Node failure timeout is hard-coded to 15s

[mojixcoder]

When a node fails, it takes 15s for go-redis to mark this node as a failed node. You can see it in this line.

Possible Solution

This timeout should be configurable.

Or

Checking cluster nodes in some intervals and discover failed nodes and mark these nodes as a failed node and don't send requests to these nodes.

[mojixcoder]

And there is another thing, in case of failures we call node.MarkAsFailing() multiple times which updates the timestamp of failure.

As we call this method the timestamp gets updated and we never reach this 15s threshold unless node.MarkAsFailing() is not called more than once.

[mojixcoder]

Any updates guys?