I am having trouble conceptualizing how load balance and fail over processes across nodes for dbms_jobs in a RAC environment work. For example, if I want job1 to run in instance1 and job2 to run in instance2 so that memory and CPU is distributed, do I just submit the job while connected to the instance I want?

If jobs are submitted to a particular instance, then what happens in failover, if that instance breaks?

In another train-of-thought, I went to the RAC Administration index of the docs, I went to "Jobs" and then clicked "creating for cluster database". The link brought me into the OEM admin guide to setup via oem/agent where I would chose a destination. If this is the only way to configure, then does oem try a second instance if the defined (destination) instance is broken?

Any more insight? Thnx!