> We have a job that runs across several compute nodes, and all but > the last process starts. The last process (on the master node) fails with a > "Text file busy" error on the program binary. Do you have anything in the client logfile on the master node? Avati