Should a "retry_on_result" not include a "retry_on_exception=never_retry_on_exception"?

[spex66]

As mentioned in https://github.com/rholder/retrying/issues/25, I hoped to have control over a retry, based on a function, which explicits output a print-statement.

A suspicious long run (in a detached process, which I couldn't debug/interupt easily), had sent me into the internal retrying logic.

With a print statement, after the "sleep" calculation, I was able to observe that it run out of my control into endless retries (with wait_exponential_multiplier).

                sleep = self.wait(attempt_number, delay_since_first_attempt_ms)

                # xxx pa 141117 debug wait
                print "RETRY: ", sleep, attempt_number, delay_since_first_attempt_ms

Reason was that my code had hit an exception, and the default logic is "always_reject" due to that part:

        if retry_on_exception is None:
            self._retry_on_exception = self.always_reject

which ended in endless retries:

RETRY:  2000 1 0
RETRY:  4000 2 2003
RETRY:  8000 3 6004
RETRY:  10000 4 14006
RETRY:  10000 5 24008
RETRY:  10000 6 34010
...

To stop endless loop, I had to override the "retry_on_exception"

    def never_retry_on_exception(response):
        """Return always False to raise on Exception (will not happen by default!)"""
        return False

    # and has to add it to my decorator
    retry429_decorator = retry(
        retry_on_result=retry_if_result_429, 
        wait_exponential_multiplier=1000, 
        wait_exponential_max=10000,
        retry_on_exception=never_retry_on_exception
        )

Summary:

• for me it was a trap, because I thought to have control
• is this intended by design?
• was my approach the correct one?
• If yes, worth to mention more explicit in the doc, that your solution trys by default to retry on tracebacks?

[Trogluddite]

I've encountered this pain as well -- your approach seems correct to me.

I usually use a lambda to make it a little more concise, though:

@retry(retry_on_exception=lambda y: False)
def method_to_retry():
    raise Exception("oh darn")

should raise the exception without retrying at all