HGETALL return long type value

I deployed a redis cluster with six nodes. And saved a hashmap into it. All redis.conf “timeout” options is 600. All logical are working as expected when my app connect to this cluster using redis-py-cluster. But after one night, I got the following exception( it is a part of the traceback).

NOTE: I have retried to get this key by redis-cli, it is OK. And I checked all redis cluster nodes log(debug level), no error and warning message found.

res = self.redis_conn.hgetall(res_id)
File "/opt/ksc_lbaas/.venv/lib/python2.7/site-packages/redis/client.py", line 1861, in hgetall
  return self.execute_command('HGETALL', name)
File "/opt/ksc_lbaas/.venv/lib/python2.7/site-packages/rediscluster/utils.py", line 93, in inner
  return func(*args, **kwargs)
File "/opt/ksc_lbaas/.venv/lib/python2.7/site-packages/rediscluster/client.py", line 309, in execute_command
  return self.parse_response(r, command, **kwargs)
File "/opt/ksc_lbaas/.venv/lib/python2.7/site-packages/redis/client.py", line 587, in parse_response
  return self.response_callbacks[command_name](response, **options)
File "/opt/ksc_lbaas/.venv/lib/python2.7/site-packages/redis/client.py", line 338, in <lambda>
  'HGETALL': lambda r: r and pairs_to_dict(r) or {},
File "/opt/ksc_lbaas/.venv/lib/python2.7/site-packages/redis/client.py", line 186, in pairs_to_dict
  it = iter(response)
TypeError: 'long' object is not iterable

[Grokzen]

@fossilet @huwei88 I will close this one as Fixed in 1.3.1 for now and if you see this reappearing in future releases after 1.3.1 please open one new ticket with the new issue. Thanks for helping out @fossilet 😃

[huwei88]

I’m sorry to make a mistake. After I add some log in redis-py-cluster. I got the real error. This issue should relate to redis-server socket close. Line https://github.com/Grokzen/redis-py-cluster/blob/unstable/rediscluster/pipeline.py#L466 got ConnectionError because client got empty socket data. See https://github.com/andymccurdy/redis-py/blob/master/redis/connection.py#L144. And because I use pipline, then some command can share the same connection. So if the former command got ConnnectionError, the socket and connection seems will be released. Then we get the “AttributeError: ‘NoneType’ object has no attribute ‘readline’” exceptions.

Using lock reduce this issue reproduce times. But can not fix it. May be this is because of server’s performance. And because I used lock, some requests have to be queued which reduced the redis server pressure.

However, I think redis-py-cluster should raise the real exception(ConnectionError) instead of AttributeError.@Grokzen.