git.sesse.net Git - backup.sh/blob - backup.sh

   1 #!/bin/bash
   2 # backup.sh
   3 #
   4 # Backup up Unix-like computers
   5 #
   6 # itk@samfundet.no
   7 #
   8 # This script was first checked into RCS in 2000, and has since backed up a few
   9 # servers almost every day, whenever not broken.
  10 #
  11 # It's a bit peculiar, but quite effective. The following is worth keeping in
  12 # mind when hacking:
  13 #
  14 # Because of the way we run this script from cron, informational logging goes
  15 # to stderr and warnings and errors to stdout. This way, we get mail whenever
  16 # stuff fails, and other output is redirected to the log file.
  17 #
  18 # The script is halfway Norwegian and halfway English. Newer modifications are
  19 # in English, we should probably stick to that.
  20
  21 # The computer to backup.
  22 computer=$1
  23
  24 # Configuration
  25 LOCKFILE=/home/backup/backuprun.lockfile.$computer
  26 confdir=/home/backup/conf/              # Configuration files
  27 storagedir=/backup                      # Where we keep backups
  28 maxnumfull=3                            # Number of full backups
  29 daysbetweenfull=30                      # Days between full backups
  30 daysbetweenfullforfixed=32              # Days between full backups for machines with fixed
  31                                         # full-backup date (in case the fixed day gets missed)
  32 DAY_OF_MONTH=`date "+%d" | sed s/^0//`  # from 1 to 31
  33 DATE=`date "+%Y%m%d%H%M"`               #format: touch
  34 DATEs=`date "+%Y-%m-%d %H:%M"`          #format: tar
  35
  36 # Exclude-pattern
  37 exclude=$confdir/exclude
  38 [ ! -f $exclude ] && printf "tmp\ncore\n" > $exclude
  39
  40 if [ -z "$computer" ]; then
  41         echo "Usage: $0 COMPUTER"
  42         exit 1
  43 fi
  44
  45 # Die more or less gracefully (at least notify user)
  46 die() {
  47         echo `date`": Something nasty happened."
  48         rm $LOCKFILE
  49         exit 1
  50 }
  51
  52 # Trap C-c and kill
  53 trap die SIGINT SIGTERM
  54
  55 # Don't start if we're already running
  56 if [ -e $LOCKFILE ]; then
  57         echo `date`": $LOCKFILE exists for pid `cat $LOCKFILE`, exiting."
  58         echo `date`": $LOCKFILE exists for pid `cat $LOCKFILE`, exiting." >&2
  59         exit 1
  60 fi;
  61 touch $LOCKFILE
  62 echo $$ > $LOCKFILE
  63
  64 echo `date`": Backup run starting" >&2
  65
  66 umask 027
  67
  68
  69 backup()
  70 {
  71         echo -n `date` >&2
  72         printf " $computer: $computer:$filesystem $backuplevel backup\n" >&2
  73
  74         SNARFILE="${storagedir}/${computer}/${sfilesystem}/.incremental.snar"
  75         incrementalsnar="/root/.backup/${sfilesystem}.snar"
  76         if [ "$backuplevel" = "daglig" ]
  77         then
  78                 # If incremental backup, we need to copy the incremental status to $computer.
  79                 # If it does not exist, whine a bit and then run date-based instead.
  80                 if [ -s "$SNARFILE" ]; then
  81                         if ! scp $SNARFILE root@$computer:$incrementalsnar; then
  82                                 echo `date`" $computer: Could not copy .incremental.snar to $computer"
  83                                 die
  84                         fi
  85                         lastcmd="--listed-incremental=$incrementalsnar --no-check-device"
  86                 else
  87                         echo `date`" $computer: Missing incremental.snar for $filesystem, doing date-based backup instead until next full backup" >&2
  88                         incrementalsnar=""
  89                         lastd=`cat ../.date`
  90                         lastcmd="--newer='$lastd'"
  91                 fi
  92         else
  93                 lastcmd="--listed-incremental=$incrementalsnar"
  94         fi
  95
  96         # We try to run tar on the remote computer
  97         #    c create archive
  98         #    C change to directory first
  99         #    . where to start taring (see C)
 100         #    $lastcmd only files newer than this
 101         #    --one-file-system don't traverse file systems
 102         #    --use-compress-program compress it using gzip or pigz
 103         #    --exclude-from file to get exclusion pattern from
 104         #    Pipe the stuff over ssh to ourselves, run pee to cat the contents to a
 105         #    file and build a file list, respectivly.
 106         TARFILE=$DATE.tmp
 107         TARCMD="ssh root@$computer \"nice -n 19 ionice -c3 tar --one-file-system --use-compress-program $compressor -cf - -C $filesystem . $lastcmd \
 108                 --exclude-from=/root/.backup/exclude\" | pee \"cat > $TARFILE\" \"tar tzvf -\""
 109         echo `date`" $computer: Running $TARCMD" >&2
 110         eval $TARCMD > $DATE.idx
 111
 112         if [ $? -eq 0 ] && [ -s $TARFILE ]; then
 113                 # File is >0 in size and neither cat or tar tzvf failed; we assume it worked.
 114
 115                 if ! scp root@$computer:$incrementalsnar $SNARFILE.tmp; then
 116                         echo `date`" $computer: Could not copy .incremental.snar from $computer"
 117                         die
 118                 fi
 119
 120                 # Move tar file in place
 121                 mv $TARFILE $DATE.tgz
 122                 mv $SNARFILE.tmp $SNARFILE
 123
 124                 # Update timestamp
 125                 echo $DATEs > ../.date && touch -t $DATE ../.date
 126
 127                 # Make a sorted file list as well
 128                 sort -k6 < $DATE.idx > $DATE.sdx
 129
 130                 # Fix permissions
 131                 chmod 600 *tgz                      #only for us
 132                 chmod 644 *sdx *idx 2>/dev/null     #everyone can read
 133
 134                 # Let the remote computer know that we ran a successful backup (for nagios)
 135                 ssh root@$computer touch $filesystem/.lastbackup
 136         else
 137                 # Something wrong happened.
 138                 rm $TARFILE
 139                 echo `date`" $computer: $TARFILE empty. $backuplevel backup of $computer:$filesystem failed and deleted"
 140                 die
 141         fi
 142 }
 143
 144 # Check that the target filesystem is mounted (actually check that it's not
 145 # the root filesystem)
 146 if [ ! -d "$storagedir/$computer" ]; then
 147         echo `date`" $computer: Target filesystem ($storagedir/$computer) does not exist. Aborting"
 148         die
 149 fi
 150 rootfilesystem=`df -P /`
 151 targetfilesystem=`df -P "$storagedir/$computer"`
 152 if [ "$rootfilesystem" == "$targetfilesystem" ]; then
 153         echo `date`" $computer: Target filesystem ($storagedir/$computer) was mounted on /. Aborting"
 154         die
 155 fi
 156
 157 echo `date`" $computer: Backing up $computer" >&2
 158
 159 # Try to SSH to the computer without entering a password.
 160 if ! `ssh -n -o NumberOfPasswordPrompts=0 root@$computer /bin/true`; then
 161         echo `date`" $computer: Could not use passwordless SSH."
 162         die
 163 fi
 164
 165 # Check if pigz is available
 166 if ssh -n root@$computer "pigz -V 2>/dev/null"; then
 167         compressor=pigz
 168 else
 169         echo `date`" $computer: pigz missing; falling back to gzip."
 170         compressor=gzip
 171 fi
 172
 173 # Check dump bit in fstab
 174 filesystems=`ssh -n root@$computer "cat /etc/fstab" \
 175         | grep -v nfs \
 176         | grep -v "^#" \
 177         | grep -v "^$" \
 178         | awk '{ if ( $(NF-1) != "0" ) print $2}' `
 179
 180 # Clean up our dir at this client
 181 if ! ssh root@$computer "rm -r /root/.backup ; mkdir -m 700 /root/.backup"; then
 182         echo `date`" $computer: Could not create backup staging area at $computer:/root/.backup"
 183         die
 184 fi
 185
 186 # Try to copy $exclude to $computer
 187 if ! scp $exclude root@$computer:/root/.backup/exclude > /dev/null; then
 188         echo `date`" $computer: Could not copy exclude.txt to $computer"
 189         die
 190 fi
 191
 192 # Try to copy preeexec and postexec if they exist
 193 if [ -f $confdir/preexec.$computer ]; then
 194         if ! scp $confdir/preexec.$computer  root@$computer:/root/.backup/preexec >&2; then
 195                 echo `date`" $computer: Could not copy preexec.$computer to $computer:/root/.backup/preexec"
 196                 die
 197         fi
 198 fi
 199 if [ -f $confdir/postexec.$computer ]; then
 200         if ! scp $confdir/postexec.$computer root@$computer:/root/.backup/postexec >&2; then
 201                 echo `date`" $computer: Could not copy postexec.$computer to $computer:/root/.backup/postexec"
 202                 die
 203         fi
 204 fi
 205
 206 # Try to run preexec if it exists
 207 if ! ssh root@$computer "[ ! -f /root/.backup/preexec ] || /bin/bash -x /root/.backup/preexec" >&2; then
 208         echo `date`" $computer: Could not run $computer:/root/.backup/preexec"
 209         die
 210 fi
 211
 212 for filesystem in $filesystems
 213 do
 214         # Remove / and $ (for Windows) for path names
 215         sfilesystem=`echo $filesystem | tr '\/\$' '__'`
 216
 217         # Prepare storage area
 218         mkdir -m 755 -p $storagedir/$computer/$sfilesystem/full 2>/dev/null
 219         mkdir -m 755 -p $storagedir/$computer/$sfilesystem/daglig 2>/dev/null
 220
 221         # Default backuplevel
 222         backuplevel=daglig
 223
 224         if [ ! -f $storagedir/$computer/$sfilesystem/.date ]
 225         then
 226                 # Take the first full backup of this filesystem on this computer
 227                 backuplevel=full
 228                 echo $DATEs > $storagedir/$computer/$sfilesystem/.date
 229         fi
 230
 231         # Check if we want a full backup
 232         if [ -f $confdir/fastfullbackupdag.$computer.$sfilesystem ]; then
 233                 fullbackup_min_for_this_machine=$daysbetweenfullforfixed
 234                 if [ "$DAY_OF_MONTH" = "`cat $confdir/fastfullbackupdag.$computer.$sfilesystem`" ]; then
 235                         backuplevel=full
 236                 fi
 237         elif [ -f $confdir/fastfullbackupdag.$computer ]; then
 238                 fullbackup_min_for_this_machine=$daysbetweenfullforfixed
 239                 if [ "$DAY_OF_MONTH" = "`cat $confdir/fastfullbackupdag.$computer`" ]; then
 240                         backuplevel=full
 241                 fi
 242         else
 243                 fullbackup_min_for_this_machine=$daysbetweenfull
 244         fi
 245
 246         if [ -z "`find $storagedir/$computer/$sfilesystem/full/ -name \*tgz -mtime -$fullbackup_min_for_this_machine`" ]; then
 247                 backuplevel=full
 248         fi
 249
 250         # We want to be here
 251         cd $storagedir/$computer/$sfilesystem/$backuplevel || die
 252
 253         # Perform the actual backup
 254         backup
 255
 256         # Check if this box has a custom number of full backups
 257         if [ -f $confdir/maksfulle.$computer ] ; then
 258                 mf=$((`cat $confdir/maksfulle.$computer`+1))
 259         else
 260                 mf=$(($maxnumfull+1))
 261         fi
 262
 263         # Delete old full backups
 264         for full in `ls -1t $storagedir/$computer/$sfilesystem/full/*tgz | tail -n +$mf`
 265         do
 266                 prefix=`echo $full | sed "s/\.[^.]*$//"`
 267                 echo `date`": $computer:$filesystem Deleting full backup $prefix" >&2
 268                 rm $prefix*
 269         done
 270
 271         # Delete incremental backups older than the oldest full backup
 272         oldf=`ls -t1 $storagedir/$computer/$sfilesystem/full/*tgz | tail -1`
 273         find \
 274                 $storagedir/$computer/$sfilesystem/daglig \
 275                 -type f \
 276                 \! -newer $oldf \
 277                 -printf "`date`: $computer: Deleting old incremental backup: %p\n" \
 278                 -exec rm {} \; >&2
 279 done
 280
 281 # Try to run postexec if it exist
 282 if ! ssh root@$computer "[ ! -f /root/.backup/postexec ] || /bin/bash -x /root/.backup/postexec" >&2; then
 283         echo `date`" $computer: Could not run $computer:/root/.backup/postexec"
 284         die
 285 fi
 286
 287 # Remove lockfile
 288 rm $LOCKFILE