Hi,
I am running Amanda-2.4.1p1 to backup an Intel/RedHat6.2 cluster running
PBS and MPI. Quite intermittently some job dies with an error like:
> Broken pipe
> rm_l_10_19092: p4_error: net_recv read: probable EOF on socket: 1
If the same job is resubmitted again there are very good chances it will
finish fine. I wonder if this "probable EOF on socket" may be caused by
a time-out because amanda was doing backup on this node at that time? Is
there a way to run amanda with nice so that the remote gtar be invoked
also with lower priority? Or even better, is there a way to specify the
bandwidth and the priority on a node-by-node basis?
Thanks in advance,
Ted
