]> git.sesse.net Git - backup.sh/blobdiff - backup.sh
Add a wrapper script for running everything in parallel again and output the logs...
[backup.sh] / backup.sh
index 307a14f23aae338f15b76d4c795aa0823ccdd180..6352fcebdff1b3a5eef94053292b9f58aea60ddc 100755 (executable)
--- 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.
@@ -28,9 +25,9 @@ computer=$1
 LOCKFILE=/home/backup/backuprun.lockfile.$computer
 confdir=/home/backup/conf/              # Configuration files
 storagedir=/backup                      # Where we keep backups
-maksantallfulle=3                       # Number of full backups
-dagermellomfulle=30                     # Days between full backups
-dagermellomfulleforfast=32              # Days between full backups for machines with fixed
+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
@@ -47,251 +44,247 @@ fi
 
 # 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
 if [ -e $LOCKFILE ]; then
-       echo `date`": $LOCKFILE exists for pid `cat $LOCKFILE`, exiting."
-       echo `date`": $LOCKFILE exists for pid `cat $LOCKFILE`, exiting." >&2
-       exit 1
+       diemsg "$LOCKFILE exists for pid `cat $LOCKFILE`."
 fi;
 touch $LOCKFILE
 echo $$ > $LOCKFILE
 
-echo `date`": Backup run starting" >&2
+infomsg "Backup run starting"
 
 umask 027
 
-
 backup()
 {
+       infomsg "$computer:$filesystem $backuplevel backup"
+
+       localsnar="${storagedir}/${computer}/${sfilesystem}/.incremental.snar"
+       remotesnar="/root/.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
+                               diemsg "Could not copy .incremental.snar to $computer"
+                       fi
+                       lastcmd="--listed-incremental=$remotesnar --no-check-device"
+               else
+                       infomsg "Missing incremental.snar for $filesystem, doing date-based backup instead until next full backup"
+                       remotesnar=""
+                       lastd=`cat ../.date`
+                       lastcmd="--newer='$lastd'"
+               fi
+       else
+               # 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
+
+       # We try to run tar on the remote computer
+       #    c create archive
+       #    C change to directory first
+       #    . where to start taring (see C)
+       #    $lastcmd only files newer than this
+       #    --one-file-system don't traverse file systems
+       #    --use-compress-program compress it using gzip or pigz
+       #    --exclude-from file to get exclusion pattern from
+       #    Pipe the stuff over ssh to ourselves, run pee to cat the contents to a
+       #    file and build a file list, respectivly.  
+       TARFILE=$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 -\""
+       infomsg "Running $TARCMD"
+       eval $TARCMD > $DATE.idx
+
+       if [ $? -eq 0 ] && [ -s $TARFILE ]; then
+               # File is >0 in size and neither cat or tar tzvf failed; we assume it worked.
+               if [ "$remotesnar" ]; then
+                       if ! scp root@$computer:$remotesnar $localsnar.tmp; then
+                               diemsg "Could not copy .incremental.snar from $computer"
+                       fi
+               fi
+
+               # Move tar file in place
+               mv $TARFILE $DATE.tgz
+               if [ "$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 > $DATE.sdx
+
+               # Fix permissions
+               chmod 600 *tgz                      #only for us
+               chmod 644 *sdx *idx 2>/dev/null     #everyone can read
+
+               # Let the remote computer know that we ran a successful backup (for nagios)
+               ssh root@$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
+}
 
- echo -n `date` >&2
- printf " $computer: $computer:$filesystem $backuplevel backup\n" >&2
-
- SNARFILE="${storagedir}/${computer}/${sfilesystem}/.incremental.snar"
- incrementalsnar="/root/.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 $computer"
-       die
-     fi
-     lastcmd="--listed-incremental=$incrementalsnar --no-check-device"
-   else
-     echo `date`" $computer: Missing incremental.snar for $filesystem, doing date-based backup instead until next full backup" >&2
-     incrementalsnar=""
-     lastd=`cat ../.date`
-     lastcmd="--newer='$lastd'"
-   fi
- else
-    lastcmd="--listed-incremental=$incrementalsnar"
- fi
-
- # expand the exclude-path for use with tar
- exf=`ssh root@$computer "ls ~/.backup/exclude"`
-
- # We try to run tar on the remote computer
- #    c create archive
- #    C change to directory first
- #    . where to start taring (see C)
- #    $lastcmd only files newer than this
- #    --one-file-system don't traverse file systems
- #    --use-compress-program compress it using gzip or pigz
- #    --exclude-from file to get exclusion pattern from
- #    Pipe the stuff over ssh to ourselves, run pee to cat the contents to a
- #    file and build a file list, respectivly.  
- TARFILE=$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
- 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
-  fi
-
-  # Move tar file in place
-  mv $TARFILE $DATE.tgz
-  mv $SNARFILE.tmp $SNARFILE
-
-  # Update timestamp
-  echo $DATEs > ../.date && touch -t $DATE ../.date
-
-  # Make a sorted file list as well
-  sort -k6 < $DATE.idx > $DATE.sdx
-
-  # Fix permissions
-  chmod 600 *tgz                      #only for us
-  chmod 644 *sdx *idx 2>/dev/null     #everyone can read
-
-  # Let the remote computer know that we ran a successful backup (for nagios)
-  ssh root@$computer touch $filesystem/.lastbackup
-
- else
-  # Something wrong happened.
-  rm $TARFILE
-  echo `date`" $computer: $TARFILE empty. $backuplevel backup of $computer:$filesystem failed and deleted"
-  die
- fi
+# Check that the target filesystem is mounted (actually check that it's not
+# the root filesystem)
+if [ ! -d "$storagedir/$computer" ]; then
+       diemsg "Target filesystem ($storagedir/$computer) does not exist."
+fi
+rootfilesystem=`df -P /`
+targetfilesystem=`df -P "$storagedir/$computer"`
+if [ "$rootfilesystem" == "$targetfilesystem" ]; then
+       diemsg "Target filesystem ($storagedir/$computer) was mounted on /."
+fi
 
-}
+infomsg "Backing up $computer"
 
- # 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/$computer) does not exist. Aborting"
-       die
- fi
- rootfilesystem=`df -P /`
- targetfilesystem=`df -P "$storagedir/$computer"`
- if [ "$rootfilesystem" == "$targetfilesystem" ]; then
-       echo `date`" $computer: Target filesystem ($storagedir/$computer) was mounted on /. Aborting"
-       die
- fi
-
- echo `date`" $computer: Backing up $computer" >&2
-
- # Try to SSH to the computer without entering a password.
- if ! `ssh -n -o NumberOfPasswordPrompts=0 root@$computer /bin/true`; then
-  echo `date`" $computer: Could not use passwordless SSH."
-  die
- fi
-
- # Check if pigz is available
- if ssh -n root@$computer "pigz -V 2>/dev/null"; then
-   compressor=pigz
- else
-   echo `date`" $computer: pigz missing; falling back to gzip."
-   compressor=gzip
- fi
- # Check dump bit in fstab
- filesystems=`ssh -n root@$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
- 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
- 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
-   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 $computer:~/.backup/postexec"
-     die
-   fi
- fi
-
- # Try to run preexec if it exists
- if ! ssh root@$computer "[ ! -f ~/.backup/preexec ] || /bin/bash -x ~/.backup/preexec" >&2; then
-       echo `date`" $computer: Could not run $computer:~/.backup/preexec"
-       die
- fi
-
-
- for filesystem in $filesystems
- do
-  # Remove / and $ (for Windows) for path names
-  sfilesystem=`echo $filesystem | tr '\/\$' '__'`
-
-  # Prepare storage area
-  mkdir -m 755 -p $storagedir/$computer/$sfilesystem/full 2>/dev/null
-  mkdir -m 755 -p $storagedir/$computer/$sfilesystem/daglig 2>/dev/null
-
-  # Default backuplevel
-  backuplevel=daglig
-
-  if [ ! -f $storagedir/$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
-  if [ -f $confdir/fastfullbackupdag.$computer.$sfilesystem ]; then
-    fullbackup_min_for_this_machine=$dagermellomfulleforfast
-    if [ "$DAY_OF_MONTH" = "`cat $confdir/fastfullbackupdag.$computer.$sfilesystem`" ]; then
-      backuplevel=full
-    fi
-  elif [ -f $confdir/fastfullbackupdag.$computer ]; then
-    fullbackup_min_for_this_machine=$dagermellomfulleforfast
-    if [ "$DAY_OF_MONTH" = "`cat $confdir/fastfullbackupdag.$computer`" ]; then
-      backuplevel=full
-    fi
-  else
-    fullbackup_min_for_this_machine=$dagermellomfulle 
-  fi
-
-  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
-
-  # Perform the actual backup
-  backup
-
-  # Check if this box has a custom number of full backups
-  if [ -f $confdir/maksfulle.$computer ] ; then
-    mf=$((`cat $confdir/maksfulle.$computer`+1))
-  else
-    mf=$(($maksantallfulle+1))
-  fi
-
-  # Delete old full backups
-  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
-   rm $prefix*
-  done
-
-  # Delete incremental backups older than the oldest full backup
-  oldf=`ls -t1 $storagedir/$computer/$sfilesystem/full/*tgz | tail -1`
-  find \
-     $storagedir/$computer/$sfilesystem/daglig \
-     -type f \
-     \! -newer $oldf \
-     -printf "`date`: $computer: Deleting old incremental backup: %p\n" \
-     -exec rm {} \; >&2
- done
-
- # Try to run postexec if it exist
- if ! ssh root@$computer "[ ! -f ~/.backup/postexec ] || /bin/bash -x ~/.backup/postexec" >&2; then
-       echo `date`" $computer: Could not run $computer:~/.backup/postexec"
-       die
- fi
+# Try to SSH to the computer without entering a password.
+if ! ssh -n -o NumberOfPasswordPrompts=0 root@$computer /bin/true; then
+       diemsg "Could not use passwordless SSH."
+fi
+
+# Check if pigz is available
+if ssh -n root@$computer "pigz -V 2>/dev/null"; then
+       compressor=pigz
+else
+       infomsg "pigz missing; falling back to gzip."
+       compressor=gzip
+fi
+
+# Check dump bit in fstab
+filesystems=`ssh -n root@$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"
+fi
+
+# Try to copy $exclude to $computer
+if ! scp $exclude root@$computer:/root/.backup/exclude > /dev/null; 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"
+       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 $computer:/root/.backup/postexec"
+       fi
+fi
+
+# Try to run preexec if it exists
+if ! ssh root@$computer "[ ! -f /root/.backup/preexec ] || /bin/bash -x /root/.backup/preexec" >&2; then
+       diemsg "Could not run $computer:/root/.backup/preexec"
+fi
+
+for filesystem in $filesystems; do
+       # Remove / and $ (for Windows) for path names
+       sfilesystem=`echo $filesystem | tr '\/\$' '__'`
+
+       # Prepare storage area
+       mkdir -m 755 -p $storagedir/$computer/$sfilesystem/full 2>/dev/null
+       mkdir -m 755 -p $storagedir/$computer/$sfilesystem/daglig 2>/dev/null
+
+       # Default backuplevel
+       backuplevel=daglig
+
+       if [ ! -s $storagedir/$computer/$sfilesystem/.date ]; then
+               # Take the first full backup of this filesystem on this computer
+               backuplevel=full
+       fi
+
+       # Check if we want a full backup
+       if [ -f $confdir/fastfullbackupdag.$computer.$sfilesystem ]; then
+               fullbackup_min_for_this_machine=$daysbetweenfullforfixed
+               if [ "$DAY_OF_MONTH" = "`cat $confdir/fastfullbackupdag.$computer.$sfilesystem`" ]; then
+                       backuplevel=full
+               fi
+       elif [ -f $confdir/fastfullbackupdag.$computer ]; then
+               fullbackup_min_for_this_machine=$daysbetweenfullforfixed
+               if [ "$DAY_OF_MONTH" = "`cat $confdir/fastfullbackupdag.$computer`" ]; then
+                       backuplevel=full
+               fi
+       else
+               fullbackup_min_for_this_machine=$daysbetweenfull 
+       fi
+
+       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 || diemsg "$storagedir/$computer/$sfilesystem/$backuplevel does not exist"
+
+       # Perform the actual backup
+       backup
+
+       # Check if this box has a custom number of full backups
+       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
+               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`
+       find \
+               $storagedir/$computer/$sfilesystem/daglig \
+               -type f \
+               \! -newer $oldf \
+               -printf "`date`: $computer: Deleting old incremental backup: %p\n" \
+               -exec rm {} \; >&2
+done
+
+# Try to run postexec if it exist
+if ! ssh root@$computer "[ ! -f /root/.backup/postexec ] || /bin/bash -x /root/.backup/postexec" >&2; then
+       diemsg "Could not run $computer:/root/.backup/postexec"
+fi
+
+# Use warn so that we will get an email even on success.
+warnmsg "$computer: Backup completed successfully."
 
 # Remove lockfile
 rm $LOCKFILE