Feature request: Timeouts / Retry about terraform-provider-shell HOT 3 OPEN

resource timeouts are definitely a good idea and not too difficult to implement. retry also sounds like a good idea, except i would probably add a max_retries attribute or something

from terraform-provider-shell.

Thanks for the speedy answer :)

max_retries seem sensible indeed as a failsafe. I'm not sure how the retry logic intersects with timeouts. IF timeouts are counted against the whole operation - retries included - then max_retries doesn't add much to safety ; just to expressiveness. I'm not sure the corner cases are worth it. On the other hand, IF timeouts are per-attempt, then max_retries seems pretty must-have for safety.

In the context of retrying, passing some additional context through the environment could help the script make a better decision on "should we keep trying or not":

• TF_SHELL_RETRY_ATTEMPT : 1 means first attempt, then 2, 3, 4. Conveniently this would allow to implement some flavor of exponential backoff script-side if desired. Also enables script to enforce max_retries if so desired (but provides less safety than handling that aspect in the provider).
• TF_SHELL_RETRY_FIRST_ATTEMPT_TS : unix timestamp of first attempt. Enables script to give up before timeout if so desired. Convenient to reconcile multiple attempts in logs (barring some kind of TF_SHELL_INVOCATION_ID that isn't there either)

from terraform-provider-shell.

What's the actual behavior with timeout, is the go code hardcoded to timeout after 1 minute?

from terraform-provider-shell.