Folks -
I spoke too soon on the iobuf "fix" working to solve the disk stalling
problems at PSC. I have been running more tests, collecting more data, and
the earlier "success" was a matter of good (well, actually bad) timing.
Both Marcela Madrid (PSC apps support person) and I are seeing horrendous
stalling in runs here and there with the iobuf library enabled and running
(as evidenced by verbose output). So stay tuned; I am bringing it up again
with the PSC folks, and Marcela seems committed to figuring out what can be
done (good to have one of the people there actually working to get to the
bottom of things). Amazing (to me, anyway) that a supercomputer company can
overlook a rather large shared component (the disk).
Regards - Bob
Received on Wed May 24 2006 - 06:07:15 PDT
