This information might help you, even though your post indicates that you are using "schedmode polled" rather than our practice, which is "schedmode prompted".
We noticed that, over time, our IRIX 6.5.x clients' allocation of memory (especially swap space) grew out of proportion, eventually requiring either (1) recycling the client, or (2) rebooting the system. Most of our administrators had been using "kill -9 <dsmc-pid>", but that did not return allocated resources to the system. The alternative we found to work was "kill -HUP <dsmc-pid>", which did allow cleanup. However, all this monitoring and recycling turned out to be a bit much of a chore for a couple hundred clients, so we looked into other options, and found the Client Acceptor Daemon/Service (dsmcad). This great little program handles the problem nicely for us, since it is small (SZ/RSS~=180/150, and never grows), and it handles the cycling of the "dsmc sched" program properly, so that system resources are returned at the end of every backup. I suggest that you check the on-line documentation and try the Client Acceptor Daemon or Service (Daemon for UNIX, Service for the Monster from Microsoft), and see if it helps. You will have to ensure the line: "managedservices schedule" is in your dsm.opt file. Dennis Glover Systems Analyst/Backup Administrator Integrated System Test Capability Advanced Research Center Colsa Corporation Huntsville, Alabama USA!
