SAP Java Scheduler - Cluster Behavior

[Former Member]

Hello,

we are using SAP Java Scheduler in 7.30 environment and are focussing some Cluster-related questions.

Due to the Message Driven Bean approach, it seems like every task will be executed on one server node in the cluster, and that the selection of this  node seems to work in a random way. From my point of view this is OK and well-implemented. But as we are performing some node-specific operations in our Custom Job implementations, we need to have some more information about this behavior.

Does anyone have experience about Cluster-behavior of the Java Scheduler. Is the node-selection deterministic or can it be influenced by some configurations?

I did not find any information about this topic in the official doc like http://scn.sap.com/docs/DOC-7700, so maybe there are some real world experiences

Thanks in advance

Mirco

[satishra · Accepted solution]

Hi Mirco,

Cluster-behavior of Java Scheduler is handled through JMS

Jobs are implemented on the basis of message-driven beans. A message-driven bean

containing a job is called a JobBean. The execution of JobBeans is handled by the EJB

container. A JobBean is executed when it receives a Java Messaging Service (JMS)

message from the scheduler runtime service. In cluster environment, the JMS is responsible

for load balancing: it decides which JobBean instance on which node gets the request to run.

You can details info in the following guide

http://www.sdn.sap.com/irj/scn/go/portal/prtroot/docs/library/uuid/e8ab5056-0e01-0010-5581-d3e51724e...

[Former Member · Accepted solution]

Hi Satish,

so it seems like the whole cluster/load-balancing in SAP Java Scheduler is just based on "plain JMS".

Therefore the behavior is configurable in NWA. I had a look at the JMS Server Configuration in NWA, you can configure the JMS queue "jmsqueues/default/JobQueue", which is configured as "roundrobin" per default.

Thanks for your support!!

Best regards

Mirco