ec2spotter-remount-root

#!/bin/bash
# "Phase 2" this script is invoked inside of the spot VM at boot time to find the root volume.

manual_invoke=1

# Read the input args
while [[ $# -gt 0 ]]
do
key="$1"
case $key in
	--force)
	manual_invoke=0
	shift # past argument
	;;
    --vol_name)
    ROOT_VOL_NAME="$2"
    shift # past argument
    ;;
    --vol_region)
    ROOT_REGION="$2"
    shift # past argument
    ;;
    --elastic_ip)
    ec2spotter_elastic_ip="$2"
    shift # past argument
    ;;
    *)
            # unknown option
    ;;
esac
shift # past argument or value
done

if [ manual_invoke = 1 ]
then 
  # this is here just to cover accidental invocation from the command-line. it should "never happen"
  echo "This script is destructive when invoked at the wrong time. If you are seeing this message you are doing something wrong."
  exit -1
fi

if [ "x$ROOT_VOL_NAME" = "x" ]; then echo "missing volume name"; exit -1; fi


if [ "x$ROOT_REGION" = "x" ]
then 
	echo "missing volume region"; 
	exit -1; 
fi

# Get Instance ID and current zone from AWS
export INSTANCE_ID=`curl -s http://169.254.169.254/latest/meta-data/instance-id`
export SPOT_ZONE=`curl -s http://169.254.169.254/latest/meta-data/placement/availability-zone`
export SPOT_REGION=${SPOT_ZONE::-1} # Get the region by stripping the last character of the ROOT_ZONE

export AWS_CREDENTIAL_FILE=/root/.aws.creds
. /root/.aws.creds
export AWS_ACCESS_KEY=$AWSAccessKeyId
export AWS_SECRET_KEY=$AWSSecretKey

# acquire the elastic ip
if ! [ "$ec2spotter_elastic_ip"x = x ]
then
	aws ec2 associate-address --instance-id $INSTANCE_ID --allocation-id $ec2spotter_elastic_ip --allow-reassociation --region ${SPOT_REGION}
	echo "Elastic IP associated."
fi

echo "Looking for a volume in $ROOT_REGION with Name = $ROOT_VOL_NAME"
aws ec2 describe-volumes \
        --filters Name=tag-key,Values="Name" Name=tag-value,Values="$ROOT_VOL_NAME" \
        --region ${ROOT_REGION} > volumes.tmp || exit -1

ROOT_ZONE=$(jq -r '.Volumes[0].AvailabilityZone' volumes.tmp)
ROOT_VOL=$(jq -r '.Volumes[0].VolumeId' volumes.tmp)
ROOT_TYPE=$(jq -r '.Volumes[0].VolumeType' volumes.tmp)

# Are we copying the volume from the same region?
if [[ $ROOT_ZONE != $SPOT_ZONE ]]; then
    # need to copy the volume across
    echo "Volume $ROOT_VOL is in another Availability Zone"
    echo "Creating a snapshot of the volume"
    # SNAPSHOT=$(${APIBIN}/ec2-create-snapshot --region ${ROOT_REGION} $ROOT_VOL --description 'ec2-spotter temporary snapshot ok to delete' | awk '{print $2}')
	aws ec2 create-snapshot --region ${ROOT_REGION} --volume-id $ROOT_VOL --description 'ec2-spotter temporary snapshot ok to delete' > snapshot.tmp || exit -1
	SNAPSHOT=$(jq -r '.SnapshotId' snapshot.tmp)

    echo "Snapshot $SNAPSHOT created. Waiting for completion"
    # Keep checking to see that snapshot has been created
    count=0
    while /bin/true
    do
        sleep 5
        eval count=$((count+5))
        echo "... $count seconds gone. Still waiting..."
        # STATUS=$(${APIBIN}/ec2-describe-snapshots --region ${ROOT_REGION} ${SNAPSHOT} | grep completed)
		STATUS=$(aws ec2 describe-snapshots --snapshot-ids ${SNAPSHOT} --region ${ROOT_REGION} | grep completed)
        [[ ! -z $STATUS ]] && break
    done
    echo "Snapshot $SNAPSHOT created successfully"
    echo "------------------------------------------------"
    echo ""
    
    if [[ $SPOT_REGION != $ROOT_REGION ]]; then
	    # NEW_SNAPSHOT=$(${APIBIN}/ec2-copy-snapshot -r ${ROOT_REGION} -s ${SNAPSHOT} | awk '{print $2}')
		aws ec2 copy-snapshot \
			--source-region ${ROOT_REGION} --region ${SPOT_REGION} \
			--source-snapshot-id ${SNAPSHOT} --destination-region ${SPOT_REGION} > new_snapshot.tmp || exit -1
		NEW_SNAPSHOT=$(jq -r '.SnapshotId' new_snapshot.tmp)
	    echo "Copying snapshot from $ROOT_REGION to $SPOT_REGION with name $NEW_SNAPSHOT. Waiting for completion"		
			
	    # Keep checking to see that snapshot has been copied		
	    count=0		
	    while /bin/true		
	    do		
			sleep 5		
			eval count=$((count+5))		
			echo "... $count seconds gone. Still waiting..."		
			# STATUS=$(${APIBIN}/ec2-describe-snapshots ${NEW_SNAPSHOT} | grep completed)
			STATUS=$(aws ec2 describe-snapshots --snapshot-ids ${NEW_SNAPSHOT} --region ${SPOT_REGION} | grep completed)
			
	       [[ ! -z $STATUS ]] && break		
	    done		
			
	    echo "Snapshot $NEW_SNAPSHOT created successfully"		
	    echo "------------------------------------------------"		
	    echo ""
    else
	    NEW_SNAPSHOT=$SNAPSHOT
    fi

    # create volume from this new snapshot
    #NEW_VOLUME=$(${APIBIN}/ec2-create-volume --snapshot ${NEW_SNAPSHOT} -z ${SPOT_ZONE} --type $ROOT_TYPE | awk '{print $2}')
	aws ec2 create-volume --snapshot-id ${NEW_SNAPSHOT} --availability-zone ${SPOT_ZONE} --volume-type $ROOT_TYPE --region ${SPOT_REGION} > new_volume.tmp || exit -1
	NEW_VOLUME=$(jq -r '.VolumeId' new_volume.tmp)
    echo "Creating volume $NEW_VOLUME from $NEW_SNAPSHOT. Waiting for completion"
    
    # Keep checking to see that volume has been created
    count=0
    while /bin/true
    do
        sleep 5
        eval count=$((count+5))
        echo "... $count seconds gone. Still waiting..."
        #STATUS=$(${APIBIN}/ec2-describe-volumes ${NEW_VOLUME} | grep available)
		STATUS=$(aws ec2 describe-volumes --volume-ids ${NEW_VOLUME} --region ${SPOT_REGION} | grep available)
        [[ ! -z $STATUS ]] && break
    done
    
    echo "Volume $NEW_VOLUME created successfully"
    echo "------------------------------------------------"

    echo "Renaming the old volume"
    aws ec2 create-tags --resources ${ROOT_VOL} --tags Key=Name,Value="${ROOT_VOL_NAME}-is-now-$NEW_VOLUME" --region ${ROOT_REGION}
	echo "Renaming the new volume"
	aws ec2 create-tags --resources ${NEW_VOLUME} --tags Key=Name,Value="${ROOT_VOL_NAME}" --region ${SPOT_REGION}
    PERSISTENT_ROOT_VOLUME=${NEW_VOLUME}
else
    echo "Using the volume as-is, no migration needed"
    PERSISTENT_ROOT_VOLUME=${ROOT_VOL}
fi

echo ""
echo "Attaching volume $PERSISTENT_ROOT_VOLUME as /dev/sdf"
# Attach volume
#${APIBIN}/ec2-attach-volume $PERSISTENT_ROOT_VOLUME -d /dev/sdf --instance $INSTANCE_ID || exit -1
aws ec2 attach-volume --volume-id $PERSISTENT_ROOT_VOLUME --instance-id $INSTANCE_ID --device /dev/sdf --region ${SPOT_REGION} || exit -1

while ! lsblk /dev/xvdf
do
  echo "waiting for device to attach"
  sleep 5
done

# Remove .aws.creds file because it's not very safe to keep them on a random server
unset AWS_ACCESS_KEY
unset AWS_SECRET_KEY
unset AWSAccessKeyId
unset AWSSecretKey
rm $AWS_CREDENTIAL_FILE

# Ready up for the swap
DEVICE=/dev/xvdf1
NEWMNT=/permaroot
OLDMNT=old-root
e2label $DEVICE permaroot
tune2fs $DEVICE -U `uuidgen`
mkdir $NEWMNT

#
# point of no return... 
# modify /sbin/init on the ephemeral volume to chain-load from the persistent EBS volume, and then reboot.
#
mv /sbin/init /sbin/init.backup
cat >/sbin/init <<EOF11
#!/bin/sh
mount $DEVICE $NEWMNT
[ ! -d $NEWMNT/$OLDMNT ] && mkdir -p $NEWMNT/$OLDMNT

cd $NEWMNT
pivot_root . ./$OLDMNT

for dir in /dev /proc /sys /run; do
    echo "Moving mounted file system ${OLDMNT}\${dir} to \$dir."
    mount --move ./${OLDMNT}\${dir} \${dir}
done
exec chroot . /sbin/init
EOF11
chmod +x /sbin/init
shutdown -r now