summary: restartability problem

[Kelley Wittmeyer]

thank you to all who responded to our 'restartability' problem (see
original post for the problem).

we didnt really get it fixed for how we were writing our files.  we
just recoded the model to gather everything back to the calling
process which then writes to the files.

i tried increasing the nfs daemons from our redhat default of 8.  this
didnt help the problem, but also added significant wallclock overhead
for jobs.

also tried modifying the nfs mount options to the nodes for the /home
filesystem to which we write and alltho i got improved performance
adding the xsize and wsize options (set to 8192), this did not help
the problem either.

we're still looking at file locks and buffering w/r/t nfs.

so that's it.
kelley wittmeyer
atmospheric science
colorado state university