Caching of the SES rate limit is broken when using an IAM role about django-ses

[pcraciunoiu]

@adarnimrod thanks for reporting this issue. Would you be willing to open a pull request with a proposed fix? (and ideally unit test)

[mdw123]

Same issue. Will look into a PR for this.

I'm glad to see this bug because I was just talking to my co-worker, saying, "What we need to do is cache the GetSendQuota result."

[mdw123]

Reading the code, I don't think this is an actual problem. There's probably a different reason you're calling get_send_quota too much.

cached_rate_limits is a dictionary.

If you use a machine role to access AWS, you correctly point out that self._access_key_id is unset. More specifically, self._access_key_id defaults to None based on line 9 of settings.py

ACCESS_KEY = getattr(settings, 'AWS_SES_ACCESS_KEY_ID',
                     getattr(settings, 'AWS_ACCESS_KEY_ID', None))

Because (None or None) is None, the cache key is None. None is a valid index in a dictionary. Because you're using a machine role to access AWS, there is only ever one SES instance involved and because the cache key is None, I believe this code is functioning as designed, and this issue can be closed.

The problem is more likely because you have too many instances of django-ses trying to find out how to throttle their send rate at the same time, perhaps too many processes, too many nodes, too many somethings. But, within the context of that single EC2 instance, I think it is successfully using the cache.

I think I've figured out what my problem is in general, but I'm not sure how to fix it yet, so you can remove my +1 from this issue.

I think the code is working sufficiently well that you can close this issue.