When running ansible playbooks you do often encounter use-cases where a task (usually shell or command) is expected to return error codes.
So far the solution was to register the result and to add ignore_errors: true and to decide later if it was a real error or not.
Now, there is one issue with this: it will mess the logging as you will see red error like
fatal: ...
...ignoring
Is there a way to avoid these so we would not have fake errors in the logs?
failed_when and changed_when is to help you:
- shell: echo good | grep bad
register: res
failed_when: false
changed_when: false
This will always be good and green despite failed shell command.
You can also define a complex failed_when statement based on registered variable.
The best way is https://stackoverflow.com/a/40430875/1849837 (use changed_when and failed_when). Note that it works from Ansible since 1.4.
For older Ansible version the only way I see (and use) is do not run potentially failing task at all. You may use conditions (when) for that.
First you need to check the condition and assign the result of check to the variable
- name: "check if condition is met"
command: some condition that populates result
register: result
- name: "some operation"
command: operation that runs only if result is not failed.
when: result|failed
This includes some boilerplate code (you need artificial step to test the condition) but I see no better alternatives. I believe that this is better than polluting the logs with fatal: ... ...ignoring.
