DSXchange

The server is a RAC (Real Application Cluster) server with Oracle 10g running in addition to DataStage. When we start a large (20-way) DataStage process, Oracle shuts down. This is because the Oracle cluster communication process "oprocd.bin" is unable to get the cpu time it needs to check in with other cluster members, and so decides to re-boot the system.

Our solution is to shut down Oracle. DataStage then finishes.

Can anyone suggest a solution to this problem?

Fiddle with priorities ("nice" numbers). Give the Oracle cluster communication process a tad more priority. Your assertion is somewhat surprising, because DataStage processes operate as background processes and should automatically receive lower priority as a result. But, then, oprocd.bin probably runs as a background process also.