mongo-backup

#!/bin/bash
#
# Backup script for mongo database backup, to be run from database server
# and subsequently grabbed by backup server. Meant to be run as a cronjob

PROGNAME=$(basename $0)

if [ -z "$1" ]; then
cat << EOF
Usage:
    $PROGNAME [dbname]
    $PROGNAME all

'$PROGNAME' will back up the nominated database, or 'all' databases. It is a wrapper for mongo_dump. It will only work with the first argument nominated.

'$PROGNAME' is designed to be called several times per day. The first time the script is run each day, it will create a daily backup (with a simlink marked 'daily', dated with day-of-month), and every time the script is run it will create an 'on-demand' timestamped copy. The daily backups are overwritten the next month(ish) when the same match occurs. The 'on-demand' copies will be destroyed if they're over 48 hours old when the script runs. These times are governed by the ages of the backup files.


To restore from the backup, use the native restore command:

    mongorestore -d [database name] [--drop] --objcheck [-w 3] [directory with extracted db backup]
    mongorestore -d testdb --drop --objcheck -w 3 testdb/

  --drop destroys the existing database
  --objcheck validates all objects being inserted
  -w [NUM] ensures a write to at least NUM replicas

EOF
exit 1 # run as a cronjob, nice to error if no args :)
fi

DAY=`date +%d`
MONTH=`date +%m`
YEAR=`date +%y`
DATE=`date +%e-%m-%y`
CTIME=$(date +%y-%m-%d-%H%M)
HDIR="$(echo ~)"
MDIR="mongo"
WDIR="$HDIR/$MDIR"
TDIR="$HDIR/tars"
ALERT_EMAIL="alerts@EXAMPLE.COM"

##TODO: differentiate between AWS and DD locations
## this location is Dimension Data specific
#DISKSPACE=$(df -H | grep '/dev/xvda1' | awk '{ print $5 }' | cut -d'%' -f1)
DISKSPACE=$(df -H | grep '/dev/mapper/rootvol00-rootlvol00' | awk '{ print $5 }' | cut -d'%' -f1)
if [ $DISKSPACE -gt "85" ]; then
	echo -e "Subject: ${HOSTNAME} disk nearly full\r\n\r\nThe disk on the ${HOSTNAME} server is at $DISKSPACE% usage" | sendmail $ALERT_EMAIL
fi

errCheck () {
	# check if the previous command returned non 0 exit code and abort
	if [ $? != 0 ]; then
		echo !! exited abnormally $1
		echo -e "Subject: ${HOSTNAME} ERROR\r\n\r\nThe mongo backup on ${HOSTNAME} exited with error message: $1" | sendmail $ALERT_EMAIL
		exit 1
	fi
}

dump_mongo() {
	#perform a local dump
	#args: dbname, destination base dir
    SINGLE_DB=""
    if [ "$1" != 'all' ]; then
        # specify single name with --db
        SINGLE_DB="--db $1"
    fi
	mongodump $SINGLE_DB -o $2
	errCheck "Failed to backup database $1 on server ${HOSTNAME}"
}

backupMongoDatabase () {
	mkdir -p $WDIR
	mkdir -p $TDIR
	#args: dbname
	dump_mongo $1 $WDIR

	for i in $WDIR/*; do
		j=$(basename $i)
		BACKUP_NAMECOMBO=${HOSTNAME}_${j}
		BACKUP_DAILY_NAME=${BACKUP_NAMECOMBO}_${DAY}_dom.mongo.tar.gz

		#if daily backup file does not exist or is not younger than 1 day 
		#overwrites previous backup with same day-of-month name
		if ! test $(find $TDIR/$BACKUP_DAILY_NAME -mtime -1 2>&1 >/dev/null); then
			tar zcf $TDIR/$BACKUP_DAILY_NAME -C $WDIR/ ${j}

			# make a local symlink so it works in offsite backup location
			cd $TDIR
			ln -sf $BACKUP_DAILY_NAME ${BACKUP_NAMECOMBO}_daily.mongo.tar.gz
			cd - 2&>/dev/null
		fi

		# on-demand backup
		BACKUP_NOW_NAME=${BACKUP_NAMECOMBO}_${CTIME}_od.mongo.tar.gz

		tar zcf $TDIR/$BACKUP_NOW_NAME -C $WDIR/ $j

		cd $TDIR
		ln -sf $BACKUP_NOW_NAME ${BACKUP_NAMECOMBO}_current.mongo.tar.gz
		cd - 2&>/dev/null

		# delete 'on-demand' files older than 48 hours
		find $TDIR/*od.mongo.tar.gz -type f -mtime +2 -exec rm {} \;
		#find $TDIR/*od.mongo.tar.gz -type f -mmin +2 -exec rm {} \;

	done
}

backupMongoDatabase $1