X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=backup.sh;h=7fc7ffb0018fecd951b3d64249637aa5ea28cc71;hb=fcc343a7f7be5c6f9348b05b7dbdd429d955d4d0;hp=c1c4439c44e73d9500cc175b14b959338c777712;hpb=d5fab66d097c6b9db8b54ad0a70ee0b63b4989df;p=backup.sh diff --git a/backup.sh b/backup.sh index c1c4439..7fc7ffb 100755 --- a/backup.sh +++ b/backup.sh @@ -18,30 +18,35 @@ # The script is halfway Norwegian and halfway English. Newer modifications are # in English, we should probably stick to that. -# The computer to backup. -computer=$1 - -# Configuration -LOCKFILE=/home/backup/backuprun.lockfile.$computer -confdir=/home/backup/conf/ # Configuration files -storagedir=/backup # Where we keep backups -maxnumfull=3 # Number of full backups -daysbetweenfull=30 # Days between full backups -daysbetweenfullforfixed=32 # Days between full backups for machines with fixed - # full-backup date (in case the fixed day gets missed) -DAY_OF_MONTH=`date "+%d" | sed s/^0//` # from 1 to 31 -DATE=`date "+%Y%m%d%H%M"` #format: touch -DATEs=`date "+%Y-%m-%d %H:%M"` #format: tar - -# Exclude-pattern -exclude=$confdir/exclude -[ ! -f $exclude ] && printf "tmp\ncore\n" > $exclude +SCRIPTPATH="$( cd "$(dirname "$0")" ; pwd -P )" # Absolute directory the script lives in +confdir=$SCRIPTPATH/conf/ # Directory configuration files +[ -r $confdir/env ] && . $confdir/env + +# Default configuration; used unless $confdir/env has set something different. +storagedir=${storagedir:-/backup} # Where we keep backups +exclude=${exclude:-$confdir/exclude} # List of exclude patterns +maxnumfull=${maxnumfull:-3} # Number of full backups +daysbetweenfull=${daysbetweenfull:-30} # Days between full backups + +# Days between full backups for machines with fixed full-backup date +# (in case the fixed day gets missed). +daysbetweenfullforfixed=${daysbetweenfullforfixed:-32} + +# End of configuration. + +computer=$1 # The computer to backup. +LOCKFILE=$SCRIPTPATH/backuprun.lockfile.$computer +DAY_OF_MONTH=`date "+%d" | sed s/^0//` # from 1 to 31 +DATE=`date "+%Y%m%d%H%M"` # format: touch +DATEs=`date "+%Y-%m-%d %H:%M"` # format: tar if [ -z "$computer" ]; then echo "Usage: $0 COMPUTER" exit 1 fi +[ ! -f $exclude ] && printf "tmp\ncore\n" > $exclude + # Die more or less gracefully (at least notify user) die() { warnmsg "Aborting." @@ -63,9 +68,10 @@ infomsg() { # Trap C-c and kill trap die SIGINT SIGTERM -# Don't start if we're already running +# Don't start if we're already running. Don't use diemsg, as that removes the lock file! if [ -e $LOCKFILE ]; then - diemsg "$LOCKFILE exists for pid `cat $LOCKFILE`." + warnmsg "$LOCKFILE exists for pid `cat $LOCKFILE`." + exit 1 fi; touch $LOCKFILE echo $$ > $LOCKFILE @@ -76,15 +82,15 @@ umask 027 backup() { - infomsg "$computer:$filesystem $backuplevel backup\n" + infomsg "$computer:$filesystem $backuplevel backup" localsnar="${storagedir}/${computer}/${sfilesystem}/.incremental.snar" - remotesnar="/root/.backup/${sfilesystem}.snar" + remotesnar="$remotehome/.backup/${sfilesystem}.snar" if [ "$backuplevel" = "daglig" ]; then # If incremental backup, we need to copy the incremental status to $computer. # If it does not exist, whine a bit and then run date-based instead. if [ -s "$localsnar" ]; then - if ! scp $localsnar root@$computer:$remotesnar; then + if ! scp $localsnar $username@$computer:$remotesnar >&2; then diemsg "Could not copy .incremental.snar to $computer" fi lastcmd="--listed-incremental=$remotesnar --no-check-device" @@ -112,8 +118,8 @@ backup() # Pipe the stuff over ssh to ourselves, run pee to cat the contents to a # file and build a file list, respectivly. TARFILE=$DATE.tmp - TARCMD="ssh root@$computer \"nice -n 19 ionice -c3 tar --one-file-system --use-compress-program $compressor -cf - -C $filesystem . $lastcmd \ - --exclude-from=/root/.backup/exclude\" | pee \"cat > $TARFILE\" \"tar tzvf -\"" + TARCMD="ssh -n $username@$computer \"nice -n 19 ionice -c3 tar --one-file-system --use-compress-program $compressor -cf - -C $filesystem . $lastcmd \ + --exclude-from=$remotehome/.backup/exclude\" | pee \"cat > $TARFILE\" \"tar tzvf -\"" infomsg "Running $TARCMD" eval $TARCMD > $DATE.idx @@ -121,7 +127,7 @@ backup() # File is >0 in size and neither cat or tar tzvf failed; we assume it worked. if [ "$remotesnar" ]; then - if ! scp root@$computer:$remotesnar $localsnar.tmp; then + if ! scp $username@$computer:$remotesnar $localsnar.tmp >&2; then diemsg "Could not copy .incremental.snar from $computer" fi fi @@ -135,15 +141,53 @@ backup() # Update timestamp echo $DATEs > ../.date && touch -t $DATE ../.date - # Make a sorted file list as well - sort -k6 < $DATE.idx > $DATE.sdx - # Fix permissions chmod 600 *tgz #only for us - chmod 644 *sdx *idx 2>/dev/null #everyone can read + chmod 644 *idx 2>/dev/null #everyone can read # Let the remote computer know that we ran a successful backup (for nagios) - ssh root@$computer touch $filesystem/.lastbackup + ssh -n $username@$computer touch $filesystem/.lastbackup + else + # Something wrong happened. + rm $TARFILE + diemsg "tar failed or $TARFILE empty. $backuplevel backup of $computer:$filesystem failed and deleted" + fi +} + +# Non-privileged backups currently have a few limitations compared to regular ones: +# +# - No .idx files. +# - No .snar-style incremental backups, only date-based. +# - No touching of .lastbackup for Nagios. +# - Failed tar operations may go undetected, since we only see the error code +# the last process in the pipe chain, which is gpg. +nonprivbackup() +{ + infomsg "$computer:$filesystem $backuplevel non-privileged backup" + + if [ "$backuplevel" = "daglig" ]; then + lastd=`cat ../.date` + lastcmd="'$lastd'" + else + lastcmd="" + fi + + TARFILE=$DATE.tmp + TARCMD="ssh -n $username@$computer \"sudo /usr/local/sbin/output-encrypted-backup $filesystem $lastcmd\"" + infomsg "Running $TARCMD" + eval $TARCMD > $TARFILE + + if [ $? -eq 0 ] && [ -s $TARFILE ]; then + # File is >0 in size and the remote side did not report any fatal errors; we assume it worked. + + # Move tar file in place + mv $TARFILE $DATE.tgz.gpg + + # Update timestamp + echo $DATEs > ../.date && touch -t $DATE ../.date + + # Fix permissions + chmod 600 *tgz.gpg #only for us else # Something wrong happened. rm $TARFILE @@ -161,16 +205,29 @@ targetfilesystem=`df -P "$storagedir/$computer"` if [ "$rootfilesystem" == "$targetfilesystem" ]; then diemsg "Target filesystem ($storagedir/$computer) was mounted on /." fi + +# Find the user to do the backup as. +if [ -f $confdir/username.$computer ] ; then + username=`cat $confdir/username.$computer` +else + username=root +fi infomsg "Backing up $computer" # Try to SSH to the computer without entering a password. -if ! ssh -n -o NumberOfPasswordPrompts=0 root@$computer /bin/true; then +if ! ssh -n -o NumberOfPasswordPrompts=0 $username@$computer /bin/true; then diemsg "Could not use passwordless SSH." fi +# Find the home directory of the backup user +remotehome=`ssh -n $username@$computer "echo ~"` +if [ -z "$remotehome" ]; then + diemsg "Could not expand ~ for user $username" +fi + # Check if pigz is available -if ssh -n root@$computer "pigz -V 2>/dev/null"; then +if ssh -n $username@$computer "pigz -V 2>/dev/null"; then compressor=pigz else infomsg "pigz missing; falling back to gzip." @@ -178,37 +235,37 @@ else fi # Check dump bit in fstab -filesystems=`ssh -n root@$computer "cat /etc/fstab" \ +filesystems=`ssh -n $username@$computer "cat /etc/fstab" \ | grep -v nfs \ | grep -v "^#" \ | grep -v "^$" \ | awk '{ if ( $(NF-1) != "0" ) print $2}' ` # Clean up our dir at this client -if ! ssh root@$computer "rm -r /root/.backup ; mkdir -m 700 /root/.backup"; then - diemsg "Could not create backup staging area at $computer:/root/.backup" +if ! ssh -n $username@$computer "rm -r $remotehome/.backup ; mkdir -m 700 $remotehome/.backup"; then + diemsg "Could not create backup staging area at $computer:$remotehome/.backup" fi # Try to copy $exclude to $computer -if ! scp $exclude root@$computer:/root/.backup/exclude > /dev/null; then +if ! scp $exclude $username@$computer:$remotehome/.backup/exclude >&2; then diemsg "Could not copy exclude.txt to $computer" fi # Try to copy preeexec and postexec if they exist if [ -f $confdir/preexec.$computer ]; then - if ! scp $confdir/preexec.$computer root@$computer:/root/.backup/preexec >&2; then - diemsg "Could not copy preexec.$computer to $computer:/root/.backup/preexec" + if ! scp $confdir/preexec.$computer $username@$computer:$remotehome/.backup/preexec >&2; then + diemsg "Could not copy preexec.$computer to $computer:$remotehome/.backup/preexec" fi fi if [ -f $confdir/postexec.$computer ]; then - if ! scp $confdir/postexec.$computer root@$computer:/root/.backup/postexec >&2; then - diemsg "Could not copy postexec.$computer to $computer:/root/.backup/postexec" + if ! scp $confdir/postexec.$computer $username@$computer:$remotehome/.backup/postexec >&2; then + diemsg "Could not copy postexec.$computer to $computer:$remotehome/.backup/postexec" fi fi # Try to run preexec if it exists -if ! ssh root@$computer "[ ! -f /root/.backup/preexec ] || /bin/bash -x /root/.backup/preexec" >&2; then - diemsg "Could not run $computer:/root/.backup/preexec" +if ! ssh -n $username@$computer "[ ! -f $remotehome/.backup/preexec ] || /bin/bash -x $remotehome/.backup/preexec" >&2; then + diemsg "Could not run $computer:$remotehome/.backup/preexec" fi for filesystem in $filesystems; do @@ -242,7 +299,7 @@ for filesystem in $filesystems; do fullbackup_min_for_this_machine=$daysbetweenfull fi - if [ -z "`find $storagedir/$computer/$sfilesystem/full/ -name \*tgz -mtime -$fullbackup_min_for_this_machine`" ]; then + if [ -z "`find $storagedir/$computer/$sfilesystem/full/ -name \*tgz\* -mtime -$fullbackup_min_for_this_machine`" ]; then backuplevel=full fi @@ -250,7 +307,11 @@ for filesystem in $filesystems; do cd $storagedir/$computer/$sfilesystem/$backuplevel || diemsg "$storagedir/$computer/$sfilesystem/$backuplevel does not exist" # Perform the actual backup - backup + if [ "$username" = "root" ]; then + backup + else + nonprivbackup + fi # Check if this box has a custom number of full backups if [ -f $confdir/maksfulle.$computer.$sfilesystem ]; then @@ -262,14 +323,14 @@ for filesystem in $filesystems; do fi # Delete old full backups - for full in `ls -1t $storagedir/$computer/$sfilesystem/full/*tgz | tail -n +$mf`; do + for full in `ls -1t $storagedir/$computer/$sfilesystem/full/*tgz* | tail -n +$mf`; do prefix=`echo $full | sed "s/\.[^.]*$//"` infomsg "$computer:$filesystem Deleting full backup $prefix" rm $prefix* done # Delete incremental backups older than the oldest full backup - oldf=`ls -t1 $storagedir/$computer/$sfilesystem/full/*tgz | tail -1` + oldf=`ls -t1 $storagedir/$computer/$sfilesystem/full/*tgz* | tail -1` find \ $storagedir/$computer/$sfilesystem/daglig \ -type f \ @@ -279,12 +340,12 @@ for filesystem in $filesystems; do done # Try to run postexec if it exist -if ! ssh root@$computer "[ ! -f /root/.backup/postexec ] || /bin/bash -x /root/.backup/postexec" >&2; then - diemsg "Could not run $computer:/root/.backup/postexec" +if ! ssh -n $username@$computer "[ ! -f $remotehome/.backup/postexec ] || /bin/bash -x $remotehome/.backup/postexec" >&2; then + diemsg "Could not run $computer:$remotehome/.backup/postexec" fi # Use warn so that we will get an email even on success. -warnmsg "$computer: Backup completed successfully." +warnmsg "Backup completed successfully." # Remove lockfile rm $LOCKFILE