Add a safety check and improve error handling.

[aajay]

• To fix https://github.com/Grokzen/redis-py-cluster/issues/419 add a safety check to ensure there is no stale data before using the connection for executing a command.

• Improve error handling to ensure on any error the connection is disconnected before getting released back to the pool.

• Set connection to None as on ClusterDownError, the connection pool is reset, so to avoid the dangling connection getting added back to the pool.

[Grokzen]

@aajay Is this the same code that upstream redis-py uses for this ready check feature?

[aajay]

@aajay Is this the same code that upstream redis-py uses for this ready check feature?

Yes @Grokzen. Added _check_ready_to_send() similar to redis-py. Apart from this, also fixed two other issues as described in the PR.

[aajay]

@Grokzen Can you please review this PR?

[AmitANetskope]

@aajay Is this the same code that upstream redis-py uses for this ready check feature?

Yes @Grokzen. Added _check_ready_to_send() similar to redis-py. Apart from this, also fixed two other issues as described in the PR.

Reference to similar check added in redis-py library - https://github.com/andymccurdy/redis-py/blob/master/redis/connection.py#L1147

[ns-smundle]

@Grokzen Would you be able to help us with this review ?

[Grokzen]

@aajay I manually implemented this into the code base as i needed to put the check logic elsewhere to mimic the place where redis-py implements it so i added it to the same place and not in the client class

https://github.com/Grokzen/redis-py-cluster/commit/95fee797dac63501f2933c8a0b46d72dbc5250d9