Failure of software, hardware, or local disk on a writer causes a temporary outage for connected clients. Read-write operations are rolled back on the failed node.
Clients connected to any other node will continue to function without disruption. Any distributed query being executed on the writer where the writer is not the leader node for that query will be completed by the leader node and thus cause no disruption to clients connected to the rest of the nodes. Any read-write transaction on the failed writer will be disrupted and the transaction rolled back.
Clients connected to the failed writer will see an outage. When the clients try to reconnect, they can be redirected to a node that is up using the login redirection feature or using a third party redirector. Depending on the severity of the failure, the failed node can be restarted if it is a software issue or restarted after fixing the hardware or disk issue.