| products | dev2dev | support | askBEA
 Download Docs   Site Map   Glossary 

WebLogic Server Frequently Asked Questions

 Previous Next Contents View as PDF  

FAQs: Clustering

Q. How do stubs work in a WebLogic Server cluster?

A. Clients that connect to a WebLogic Server cluster and look up a clustered object obtain a replica-aware stub for the object. This stub contains the list of available server instances that host implementations of the object. The stub also contains the load balancing logic for distributing the load among its host servers.

Q. What happens when a failure occurs and the stub cannot connect to a WebLogic Server instance?

A. When the failure occurs, the stub removes the failed server instance from its list. If there are no servers left in its list, the stub uses DNS again to find a running server and obtain a current list of running instances. Also, the stub periodically refreshes its list of available server instances in the cluster; this allows the stub to take advantage of new servers as they are added to the cluster.

Q. How does a server know when another server is unavailable?

A. WebLogic Server uses two mechanisms to determine if a given server instance is unavailable.

Each WebLogic Server instance in a cluster uses multicast to broadcast regular "heartbeat" messages that advertise its availability. By monitoring heartbeat messages, server instances in a cluster determine when a server instance has failed. The other server instances will drop a server instance from the cluster, if they do not receive three consecutive heartbeats from that server instance

WebLogic Server also monitors socket errors to determine the availability of a server instance. For example, if server instance A has an open socket to server instance B, and the socket unexpectedly closes, server A assumes that server B is offline.

Q. How are notifications made when a server is added to a cluster?

A. The WebLogic Server cluster broadcasts the availability of a new server instance each time a new instance joins the cluster. Cluster-aware stubs also periodically update their list of available server instances.

Q. How do clients learn about new WebLogic Server instances?

A. Once a client has done a JNDI lookup and begins using an object reference, it finds out about new server instances only after the cluster-aware stub has updated its list of available servers.

Q. How do clients handle DNS requests to failed servers?

A. If a server fails and DNS continues to send requests to the unavailable machine, this can waste bandwidth. For a Java client application, this problem occurs only during startup. WebLogic Server caches the DNS entries and removes the unavailable ones, to prevent the client from accessing a failed server twice.

Failed servers can be more of a problem for browser-based clients, because they always use DNS. To avoid unnecessary DNS requests with browser-based clients, use a third-party load-balancer such as Resonate, BigIP, Alteon, and LocalDirector. These products mask multiple DNS addresses as a single address. They also provide more sophisticated load-balancing options than round-robin, and they keep track of failed servers to avoid routing unnecessary requests.

Q. How many WebLogic Servers can I have on a multi-CPU machine?

A. There are many possible configurations and each has its own advantages and disadvantages. BEA WebLogic Server has no built-in limit for the number of server instances that can reside in a cluster. Large, multi-processor servers such as Sun Microsystems, Inc. Sun Enterprise 10000, therefore, can host very large clusters or multiple clusters.

In most cases, WebLogic Server clusters scale best when deployed with one WebLogic Server instance for every two CPUs. However, as with all capacity planning, you should test the actual deployment with your target web applications to determine the optimal number and distribution of server instances. See Performance Considerations for Multi-CPU Machines for additional information.

Q. Should I use a separate network for multicast in a cluster?

A. No. Multicast traffic is not heavy enough to require a separate network.

Q. What should I do if my cluster "hangs" or "freezes"?

A. If your WebLogic Server cluster "freezes," you will need to collect certain diagnostic information, including thread dumps and Java garbage collection metrics, before contacting BEA Technical Support. For details, see Generate a Log File.


Back to Top Previous Next