jwhite66
commented
10 years ago More details: in backend/env.py, we have code to subscribe users to our mailing list.
We add a job to _subscribe_to_mailchimp to our deferred queue. In that code, the call to mc.subscribe is not bracketed with a try/catch. The mc.subscribe code throws exceptions for known error conditions. The way the GAE code works, if a deferred job throws an exception, the job gets retried later. That leads to the current situation where we have hundred of subscribes trying, and failing, over, and over.
Note that just bracketing the mc.subscribe call would 'cure' this, but at least at one point in time, the mc service was temporarily down, and that fail safe came in awfully handy.
So, to properly 'fix' this, we'd want to selectively try/catch, and discard known failures (e.g. email invalid, or email is believed to be a spammer), but propagate any unknown exceptions. Alternately, just having the retry logic fire only a few times would also prune things.
davidxia
aaronlifshin
brad
the mc.subscribe python utility will throw an exception on error. That will force the subscribe request to re-queue. That's nice, as sometimes Mailchimp is down, and we retry.
However, for hard failures (e.g. bogus email), it essentially means we go into a loop, trying to subscribe over, and over again.
We could modify the code to catch the set of exceptions that are permanent. Simpler might be to figure out a way to queue the subscribe with a max # of retries.