Application server failure and recovery

When an application server, say Server1, fails, the following occurs:

• The connections from the client also fail.
• Other connections through TSA1 to Server2 and Server3 survive.
• Clients who were sending traffic to the failed application server must send traffic to their alternate TSA (TSA2).
• IPFE-A1 routes new connection requests to the remaining application servers (Server2 and Server3). If all application servers in a target set fail, and IPFE-A1 receives a request for a new connection to TSA1, it optionally notifies the client the request cannot be fulfilled using either a TCP RST packet (for TCP connections) or a configurable ICMP message.

When Server1 recovers:

• IPFE-A1 detects Server1's availability.
• IPFE-A1 routes new connection requests to Server1.
• Some imbalance across application servers in TSA1 exists after recovery. IPFE-A1 monitors for imbalances in traffic and distributes new connections to reduce the imbalance.