|
NFS networked cluster, computers hang on shutdown
Help.
Here's the situation. I have one computer with a complete Slackware Linux 14.0 install. This computer will eventually serve as the command center for a cluster of thirty computers. I have a second computer, a prototype for the thirty computers in the cluster, with a slim set of packages installed to reduce the overhead and optimize performance. Eventually, I will clone the second computer 29 times to create a cluster of 30 machines designed to carry out some extensive N-body integrations. I'm modeling extra-solar planetary systems.
At the moment, the hard drives on the command computer and the prototype are cross mounted so that I can move files back and forth easily. The cross-mounting is working quite well - the only problem is that one computer will hang on shutdown if the other computer is already shutdown. Granted, I don't expect to shutdown these computers often, so this is more an annoyance than a real problem, but I'd like it to go away.
In detail... After shutting down one computer, if I shut down the second (using shutdown -h now), I get the following notices:
killing process holding NSF mount /proc/fs/nfs open...
killing process holding NSF mount /proc/fs/nfsd open...
killing process holding NSF mount /home/mercury6/c01
[The last directory is the directory on the machine that is shutting down that is cross mounted to the machine that is now shut down.]
nfsd has been unmounted
umount: /prc/fs/nfs device is busy
Remounting root filesystem rad only
/dev/sda1 on / type ext2 (ro)
[HANG]
I'm not sure why there is /proc/fs/nfs and a proc/fs/nfsd. I didn't see that when I had the computers running Slackware 9.1. Why are there two nfs processes running, nfs and nfsd? and what's the difference between them? Is this a significant clue?
jjs
|
