I have a working playbook that searches and collects a string from a file and, based on that string, sets another variable.
But I’m retrieving the string using grep/awk:
I’m curious how these notions caught on that very clearly expressed shell expressions are inherently “bad”, and that unfathomably complex jinja2 expressions are inherently “better”.
I seriously doubt you’re going to find a more maintainable, all-Ansible expression to accomplish what you’re doing.
However, I would suggest a couple alterations.
(A) I’d make that “changed_when:” into a “failed_when:”, and
(B) I’d add “changed_when: false” (because it doesn’t ever change anything).
(C) I might also add a “vars:” to your set_fact task that maps “location[XYZ]” to your smtp.conf expressions:
Then your set_fact becomes "snmpd_conf: ‘{{ locmap[grep_result.stdout_lines[0]] }}’
That way you don’t have to change “code” to update your location map data.
Thanks a lot for the insights, Tod. Very appreciated…
And suggestions as well. I’ll give them a try.
I’m still in an Ansible learning process and sometimes I also come upon that curiosity of yours…
How come the notion of avoiding shell “at all cost” exists even for very specific situations, like this one.
That’s why I, based on the “no shell” assumption, tried to make my playbook better.
I suspect the broad notion that shell == bad and native statements == good is probably just conflation/extrapolation. And human nature
The generally received wisdom when I was taught Ansible by Ansible folk was native modules > command > shell.
Note this is not ‘no shell at any costs’, but rather, try and avoid shell because it can be fragile with globbing, expansion, unexpected behaviour with weird file names etc.
Second thought, on this topic is you could collapse the shell to command here like so:
And finally, while the general pattern probably fits your purposes Alex, it kinda feels like getting stuff that might more naturally be considered inventory information from flat/static files? That’s probably totally cool for your current needs but I’d potentially think about exposing location as a fact, either through inference from dynamic inventory/inventory plugins, or baking into custom facts so the host can report its location (though you might have some chicken-and-egg constraints guiding you down your current path.)
Yes, that’s how I usually go with (no grep + awk).
But back then, I was having so many “trouble” escaping the special characters associated with awk’s syntax that I decided to go that way.
Now that things are looking better, I’ll give that a try again. Otherwise I’ll miss learning…
Thanks, Will
Yes, I may have exaggerated a bit by going with ‘avoid at all cost’… You are right about the modules > command > shell statement. But that still made me think the solution I used was not the way to go.
“(though you might have some chicken-and-egg constraints guiding you down your current path.)” Yup, that’s exactly the reason.
For now I have to rely on collecting that information from an already established way of locating our servers.
Eventually will migrate away from that and have something more coupled with Ansible. Thank you for your suggestions, by the way…