S3 transfer does not properly handle running out of disk space

[jamesls]

If we run out of disk space (including inodes) we keep trying to sync. We also don't throw any kind of exception when this happens. I believe the root cause if from this code snippet:

def save_file(filename, response_data, last_update):
    body = response_data['Body']
    etag = response_data['ETag'][1:-1]
    d = os.path.dirname(filename)
    try:
        if not os.path.exists(d):
            os.makedirs(d)
    except Exception:      # <---- bad
        pass

We should be more selective in this case and only allow errno.EEXIST errors.