Seitenhistorie

#!/bin/bash

echo "node: $PBS_NODENUM"
echo "task: $PBS_TASKNUM"

numnodes=$(wc -l $PBS_NODEFILE)

echo "node number: $numnodes"

index=$(expr $PBS_TASKNUM - 2)

master_info=/gpfs/scratch/$PBS_JOBID/master.info

echo "rank: $index"

if [[ $index -eq 0 ]] ; then
	echo "running on master"
	port=$(shuf -i 2000-65000 -n 1)
	echo "" > $master_info
	echo "master=$HOSTNAME-ib" >> $master_info
	echo "port=$port" >> $master_info
	echo "wrote master info to $master_info"
	echo "starting master process"

	.... now you call/module load

else
	echo "running on slave"
	echo "waiting for master for some seconds"
	while [ ! -f $master_info ]
	do
		sleep 1 # or less like 0.2
	done
	echo "found info file $master_info"
	sleep 1

	master=$(grep master $master_info | awk -F'=' '{print $2}')
	port=$(grep port $master_info | awk -F'=' '{print $2}')

	echo "found master at $master:port"

	... now your call/module load

f