Tweepy SSLError

Question

I have a Django management command, launched via supervisord, that uses tweepy to consume the twitter streaming API.

The agent works quite well however I notice in the logs there's an SSLError every 10-15 minutes and supervisord is re-launching the agent.

The tweepy package is latest, version 1.11. The server is ubuntu 12.04 LTS. I've tried installing the cacert into the key chain as mentioned in the link below, but no luck.

Twitter API SSL Root CA Certificate

Any suggestions?

[2012-08-26 19:28:15,656: ERROR] Error establishing the connection
Traceback (most recent call last):.../.../datasinks.py", line 102, in start
    stream.filter(locations=self.locations)
  File "/site/pythonenv/local/lib/python2.7/site-packages/tweepy/streaming.py", line 228, in filter
    self._start(async)
  File "/site/pythonenv/local/lib/python2.7/site-packages/tweepy/streaming.py", line 172, in _start
    self._run()
  File "/site/pythonenv/local/lib/python2.7/site-packages/tweepy/streaming.py", line 117, in _run
    self._read_loop(resp)
  File "/site/pythonenv/local/lib/python2.7/site-packages/tweepy/streaming.py", line 150, in _read_loop
    c = resp.read(1)
  File "/usr/lib/python2.7/httplib.py", line 541, in read
    return self._read_chunked(amt)
  File "/usr/lib/python2.7/httplib.py", line 574, in _read_chunked
    line = self.fp.readline(_MAXLINE + 1)
  File "/usr/lib/python2.7/socket.py", line 476, in readline
    data = self._sock.recv(self._rbufsize)
  File "/usr/lib/python2.7/ssl.py", line 241, in recv
    return self.read(buflen)
  File "/usr/lib/python2.7/ssl.py", line 160, in read
  return self._sslobj.read(len)
SSLError: The read operation timed out

Following is an outline of the code.

from tweepy import API, OAuthHandler
from tweepy.streaming import StreamListener, Stream
# snip other imports

class TwitterSink(StreamListener, TweetSink):

  def __init__(self):
    self.auth = OAuthHandler(settings.TWITTER_OAUTH_CONSUMER_KEY, settings.TWITTER_OAUTH_CONSUMER_SECRET)
    self.auth.set_access_token(settings.TWITTER_OAUTH_ACCESS_TOKEN_KEY, settings.TWITTER_OAUTH_ACCESS_TOKEN_SECRET)
    self.locations = '' # Snip for brevity

  def start(self):
    try:
        stream = Stream(self.auth, self,timeout=60, secure=True)
        stream.filter(locations=self.locations)
    except SSLError as e:
        logger.exception("Error establishing the connection")
    except IncompleteRead as r:
        logger.exception("Error with HTTP connection")

  # snip on_data()
  # snip on_timeout()
  # snip on_error()

Answer 1

The certificate doesn't seem to be the problem. The error is just a timeout. Seems like an issue with tweepy's SSL handling to me. The code is equipped to handle socket.timeout and reopen the connection, but not a timeout arriving through SSLError.

Looking at the ssl module code (or docs), though, I don't see a pretty way to catch that. The SSLError object is raised without any arguments, just a string description. For lack of a better solution, I'd suggest adding the following right before line 118 of tweepy/streaming.py:

except SSLError, e:
  if 'timeout' not in exception.message.lower(): # support all timeouts
    exception = e
    break
  if self.listener.on_timeout() == False:
    break
  if self.running is False:
    break
  conn.close()
  sleep(self.snooze_time)

Why it's timing out in the first place is a good question. I have nothing better than repeating Travis Mehlinger's suggestion of setting a higher timeout.