Nortel Networks WEB OS 212777 - Manual (Page 246)

Web OS 10.0 Application Guide
246 n Chapter 10: Health Checking
212777-A, February 2002
Failure Types
Service Failure
If a certain number of connection requests for a particular service fail, the session switch
places the service into the service failed state. While in this state, no new connection requests
are sent to the server for this service; however, if graceful real server failure is enabled
(
/cfg/slb/adv/grace ena ) , state information about existing sessions is maintained and
traffic associated with existing sessions continues to be sent to the server. Connection requests
to, and traffic associated with, other load-balanced services continue to be processed by the
server.
Example : A real server is configured to support HTTP and FTP within two real server groups.
If a session switch detects an HTTP service failure on the real server, it removes that real
server group from the load-balancing algorithm for HTTP but keeps the real server in the mix
for FTP. Removing only the failed service from load balancing allows users access to all
healthy servers supporting a given service.
When a service on a server is in the service failed state, the session switch sends Layer 4
connection requests for the failed service to the server. When the session switch has
successfully established a connection to the failed service, the service is restored to the load-
balancing algorithm.
Server Failure
If all load-balanced services supported on a server fail to respond to switch connection requests
within the specified number of attempts, then the server is placed in the server failed state.
While in this state, no new connection requests are sent to the server; however, if graceful real
server failure is enabled (
/cfg/slb/adv/grace ena ), state information about existing
sessions is maintained and traffic associated with existing sessions continues to be sent to the
server.
N OTE – All load-balanced services on a server must fail before the switch places the server in
the server failed state.
The server is brought back into service as soon as the first service is proven to be healthy.
Additional services are brought online as they are subsequently proven to be healthy.