We are seeing a lot of
Could not dispatch request, cancelling instead of retrying exceptions in our logs,
after one node in the cluster went down (even for a few seconds of network problems).
The real issue is that even when the node is back up and normal (and able to answer to new client) -
currently running application keep failing and only a restart of the application resolve this issue.
We looked around in the SDK an did not find anything that can "refresh" the existing connection.
We've considered creating a new connection (create a new Cluster obj and Buckets) in the application once too many of this exception is being thrown - But it really seems like not what we want to do.
Are we missing something ?
Shouldn't the connector try to re-open the connection to that specific node ?
Any best-practice on how to solve this issue?