Cp2Nodes

#!/bin/bash

#**************************************************************
#  EPPIC Pre-Computation Script
#  Author : Kumaran Baskaran
#  Date   : 05/07/2013
#***************************************************************

CONCURRENCY=4

help=" Usage: $0 \n
	Script to copy files to all nodes in cluster. Output and error are logged to log directory. Use chk_err.sh and chk_out.sh to check them.\n
        Performs a 'cp -R' pulling files from a common NFS file system to the nodes.\n
        The copy will be done concurrently from $CONCURRENCY nodes at a time. Unless another concurrency valued specified in third parameter.\n
        \t\t    [ -s <dir>  : Source Directory (must be in an nfs exported file system from the head nodes)\n
        \t\t      -d <dir>  : Destination Directory ( must be local to the node. Can only be of form /scratch/<YOUR_FOLDER> )\n
        \t\t      -n <file> : File specifying set of host nodes where the files are to be copied \n
	\t\t	  -l <dir>  : Logs directory \n
        \t\t     [-c <int>] : Concurrency value, number of host nodes where the files are copied concurrently ] \n
Example: ./cp2cluster -s /gpfs/home/duarte_j/data/uniprot_2013_02 -d /scratch/biyani_n -n /blast/merlin-nodes.list -l /blast/copy-logs/\n"

while getopts :s:d:n:l:c:h option
do
        case "${option}"
        in
                s) srcdir=${OPTARG};;
                d) destdir=${OPTARG};;
                n) hfile=${OPTARG};;
		l) logsdir=${OPTARG};;
                c) CONCURRENCY=${OPTARG};;
                h) echo -e $help;;
               \?) print >&2 $help
                   exit 1;;
                :) echo "`date +%d/%m/%y-%H:%M:%S` ERROR: Option -$OPTARG requires an argument." >&2
                   exit 1;;
        esac
done

if [ -z $srcdir ] || [ -z $destdir ] || [ -z $hfile ] || [ -z $CONCURRENCY ] || [ -z $logsdir ]
then
        echo -e "\n`date +%d/%m/%y-%H:%M:%S` INFO: ---- SOME OPTIONS NOT SPECIFIED CORRECTLY ---- " >&2
        echo -e $help
        exit 1
fi

if [ ! -f $hfile ]
then
	echo "`date +%d/%m/%y-%H:%M:%S` ERROR: File $hfile can\'t be read" >&2
	exit 1
fi	

if [ ! -d $logsdir ]
then
	mkdir -p $logsdir
fi
allnodes=`cat $hfile`
echo "`date +%d/%m/%y-%H:%M:%S` INFO: Copy being done from $CONCURRENCY nodes at a time." 

cmd="cp -vR $srcdir $destdir"
count=1

# deleting old output/error logs
rm -f $logsdir/*.out $logsdir/*.err

for node in $allnodes
do
	mod=`expr $count % $CONCURRENCY`
	if [ "$mod" -ne "0" ]
	then
		echo "`date +%d/%m/%y-%H:%M:%S` INFO: Copying $node..."; ssh $node $cmd 1>$logsdir/$node.out 2>$logsdir/$node.err &
		# we introduce a 0.1s delay between the start of each ssh job
		# this speeds up considerably the ssh startup as the mount requests from nodes don't come all exactly at the same moment
		sleep 0.1			
	else
		echo "`date +%d/%m/%y-%H:%M:%S` INFO: Copying $node..."; ssh $node $cmd 1>$logsdir/$node.out 2>$logsdir/$node.err
	fi
	count=`expr $count + 1`
done

# making log files writable (and removable) to the group
chmod g+rw $logsdir/*.out $logsdir/*.err