On May 20, 2009, at 5:55 PM, Eric Thompson wrote:

...
"kill off the defective scheduler process"

How is this accomplished?  Can this be done on the TSM server, or is
this the process running on the node?

I've read that this may occur if the queryschedperiod and the Backup
Window are the same size or if the queryschedperiod is smaller than
the
Backup window.  Is this a possibility?

This seems to be occurring intermittently, but on the same nodes
most of
the times.

The scheduler process runs on the client.
Your postings don't say what you've looked at on the client to analyze
the problem.  Start with the scheduler log, and then the
dsmerror.log.  If running with CAD, examine its log as well.  And see
what dsm* processed are running on that system: if multiple, the older
one is likely defunct and clogging the works.  Also assure that all
your systems are syncing to a time standard.  (Many Windows shops fail
to do this.)  If the client's TSM software maintenance level is low
compared to what is available for it, consider boosting that.

    Richard Sims