1

I have a Websphre 7 cluster with nodes running on different servers. When a server with one node loses connection to the network, it takes about a minute, after which the Websphre knows that the member is unavailable.

How can I speed up the status updates?

UPD. The cluster is used only for EJB. EJB called from the local network.

Konstantin
  • 21
  • 4

1 Answers1

3

I think this is always going to be a tradeoff between performance during normal operations and how quickly a down cluster member is detected.

See this article, Understanding HTTP plug-in failover in a clustered environment and this plugin-cfg.xml reference in the WebSphere 7 InfoCenter.

From the article, the answer will involve the ConnectTimeout, ServerIOTimeout, and RetryInterval settings, but note the warning that:

In an environment with busy workload or a slow network connection, setting this value too low could make the HTTP plug-in mark a cluster member down falsely. Therefore, caution should be used whenever choosing a value for ConnectTimeout.​​

dbreaux
  • 4,982
  • 1
  • 25
  • 64
  • Good response. It only address the Web Traffic though (between HTTP Plugin and the application server). Similar operations happen when a component in one application server talks to another component running in another cluster member(for e.g on an EJB call where the JVM affinity is not set) – Manglu Mar 12 '12 at 22:52
  • Good point, @Manglu. I don't do EJBs, so I never considered that. – dbreaux Mar 12 '12 at 23:48
  • The cluster is used only for EJB. EJB called from the local network. – Konstantin Mar 13 '12 at 07:29