X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=backup.sh;h=2838aae5d9bd73a0faf14c74af983fe022e1079c;hb=8418414e49c7bb8e44bb07a3082f9bbb9ffa8d2b;hp=d524d5bfc4c5b5070075e1bef640585b6b6cc3ca;hpb=4af4594ad8ee33fb03b584a777f7f58a6f8b720e;p=backup.sh

diff --git a/backup.sh b/backup.sh
index d524d5b..2838aae 100755
--- a/backup.sh
+++ b/backup.sh
@@ -1,4 +1,28 @@
-#!/bin/bash
+#!/bin/sh
+
+# Locking
+LOCKFILE=/home/backup/backuprun.lockfile
+
+# Initially, we consider everything to be just fine.
+R=0
+
+# Die gracefully (ie. remove lockfile)
+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
+}
+
+# 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, exiting."
+	exit 1
+fi;
+touch $LOCKFILE
 
 #dirs
 confdir=/home/backup/conf/           #configuration files
@@ -20,21 +44,22 @@ PATH=/local/bin:$PATH:/store/bin
 export PATH
 
 #start the logfile
-date
-echo
-echo "Exclude pattern:"
-cat $exclude
-echo
-echo
+echo `date`": Backup run starting" >&2
 
 umask 027
 
 # The computers we want to back up
-unixcomputers=`cat $confdir/computers.unix`
-###
+unixcomputers=`cat $confdir/computers.unix \
+  | grep -v "^#" \
+  | grep -v "^$" `
+
+# Backup only one computer, from command line?
+if [ $1 ] ; then 
+  unixcomputers=$1
+fi;
 
 #disse bør kunne varieres fra fs til fs?
-maksantallfulle=4          #hvor mange fulle vi tar vare på
+maksantallfulle=3          #hvor mange fulle vi tar vare på
 dagermellomfulle=30        #antall dager før det er på tide med ny full
 logw=40
 ###
@@ -50,12 +75,13 @@ DATEs=`date "+%Y-%m-%d %H:%M"`       #format: tar
 backup()
 {
 
- printf "%-${logw}s %s\n" "$computer:$filesystem" "$backuplevel backup"
+ echo -n `date` >&2
+ printf " $computer: %-${logw}s %s\n" "$computer:$filesystem" "$backuplevel backup" >&2
 
  if [ "$backuplevel" = "daglig" ] || [ "$backuplevel" = "incremental" ]
  then
   lastd=`cat ../.date`
-  lastcmd="--newer=\"$lastd\""
+  lastcmd="--newer='$lastd'"
  else
   lastcmd=""
  fi
@@ -69,19 +95,26 @@ backup()
  fi
 
  #expand the exclude-path for use with tar
- exf=`ssh $computer "ls ~/.backup/exclude"`
+ exf=`ssh root@$computer "ls ~/.backup/exclude"`
 
  #We try to run tar on the remote computer
- #    C is instead of 'cd $filesystem'
- #    - is 'this directory'
- #    l is 'stay in local filesystem
- if `ssh $computer "$tar cfCl - $filesystem . $lastcmd \
-      --exclude-from=$exf \
-        | gzip" \
-   |  dd of=$DATE.tmp 2>/dev/null` 
- then
-  #perhaps it did work
-  mv $DATE.tmp $DATE.tgz
+ #    z gzip it
+ #    c create archive
+ #    C change to directory first
+ #    - output to stdout (we pipe to gzip, then to dd)
+ #    . where to start taring (see C)
+ #    $lastcmd only files newer than this
+ #    --exclude-from file to get exclusion pattern from
+ #    pipe to gzip, which in turn pipes over the ssh-stream
+ #    ..to dd, to output to a file. We surpress messages from dd.
+ #    And at last, redirect stderr to stdout, to get output logged.
+ TARFILE=$DATE.tmp
+ TARCMD="ssh root@$computer \"$tar --one-file-system -zcf - -C $filesystem . $lastcmd \
+	--exclude-from=$exf\" > $TARFILE"
+ echo `date`" $computer: Running $TARCMD" >&2
+
+ if eval $TARCMD; then
+  mv $TARFILE $DATE.tgz
   #make a filelist.
   #update the datefile if the filelist is ok.
   tar tvfz $DATE.tgz > $DATE.idx 2>&1 &&
@@ -97,8 +130,11 @@ backup()
 
  else
   #it did not work
-  rm $DATE.tmp 2>/dev/null
-  echo "$backuplevel backup of $computer:$filesystem failed and deleted"
+  rm $TARFILE
+  echo `date`" $computer: $TARFILE empty. $backuplevel backup of $computer:$filesystem failed and deleted"
+
+  # We don't want to return 0
+  R=1
  fi
 
 }
@@ -106,31 +142,69 @@ backup()
 #Løper gjennom listen av unixmaskiner som vi skal ta backup av
 for computer in $unixcomputers
 do
+( 
+ 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."
+
+  # We don't want to return 0
+  R=1
+  # Do next computer
+  continue;
+ fi
  
  #tømmer variabelen for sikkerhets skyld
  filesystems=""
 
  #Sjekker nest siste felt i fstab. Om det er 0 tar vi ikke backup
- filesystems=`ssh -n $computer "cat /etc/fstab" \
+ filesystems=`ssh -n root@$computer "cat /etc/fstab" \
   | grep -v nfs \
-  | grep -v ^# \
+  | grep -v "^#" \
+  | grep -v "^$" \
   | awk '{ if ( $(NF-1) != "0" ) print $2}' `
 
  #clean up our dir at this client
- ssh $computer "rm -r ~/.backup ; mkdir -m 700 ~/.backup"
+ if ! ssh root@$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"
+  # We don't want to return 0
+  R=1
+  continue;
+ fi
 
  #try to copy $exclude to $computer
- scp $exclude $computer:~/.backup/exclude ||
-    echo "could not copy exclude.txt 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"
+  # We don't want to return 0
+  R=1
+  continue;
+ fi
 
  #try to copy preeexec and postexec if they exist
- [ -f $confdir/preexec.$computer ] &&
-    scp $confdir/preexec.$computer  $computer:~/.backup/preexec
- [ -f $confdir/postexec.$computer ] &&
-    scp $confdir/postexec.$computer $computer:~/.backup/postexec
+# TODO: Gah, clean this mess!
+ [ -f $confdir/preexec.$computer ] && (
+    scp $confdir/preexec.$computer  root@$computer:~/.backup/preexec >&2 ||
+     ( echo `date`" $computer: Could not copy preexec.$computer to $computer:~/.backup/preexec - skipping backup of $computer"
+       R=1
+       continue
+     )
+    )
+ [ -f $confdir/postexec.$computer ] && (
+    scp $confdir/postexec.$computer root@$computer:~/.backup/postexec >&2 ||
+     ( echo `date`" $computer: Could not copy postexec.$computer to $computer:~/.backup/postexec - skipping backup of $computer"
+       R=1
+       continue
+     )
+    )
 
  #try to run preexec if it exist
- ssh $computer "[ -f ~/.backup/preexec ] && /bin/bash -x ~/.backup/preexec"
+ if ! ssh root@$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
+ fi
+
 
  for filesystem in $filesystems
  do
@@ -140,7 +214,6 @@ do
   #lager det som trengs av kataloger
   mkdir -m 755 -p $storagedir/$computer/$sfilesystem/{full,daglig} 2>/dev/null
 
-  #
   echo $filesystem > ${storagedir}/${computer}/.${sfilesystem}.name 
   chmod 644 ${storagedir}/${computer}/.${sfilesystem}.name
 
@@ -155,24 +228,29 @@ do
   fi
 
   #sjekker om det er på tide med en full
-  if 
- [ -z "`find $storagedir/$computer/$sfilesystem/full/ -name \*tgz -mtime -$dagermellomfulle`" ]
-  then
+  if [ -z "`find $storagedir/$computer/$sfilesystem/full/ -name \*tgz -mtime -$dagermellomfulle`" ]; then
    backuplevel=full
   fi
  
-  #gå ned i rett katalog
-  cd $storagedir/$computer/$sfilesystem/$backuplevel
+  #gå ned i rett katalog, eller dø 
+  # TODO bør sende mail om dette skjer!
+  cd $storagedir/$computer/$sfilesystem/$backuplevel || die
 
   #perform the actual backup
   backup
 
+  # Sjekk om det skal være et annet antall fulle backuper av en boks
+  if [ -f $confdir/maksfulle.$computer ] ; then
+    mf=$((`cat $confdir/maksfulle.$computer`+1))
+  else
+    mf=$(($maksantallfulle+1))
+  fi
+
   #delete complete backups
-  mf=$(($maksantallfulle+1))
   for full in `ls -1t $storagedir/$computer/$sfilesystem/full/*tgz | tail +$mf`
   do
-   prefix=`echo $full|cut -d . -f 1`
-   echo "$computer:$filesystem sletter full $prefix (for mange)"
+   prefix=`echo $full | sed "s/\.[^.]*$//"`
+   echo `date`": $computer:$filesystem sletter full $prefix (for mange)" >&2
    rm $prefix*
   done
 
@@ -183,22 +261,28 @@ do
      -type f \
      \! -newer $oldf \
      -exec rm {} \;
-     #denne funker bare med gnu find
-     #-printf "$computer:$filesystem sletter daglig %f (for gammel)\n" \
-
  done
 
  #try to run postexec if it exist
- ssh $computer "[ -f ~/.backup/postexec ] && /bin/bash -x ~/.backup/postexec"
+ if ! ssh root@$computer "[ ! -f ~/.backup/postexec ] || /bin/bash -x ~/.backup/postexec" >&2; then
+ 	echo `date`" $computer: Could not run $computer:~/.backup/postexec"
+	R=1
+ fi
 
-done
+)  
+done &
 
-#create report
-/home/backup/report.sh $storagedir > /tmp/report.txt
-#[ ! -z $remotestat ] && scp /tmp/report.txt $remotestat
-###
+wait
 
-# print diskusage to logfile
-df -k
+# Remove lockfile
+rm $LOCKFILE
 
-date
+# 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