I was wondering how you were running the tests with the controller and managed host being different.
18:04:37 <mattclay> Since that's not supported by ansible-test.
18:05:25 <mattclay> I'll review those two PRs and merge after the meeting if there are no issues.
18:05:31 * alikins reads backlog and +1's potentially getting IBM/POWER/AIX/LINUX/POWERVM/ test systems (and related modules merged)
18:05:38 <Pilou> thanks !
18:05:45 <mattclay> #topic broken import #34479
18:06:24 <mattclay> Pilou: This one just needs review?
18:06:31 <Pilou> yep
18:07:45 <mattclay> The changes look OK to me, but you might want to have shertel or ryansb review.
18:09:23 <Pilou> ok. Only 10 exceptions remaining in skip.txt !
18:10:01 <mattclay> If it didn't take so long to provision a Windows instance I'd say add an integration test for it so we could run it in CI.
18:11:33 <mattclay> #topic Open Floor
18:11:57 <mattclay> Anything quick to discuss before the meeting ends? We're already 10 minutes past our scheduled end time.
18:12:20 <gundalow> Nothing else from me
18:13:51 <mattclay> #endmeeting