Add a backoff and retry to S3 puts to help with Slow Down messages.

[zerebubuth]

No description provided.

[rmarianski]

Are you sure that the retry will actually catch the exception? I think it won't end up working that way. Here's a simple example:

def retry():
    for _ in range(3):
        try:
            print 'yielding'
            yield
            print 'break'
            break
        except Exception:
            print 'caught exception'
    else:
        print 'else'
        yield


for _ in retry():
    raise Exception('foo')

[rmarianski]

Interesting idea to use a generator, but I don't think that it'll actually catch the exception here and retry.

[zerebubuth]

Are you sure that the retry will actually catch the exception? I think it won't end up working that way.

Oooops. Wow, I hadn't realised that yield wasn't in the path to catch exceptions. Thanks for pointing that out.

I just pushed a different, and sadly more complex, version in 2520055. Please could you have a look and let me know if you think that's better?

[iandees]

Looks good -- it'd be nice to add a logger as mentioned but not required.

[iandees]

It'd be nice if there was a logger we could pass in so we know when things retry rather than stuff just silently taking forever.

[zerebubuth]

@iandees, @rmarianski does that look good with the logger threaded through?

[rmarianski]

@iandees, @rmarianski does that look good with the logger threaded through?

LGTM 👍