A backup script I wrote as a system administration course assignment in 2011.

ec2-backup.sh

#!/bin/bash

set -e
# print usage information if -h flag is passed
usage()
{
	cat 1>&2 <<EOF
usage: $0 [-hv] [-i instance] [-m method] [-t tag] -d dir
	-h             Print a usage statement and exit.
	-v             Be verbose.
	-e recipient   Encrypt the data for the specified recipient. 
	-i instance    Attach the volume in question to the given instance.
	-m (dd|rsync)  Use the given method to perform the backup.
	-t tag         Use the volume with the given tag.
EOF
}

VERBOSE=0

ENCRYPT=cat
INSTANCE=
TERMINATE=0
# default value of METHOD for when -m isn't specified:
METHOD=dd
# default value of TAG for when -t isn't specified:
TAG=ec2-backup
DIRECTORY=

# parse command line options using getopts
while getopts "hve:i:m:t:d:" OPTION
do
	case $OPTION in
		h) usage ; exit ;;
		v) VERBOSE=1 ;;
		e)
			if [ "${ENCRYPT}" = "cat" ] ; then
				ENCRYPT="gpg -e"
			fi
			ENCRYPT="$ENCRYPT -r $OPTARG" ;;
		i) INSTANCE=$OPTARG ;;
		m)
			case $OPTARG in
				dd)    METHOD=dd ;;
				rsync) METHOD=rsync ;;
				*) # exit with an error code if an invalid method is given
					echo $0: "invalid -m value (should be dd or rsync)" 1>&2
					exit 2
					;;
			esac
			;;
		t) TAG=$OPTARG ;;
		d)
			if [ -d $OPTARG ] ; then
				DIRECTORY=$OPTARG
			else
				# exit with an error code if argument to -d is not a directory
				echo $0: $OPTARG" is not a valid directory" 1>&2
				exit 3
			fi
			;;
		# show usage and exit with an error code if invalid arguments given
		# note: getopts already shows a helpful message before "usage" is called
		*) usage ; exit 1 ;;
	esac
done


# don't continue if a directory wasn't specified!
if [ x"${DIRECTORY}" = x"" ] ; then
	echo $0: "a directory must be given with '-d directory'" 1>&2
	exit 4
fi


# handling verbosity in a somewhat clever way
VERBDIRECT=true
if [ $VERBOSE = 1 ] ; then
	VERBDIRECT=cat
fi


# some encryption-related details
if [ "${ENCRYPT}" != "cat" ] ; then
	# don't continue if trying to use gpg and rsync together!
	if [ $METHOD != dd ] ; then
		echo $0: "must use method 'dd' when using PGP encryption"
		exit 10
	else
		echo "encrypting using command: $ENCRYPT"|$VERBDIRECT
	fi
fi


# get size of directory in GB
SIZE=$(du -sBGB $DIRECTORY | awk '{print $1}' | sed 's/[^0-9]//g')
# add 1GiB of size to "round up" / for overhead
SIZE=$(expr $SIZE + 1)


# availability zone (needs to be same between instance and volume)
ZONE=
if [ x"${INSTANCE}" != x"" ] ; then
	ZONE=$(ec2din | grep "^INSTANCE.$INSTANCE.*running" | awk '{print $11}')
fi
if [ x"${ZONE}" = x"" ] ; then
	ZONE=$(ec2-describe-availability-zones | head -1 | awk '{print $2}')
fi


# get information about existing volume
OUTPUT=$(ec2dvol -F "tag-key=$TAG" | grep '^VOLUME.*available' | head -1)
VOLUME=$(echo $OUTPUT | awk '{print $2}')
VOLSIZE=$(echo $OUTPUT | awk '{print $3}')
VOL_JUST_CREATED=0
# does volume exist? if not, create volume
if [ x"${VOLUME}" = x"" ] ; then
	[ $VERBOSE = 1 ] && echo "creating volume in $ZONE"
	# the value for -z is a hack, don't know how to make it not required...
	VOLUME=$(ec2addvol -s $SIZE -z $ZONE | awk '{print $2}')
	if [ $? != 0 ] ; then
		echo $0: "couldn't create volume." 1>&2
		exit 5
	fi
	# tag the new volume, or die trying
	ec2tag -t $TAG $VOLUME |$VERBDIRECT || exit 6
	VOL_JUST_CREATED=1
# if the existing volume isn't big enough, throw an error!
elif [ "$VOLSIZE" -lt "$SIZE" ] ; then
	echo $0: "specified volume is too small! $VOLSIZE < $SIZE" 1>&2
	exit 7
else
	[ $VERBOSE = 1 ] && echo "using $VOLUME"
fi


# if no instance specified at command line, start instance
if [ x"${INSTANCE}" = x"" ] ; then
	EC2_RUN_INSTANCES_FLAGS=$EC2_RUN_INSTANCES_FLAGS" -z $ZONE"
	OUTPUT=$(ec2-start-instance)
	if [ $? != 0 ] ; then
		echo $0: "couldn't create instance to attach volume." 1>&2
		exit 8
	fi
	INSTANCE=$(echo $OUTPUT | awk '{print $1}')
	HOSTNAME=$(echo $OUTPUT | awk '{print $2}')
	TERMINATE=1

	# hackish! give the machine some time to start up services, especially ssh
	sleep 20
else
	# otherwise, just get its host name
	HOSTNAME=$(ec2din|grep "^INSTANCE.$INSTANCE.*running"|awk '{print $4}')
	if [ x"${HOSTNAME}" = x"" ] ; then
		echo $0: "no such instance!" 1>&2
		exit 9
	fi
fi


# attach volume to instance, or die trying
ec2-attach-volume -i $INSTANCE -d /dev/sdh $VOLUME |$VERBDIRECT || exit 10
# give it some time to attach...
sleep 20


# do the backup by whatever method was specified
if [ $METHOD = rsync ] ; then
	# find out if we need to format
	if [ $VOL_JUST_CREATED = 1 ] ; then
		ssh $HOSTNAME "yes | mkfs /dev/sdh" 2>&1 |$VERBDIRECT
	fi
	# mount the drive, sync the files, and unmount
	ssh $HOSTNAME "mkdir -p /mnt/ec2backup" 2>&1 |$VERBDIRECT
	ssh $HOSTNAME "mount /dev/sdh /mnt/ec2backup" 2>&1 |$VERBDIRECT
	rsync -aq $DIRECTORY $HOSTNAME:/mnt/ec2backup 2>&1 |$VERBDIRECT
	ssh $HOSTNAME "umount /dev/sdh" 2>&1 |$VERBDIRECT
else
	tar c $DIRECTORY | $ENCRYPT | ssh $HOSTNAME "dd of=/dev/sdh" 2>&1 |$VERBDIRECT
fi


ec2-detach-volume $VOLUME |$VERBDIRECT
# terminate instance if we started it in this script
if [ $TERMINATE = 1 ] ; then
	ec2-terminate-instances $INSTANCE |$VERBDIRECT
fi

ec2-start-instance.sh

#!/bin/bash

# print usage information if -h flag is passed
usage()
{
	cat 1>&2 <<EOF
usage: $0 [-hv] [-a ami]
	-a AMI  Start an instance of the given AMI.
	-h      Print a usage statement and exit.
	-v      Be verbose.
EOF
}

# default value of AMI for when -a isn't specified:
AMI=ami-b232d0db
VERBOSE=0

# parse command line options using getopts
while getopts "hva:" OPTION
do
	case $OPTION in
		h) usage ; exit ;;
		v) VERBOSE=1 ;;
		a) AMI=$OPTARG ;;
		# show usage and exit with an error code if invalid arguments given
		# note: getopts already shows a helpful message before "usage" is called
		*) usage ; exit 1 ;;
	esac
done


# for the sake of not breaking 80 characters in width, rename this var...
EC2_RIF=$EC2_RUN_INSTANCES_FLAGS

# start the instance and get the instance ID
INSTANCE=$(ec2run $AMI $EC2_RIF | grep '^INSTANCE' | awk '{print $2}')

if [ .$INSTANCE != . ] ; then
	# the first part of the output, the instance id. don't output a newline yet.
	echo -n "$INSTANCE "

	# check the status of the machine until it is running and has a host
	STATUS=pending
	while [ .$STATUS == .pending ] ; do
		STATUS=$(ec2din | grep "^INSTANCE.$INSTANCE" | awk '{print $4}')
		[ $VERBOSE == 1 ] && echo "waiting for instance to be ready..." 1>&2
		sleep 5
	done
	# the fourth thing in the output of ec2din, which was its status before,
	#  just happens to be the position of the hostname!  lucky us.
	echo $STATUS
else
	echo $0: "an error occurred while starting the instance." 1>&2
	exit 2
fi