Use /dev/urandom, not /dev/random

/dev/random blocks if there is not enough entropy to return the
requested number of bytes, and this turns out to hang our AWS
servers. When generating random bytes for Jabber passwords, use
/dev/urandom instead. We figure that this will be "secure enough,"
since we'll have bigger problems if someone manages to get ahold
of our /dev/urandom's seed value.

lms/djangoapps/jabber/utils.py

@@ -4,6 +4,7 @@ functions to parse the settings, create and retrieve chat-specific
 passwords for users, etc.
 """
 import base64
+import os
 
 from django.conf import settings
 from django.core.exceptions import ImproperlyConfigured
@@ -106,8 +107,7 @@ def __generate_random_string(length):
     # so figure out how many random bytes we need to get a string of
     # just the right length
     num_bytes = length / 4 * 3
-    with open("/dev/random", "rb") as random:
-        return base64.b64encode(random.read(num_bytes))
+    return base64.b64encode(os.urandom(num_bytes))
 
 
 def __validate_settings():