Restrictions

In certain situations, the failover is not provided for 1 to 8 seconds and an SMTP/web service connection to the common NLB IP address fails. This unavailability time window occurs (always) in the following situations:

• When Message Connector node stops abruptly: In this case, it is the NLB (and not the port control) that handles this situation. NLB requires ~5 seconds to detect a failed node. NLB requires 2 to 3 seconds to remove the failed node and redistribute its load to the live nodes.

• When a node starts: Some seconds elapse between when the NLB service starts (with enabled ports - depends on cluster host configuration) and when the Message Connector with port control starts. You can eliminate most of this delay by properly configuring the cluster. See Install Windows NLB, step 4c.

• When the Message Connector process stops working. Some seconds (1 to 3) elapse until the Message Connector port control detects the issue and disables the ports for NLB.

If the SMTP/web service client support retries, the second attempt is presumed to be outside of this time window and the connection request is directed to a working node.