X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=backup.sh;h=7fc7ffb0018fecd951b3d64249637aa5ea28cc71;hb=fcc343a7f7be5c6f9348b05b7dbdd429d955d4d0;hp=448972709d69d0d1ab4f413310244bd51c1a4386;hpb=3bc98c513aa4c45391a1b57b7cf42e0ebbb7faf9;p=backup.sh diff --git a/backup.sh b/backup.sh index 4489727..7fc7ffb 100755 --- a/backup.sh +++ b/backup.sh @@ -11,9 +11,6 @@ # It's a bit peculiar, but quite effective. The following is worth keeping in # mind when hacking: # -# KISS. And by that I mean really simple. Only regular Unix commands, please -# (but GNU extensions are used and moreutils has to be installed). -# # Because of the way we run this script from cron, informational logging goes # to stderr and warnings and errors to stdout. This way, we get mail whenever # stuff fails, and other output is redirected to the log file. @@ -21,83 +18,95 @@ # 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() { - echo `date`": Something nasty happened. Ending, and not deleting lockfile $LOCKFILE." + warnmsg "Aborting." + rm $LOCKFILE exit 1 } +diemsg() { + warnmsg "$@" + die +} +warnmsg() { + echo `date`": $computer: $@" + echo `date`": $computer: $@" >&2 +} +infomsg() { + echo `date`": $computer: $@" >&2 +} # 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 - echo `date`": $LOCKFILE exists for pid `cat $LOCKFILE`, exiting." - echo `date`": $LOCKFILE exists for pid `cat $LOCKFILE`, exiting." >&2 + warnmsg "$LOCKFILE exists for pid `cat $LOCKFILE`." exit 1 fi; touch $LOCKFILE echo $$ > $LOCKFILE -echo `date`": Backup run starting" >&2 +infomsg "Backup run starting" umask 027 - backup() { - echo -n `date` >&2 - printf " $computer: $computer:$filesystem $backuplevel backup\n" >&2 + infomsg "$computer:$filesystem $backuplevel backup" - SNARFILE="${storagedir}/${computer}/${sfilesystem}/.incremental.snar" - incrementalsnar="/root/.backup/${sfilesystem}.snar" - if [ "$backuplevel" = "daglig" ] - then + localsnar="${storagedir}/${computer}/${sfilesystem}/.incremental.snar" + remotesnar="$remotehome/.backup/${sfilesystem}.snar" + if [ "$backuplevel" = "daglig" ]; then # If incremental backup, we need to copy the incremental status to $computer. # If it does not exist, whine a bit and then run date-based instead. - if [ -s "$SNARFILE" ]; then - if ! scp $SNARFILE root@$computer:$incrementalsnar; then - echo `date`" $computer: Could not copy .incremental.snar to $computer" - die + if [ -s "$localsnar" ]; then + if ! scp $localsnar $username@$computer:$remotesnar >&2; then + diemsg "Could not copy .incremental.snar to $computer" fi - lastcmd="--listed-incremental=$incrementalsnar --no-check-device" + lastcmd="--listed-incremental=$remotesnar --no-check-device" else - echo `date`" $computer: Missing incremental.snar for $filesystem, doing date-based backup instead until next full backup" >&2 - incrementalsnar="" + infomsg "Missing incremental.snar for $filesystem, doing date-based backup instead until next full backup" + remotesnar="" lastd=`cat ../.date` lastcmd="--newer='$lastd'" fi else - lastcmd="--listed-incremental=$incrementalsnar" + # Note: Since we rm-ed .backup previously, and we don't scp over + # anything in this branch, this file will not exist, and we + # will be doing a full backup (but we want the output of it). + lastcmd="--listed-incremental=$remotesnar" fi - # expand the exclude-path for use with tar - exf=`ssh root@$computer "ls ~/.backup/exclude"` - # We try to run tar on the remote computer # c create archive # C change to directory first @@ -109,113 +118,157 @@ backup() # Pipe the stuff over ssh to ourselves, run pee to cat the contents to a # file and build a file list, respectivly. TARFILE=$DATE.tmp - TARCMD="ssh root@$computer \"nice -n 19 ionice -c3 tar --one-file-system --use-compress-program $compressor -cf - -C $filesystem . $lastcmd \ - --exclude-from=$exf\" | pee \"cat > $TARFILE\" \"tar tzvf -\"" - echo `date`" $computer: Running $TARCMD" >&2 + 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 if [ $? -eq 0 ] && [ -s $TARFILE ]; then # File is >0 in size and neither cat or tar tzvf failed; we assume it worked. - - if ! scp root@$computer:$incrementalsnar $SNARFILE.tmp; then - echo `date`" $computer: Could not copy .incremental.snar from $computer" - die + + if [ "$remotesnar" ]; then + if ! scp $username@$computer:$remotesnar $localsnar.tmp >&2; then + diemsg "Could not copy .incremental.snar from $computer" + fi fi # Move tar file in place mv $TARFILE $DATE.tgz - mv $SNARFILE.tmp $SNARFILE + if [ "$remotesnar" ]; then + mv $localsnar.tmp $localsnar + fi # 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 - echo `date`" $computer: $TARFILE empty. $backuplevel backup of $computer:$filesystem failed and deleted" - die + diemsg "tar failed or $TARFILE empty. $backuplevel backup of $computer:$filesystem failed and deleted" fi } # Check that the target filesystem is mounted (actually check that it's not # the root filesystem) if [ ! -d "$storagedir/$computer" ]; then - echo `date`" $computer: Target filesystem ($storagedir/$computer) does not exist. Aborting" - die + diemsg "Target filesystem ($storagedir/$computer) does not exist." fi rootfilesystem=`df -P /` targetfilesystem=`df -P "$storagedir/$computer"` if [ "$rootfilesystem" == "$targetfilesystem" ]; then - echo `date`" $computer: Target filesystem ($storagedir/$computer) was mounted on /. Aborting" - die + diemsg "Target filesystem ($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 -echo `date`" $computer: Backing up $computer" >&2 +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 - echo `date`" $computer: Could not use passwordless SSH." - die +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 - echo `date`" $computer: pigz missing; falling back to gzip." + infomsg "pigz missing; falling back to gzip." compressor=gzip fi # Check dump bit in fstab -filesystems=`ssh -n root@$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 ~/.backup ; mkdir -m 700 ~/.backup"; then - echo `date`" $computer: Could not create backup staging area at $computer:~/.backup" - die +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:~/.backup/exclude > /dev/null; then - echo `date`" $computer: Could not copy exclude.txt to $computer" - die +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:~/.backup/preexec >&2; then - echo `date`" $computer: Could not copy preexec.$computer to $computer:~/.backup/preexec" - die + 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:~/.backup/postexec >&2; then - echo `date`" $computer: Could not copy postexec.$computer to $computer:~/.backup/postexec" - die + 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 ~/.backup/preexec ] || /bin/bash -x ~/.backup/preexec" >&2; then - echo `date`" $computer: Could not run $computer:~/.backup/preexec" - die +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 +for filesystem in $filesystems; do # Remove / and $ (for Windows) for path names sfilesystem=`echo $filesystem | tr '\/\$' '__'` @@ -226,11 +279,9 @@ do # Default backuplevel backuplevel=daglig - if [ ! -f $storagedir/$computer/$sfilesystem/.date ] - then + if [ ! -s $storagedir/$computer/$sfilesystem/.date ]; then # Take the first full backup of this filesystem on this computer backuplevel=full - echo $DATEs > $storagedir/$computer/$sfilesystem/.date fi # Check if we want a full backup @@ -248,33 +299,38 @@ do fullbackup_min_for_this_machine=$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 # We want to be here - cd $storagedir/$computer/$sfilesystem/$backuplevel || die + 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 ] ; then + if [ -f $confdir/maksfulle.$computer.$sfilesystem ]; then + mf=$((`cat $confdir/maksfulle.$computer`+1)) + elif [ -f $confdir/maksfulle.$computer ] ; then mf=$((`cat $confdir/maksfulle.$computer`+1)) else mf=$(($maxnumfull+1)) 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 $full | sed "s/\.[^.]*$//"` - echo `date`": $computer:$filesystem Deleting full backup $prefix" >&2 + infomsg "$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/$sfilesystem/daglig \ -type f \ @@ -284,10 +340,12 @@ do done # Try to run postexec if it exist -if ! ssh root@$computer "[ ! -f ~/.backup/postexec ] || /bin/bash -x ~/.backup/postexec" >&2; then - echo `date`" $computer: Could not run $computer:~/.backup/postexec" - die +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 "Backup completed successfully." + # Remove lockfile rm $LOCKFILE