I'm setting up a computational cluster in my uni lab. It is being set up to support batch execution (torque/maui) for mpi programs.
The problem is that the batch system runs different processes on different machines so they thus need a shared file system for accessing the required files for the run. I tried setting up mutual nfs sharing, i.e mount all machines in the cluster under /remote, that is to have /remote/machine1 /remote/machine2 ... The problem is that the machines don't come up at the same time so there is a problem mounting each other from the init scripts. Is there any solution to mutual access via the init scripts? (i.e something like lazy mounting. Something like what cifs seems to be doing, that is if the connection was lost to try to reconnect on request, but also at mount time and not only on connection loss). Is glusterfs a solution to this? Thanks -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
