17:00:07 <mattclay> #startmeeting Ansible Testing Working Group
17:00:07 <zodbot> Meeting started Thu Nov 15 17:00:07 2018 UTC.
17:00:07 <zodbot> This meeting is logged and archived in a public location.
17:00:07 <zodbot> The chair is mattclay. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:00:07 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
17:00:07 <zodbot> The meeting name has been set to 'ansible_testing_working_group'
17:00:11 <mattclay> #info Agenda: https://github.com/ansible/community/issues/248
17:00:13 <mattclay> #chair gundalow
17:00:13 <zodbot> Current chairs: gundalow mattclay
17:00:28 <mattclay> There is nothing new on the agenda for today.
17:00:31 <mattclay> #topic Open Floor
17:07:46 <gundalow> mattclay: Retry for packages
17:07:49 <gundalow> tasks
17:08:03 <gundalow> I think we've talked about this before, though we are still seeing it a bit
17:08:39 <gundalow> Wondering if we should run https://github.com/ansible/ansible-lint/pull/324 across `test/integration/targets/`
17:10:15 <felixfontein> sounds like a good idea to me
17:10:15 <mattclay> Instead of updating the package installation tasks to have retries I'd like to split out the requirements installation from the tests themselves, which could then include retries.
17:10:56 <mattclay> Either solution would involve updating most of our tests that include package installs.
17:11:20 <felixfontein> there are also other things which download stuff, such as docker containers being pulled
17:12:00 <mattclay> felixfontein: Which test(s) are downloading docker containers, aside from the tests using ansible-test plugins?
17:12:26 <felixfontein> some of the docker_* tests
17:12:39 <felixfontein> they need some images to play with
17:14:00 <felixfontein> I guess most cases can be replaced by something small kept nearby (maybe even loaded into docker from a tar file), except at least in one case, where we need a registry image to test pulling from and pushing to a registry
17:14:20 <felixfontein> registry image = https://hub.docker.com/_/registry/
17:15:27 <mattclay> For most cases where we need to download a file in the tests we've been putting them in our S3 bucket so we have reliable downloads.
17:15:38 <mattclay> That doesn't work in all cases though.
17:15:56 <mattclay> Have we been having failures in the docker_* tests due to download issues?
17:16:10 <felixfontein> I can't remember
17:16:25 <felixfontein> usually enough other things went wrong before it came that far ;)
17:17:12 <gundalow> mattclay: Is that on the 2.8 roadmap, how much work would it be?
17:19:40 <mattclay> gundalow: It isn't on the roadmap, and I haven't scoped it yet so I don't have a good idea of how much work it will be yet.
17:21:40 <felixfontein> ah, something else: I created a PR which adds a docker API version / docker-py version check to the docker_* tests, and fails if not all tests can be run: https://github.com/ansible/ansible/pull/48620
17:21:58 <felixfontein> (just had to rebase to get rid of a conflict, before that it was working fine ;) )
17:24:07 <mattclay> felixfontein: Nice. Hopefully it won't be an issue in the future for CI, but it should definitely help for anyone running tests locally.
17:25:01 <gundalow> I need to step out for a bit, nothing else from me
17:25:41 <felixfontein> mattclay: I hope so too :) I think you requested the fail if not all tests could be run, BTW
17:27:21 <mattclay> felixfontein: Yes, several of us actually asked for that. Having the tests fail is good, otherwise it would be too easy to think tests were passing when instead they were being skipped.
17:28:24 <felixfontein> sounds good!
17:34:29 <mattclay> Anything else?
17:35:12 <felixfontein> not from me
17:37:03 <mattclay> #endmeeting