Test idempotency and set min_ansible_version to 2.4 #50
Conversation
test_playbook now: 1 - run test.yml 2 - run test.yml again 3 - run test_idempotency.yml and fail if grep for changed=0.* does not match
Because in travis testing I'd like to use import_playbook rather than duplicating tests for idempotency. It currently passes with 2.2 ansible stable-2.4 branch last real update was in September 2018 ansible stable-2.2 branch last real update was around July 2017
|
|
||
| roles: | ||
| - ansible-role-users | ||
| post_tasks: |
There was a problem hiding this comment.
I'll take this learning opportunity. I don't understand why this post_task is needed. What would happen if it was removed? Could it have another name or comment?
There was a problem hiding this comment.
If it's removed then ansible I think will complain about syntax errors because the task to kill the screen is not in a list of tasks.
It could be changed to tasks (or pre_tasks).
I don't know which comment you refer to so can't comment on that ;))
There was a problem hiding this comment.
What I meant to ask was: Why do we need to kill all the screen sessions?
There was a problem hiding this comment.
Ah. If we don't then there's a 'user logged in', and the role runs the usermod way of changing said user ; which always(as per the task config) results in a run that has changed something.
There was a problem hiding this comment.
Which brings the next good question: why do those tasks change needlessly? :)
One possible improvement to this role would be to add more conditionals to only change/run the task 1 if the GID/UID of the user (and of the logged in process or is that overkill?) are different than what is in code. If the usermod command has some nice return codes one could use those but after some searching, I couldn't find any good ones 2
There was a problem hiding this comment.
Okay, I think I remember hearing about this before. I think I now understand everything ™... ...merging.
For reviewer: Feel free to squash merge and use the below as the squashed commit message:
After #49 was fixed idempotency test fails so yay it hasn't been working for a long time :)
Before this PR it was failing because the "usermod" way of modifying logged in users changes things.
That could perhaps also be improved, but with this PR now while we do the idempotency testing we do so without any logged in users.
As part of testing we now run ansible-playbook three times with two different playbooks:
Maybe running it three times would make testing miss some idempotency failures on the 3rd run, but I'm hoping not :)