Remove the built-in forking support; take that from cron instead.

diff --git a/backup.sh b/backup.sh
index aceb9f3a928cb951583388ab549475f34eea19e2..ce6c6d4cac516305b35b52b14c9cc613b5d713ef 100755 (executable)
--- a/backup.sh
+++ b/backup.sh
@@ -18,11 +18,6 @@
 # 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.
 # 
-# The computers are backed up in parallel (but the filesystems on each
-# computer in serial). This is implemented highly simplistic, with forking
-# subshells. The script has no concept about threads, and stuff might go wrong,
-# which usually requires manual cleanups.
-#
 # The script is halfway Norwegian and halfway English. Newer modifications are
 # in English, we should probably stick to that.
 # 
@@ -33,8 +28,11 @@
 #   right now.
 #
 
+# The computer to backup.
+computer=$1
+
 # Configuration
-LOCKFILE=/home/backup/backuprun.lockfile
+LOCKFILE=/home/backup/backuprun.lockfile.$computer
 confdir=/home/backup/conf/              # Configuration files
 storagedir=/backup                      # Where we keep backups
 maksantallfulle=3                       # Number of full backups
@@ -49,15 +47,15 @@ DATEs=`date "+%Y-%m-%d %H:%M"`          #format: tar
 exclude=$confdir/exclude
 [ ! -f $exclude ] && printf "tmp\ncore\n" > $exclude
 
-# Initially, we consider everything to be just fine.
-R=0
+if [ -z "$computer" ]; then
+       echo "Usage: $0 COMPUTER"
+       exit 1
+fi
 
 # Die more or less gracefully (at least notify user)
 die() {
-       echo `date`": Something nasty happened, and since I fork a lot:"
-        echo `date`": I CANNOT CLEAN UP THE MESS MYSELF."
-       echo `date`": You need to get rid of lost process named stuff like $0, tar and ssh."
-       exit 255
+       echo `date`": Something nasty happened. Ending, and not deleting lockfile $LOCKFILE."
+       exit 1
 }
 
 # Trap C-c and kill
@@ -85,15 +83,6 @@ echo `date`": Backup run starting" >&2
 
 umask 027
 
-# The computers we want to back up
-unixcomputers=`cat $confdir/computers.unix \
-  | grep -v "^#" \
-  | grep -v "^$" `
-
-# Backup only one computer, from command line?
-if [ $1 ] ; then 
-  unixcomputers=$1
-fi
 
 #selve backupen
 # krever at noen variabler er satt
@@ -156,21 +145,18 @@ backup()
   echo `date`" $computer: $TARFILE empty. $backuplevel backup of $computer:$filesystem failed and deleted"
 
   # We don't want to return 0
-  R=1
+  die
  fi
 
 }
 
-for computer in $unixcomputers
-do
-( 
  # Check that the target filesystem is mounted (actually check that it's not
  # the root filesystem)
  rootfilesystem=`df -P /`
  targetfilesystem=`df -P "$storagedir/$computer"`
  if [ "$rootfilesystem" == "$targetfilesystem" ]; then
        echo `date`" $computer: Target filesystem ($storagedir/$computer) was mounted on /. Aborting"
-       continue
+       die
  fi
 
  echo `date`" $computer: Backing up $computer" >&2
@@ -180,9 +166,7 @@ do
   echo `date`" $computer: Could not use passwordless SSH."
 
   # We don't want to return 0
-  R=1
-  # Do next computer
-  continue;
+  die
  fi
  
  filesystems=""
@@ -197,38 +181,33 @@ do
  # 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 - skipping backup of $computer"
-  R=1
-  continue;
+  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 - skipping backup of $computer"
-  R=1
-  continue;
+  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 - skipping backup of $computer"
-     R=1
-     continue
+     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 - skipping backup of $computer"
-     R=1
-     continue
+     die
    fi
  fi
 
  # Try to run preexec if it exist
  if ! ssh root@$computer "[ ! -f ~/.backup/preexec ] || /bin/bash -x ~/.backup/preexec" >&2; then
        echo `date`" $computer: Could not run $computer:~/.backup/preexec - skipping backup of $computer"
-       R=1
-       continue
+       die
  fi
 
 
@@ -307,23 +286,8 @@ do
  # 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"
-       R=1
+       die
  fi
 
-) &
-done
-
-wait
-
 # Remove lockfile
 rm $LOCKFILE
-
-# Did anything go wrong?
-# (IMPORTANT NOTE: The R-business does not work as expected, since this script
-# forks itself).
-if [ $R != 0 ]; then
-       echo `date`": Backup run ended with errors, check logs."
-       exit 1
-else
-       echo `date`": Backup run ended" >&2
-fi