SAP Web Dispatcher continues to load balance to failed server/cluster

[Former Member]

We're having trouble with some known issues in Java which is causing several applications and sometimes the entire server node to fail (Java heap). The dispatcher remains active so the message server sees an active application server... which is reported to the Web Dispatcher and folks get a nice big error message when they log in.

These are BW (7.0 Java+ABAP) servers using HTTP.

Is there a config step I'm missing to tell the Web Dispatcher not to send request to an app where a java piece has failed or is restarting?

[Former Member]

Hi,

Yes, this is indeed a problem. To avoid it, we always define at least 2 (or 3 ) server nodes on a Java stack.

If one server node crashes (it happens : The java stack is so much lessreliable than the abap stack), there is still one left to answer the request.

Regards,

Olivier

[Former Member]

That's weird.. By default weighted_round_robin is active, so in case a server node crashes, this will be reported by the dispatcher to the message server.

When you check the JAVA logon group, does it still has the same capacity as when all server nodes are running?

Kind regards,

Mark