X-Git-Url: https://git.sesse.net/?p=backup.sh;a=blobdiff_plain;f=backup.sh;h=a76baf99d8432b6d2dcac06d77b95ebf6d4bf568;hp=b82b8bf3cb162c339f886acad220fd10de125f80;hb=0fd93f28df10578e1e485e6883b6b832daf245c5;hpb=d732ba8a3b93b3e3e52bbc2af61dcb1cba671bf9

diff --git a/backup.sh b/backup.sh
index b82b8bf..a76baf9 100755
--- a/backup.sh
+++ b/backup.sh
@@ -1,30 +1,100 @@
 #!/bin/bash
+# backup.sh
+# 
+# Backup up Unix-like computers
+# 
+# itk@samfundet.no
+#
+# This script was first checked into RCS in 2000, and has since backed up a few
+# servers almost every day, whenever not broken.
+# 
+# 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.
+# 
+# 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.
+# 
+# Known issues:
+#
+# o $R has no proper meaning because of the forking subshell stuff. We need a
+#   more robust way of returning non-zero on errors, which we probably don't
+#   right now.
+#
 
-PATH=/local/bin:$PATH
+# Configuration
+LOCKFILE=/home/backup/backuprun.lockfile
+confdir=/home/backup/conf/              # Configuration files
+storagedir=/backup                      # Where we keep backups
+maksantallfulle=3                       # Number of full backups
+dagermellomfulle=30                     # Days between full backups
+dagermellomfulleforfast=32              # Days between full backups for machines with fixed
+                                        # full-backup date (in case the fixed day gets missed)
+DAY_OF_MONTH=`date "+%d" | sed s/^0//`  # from 1 to 31
+DATE=`date "+%Y%m%d%H%M"`               #format: touch
+DATEs=`date "+%Y-%m-%d %H:%M"`          #format: tar
+
+# Exclude-pattern
+exclude=$confdir/exclude
+[ ! -f $exclude ] && printf "tmp\ncore\n" > $exclude
+
+# Initially, we consider everything to be just fine.
+R=0
+
+# 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
+}
+
+# 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
+fi;
+touch $LOCKFILE
+echo $$ > $LOCKFILE
+
+#syntax of remotestat:
+#hostname:/directory/
+remotestatf=$confdir/remotestat
+[ -f $remotestatf ] && remotestat=`cat $remotestatf`
+###
+
+PATH=/local/bin:$PATH:/store/bin
 export PATH
 
-echo -n "Starter... "
-date
+echo `date`": Backup run starting" >&2
 
-#newgrp itk
 umask 027
 
-#liste over maskinene vi tar backup av. (space mellom maskinnavn)
-computers="cirkus backup josef"
-#for debug
-#computers="josef"
-###
+# The computers we want to back up
+unixcomputers=`cat $confdir/computers.unix \
+  | grep -v "^#" \
+  | grep -v "^$" `
 
-#disse bør kunne varieres fra fs til fs
-maksantallfulle=2          #hvor mange fulle vi tar vare på
-#maksantalldaglige=30       #hvor mange daglige vi tar vare på
-dagermellomfulle=30        #antall dager før det er på tide med ny full
-###
+# Backup only one computer, from command line?
+if [ $1 ] ; then 
+  unixcomputers=$1
+fi;
 
-#lager datovariabeler
-DATE=`date "+%Y%m%d%H%M"`            #passer for bruk med touch
-DATEs=`date "+%Y-%m-%d %H:%M"`       #passer for bruk med tar
-###
 
 #selve backupen
 # krever at noen variabler er satt
@@ -32,124 +102,225 @@ DATEs=`date "+%Y-%m-%d %H:%M"`       #passer for bruk med tar
 backup()
 {
 
- if [ "$backuplevel" = "daglig" ]
+ echo -n `date` >&2
+ printf " $computer: $computer:$filesystem $backuplevel backup\n" >&2
+
+ if [ "$backuplevel" = "daglig" ] || [ "$backuplevel" = "incremental" ]
  then
-  forrige=`cat ../.dato`
-  forrigecmd=--newer=\"$forrige\"
+  lastd=`cat ../.date`
+  lastcmd="--newer='$lastd'"
  else
-  forrigecmd=""
+  lastcmd=""
  fi
 
- if `ssh $computer "cd $filesystem ;\
-     tar cfl - . $forrigecmd \
-      --exclude=\"tmp\" \
-      --exclude=\".netscape/cache\" \
-      --exclude=\"core\" \
-      --exclude=\"*.rar\" \
-      --exclude=\"*.mp3\" \
-      --exclude=\"*.MP3\" \
-      --exclude=\"*.iso\" \
-      --exclude=\"*.rar\" \
-        | gzip" \
-   |  dd of=$DATE.tmp 2>/dev/null` 
- then
-  #Det gikk kanskje bra
-  mv $DATE.tmp $DATE.tgz
-  #lage listefil
-  tar tvfz $DATE.tgz > $DATE.idx 2>&1
-  grep -v ^d $DATE.idx | sort -n -r +2 > $DATE.sdx
+ # expand the exclude-path for use with tar
+ exf=`ssh root@$computer "ls ~/.backup/exclude"`
+
+ # We try to run tar on the remote computer
+ #    z gzip it
+ #    c create archive
+ #    C change to directory first
+ #    . where to start taring (see C)
+ #    $lastcmd only files newer than this
+ #    --one-file-system don't traverse file systems
+ #    --exclude-from file to get exclusion pattern from
+ #    Pipe the stuff over ssh to ourselves, run pee to cat the contents to a
+ #    file and build a file list, respectivly.  
+ TARFILE=$DATE.tmp
+ TARCMD="ssh root@$computer \"nice -n 19 ionice -c3 tar --one-file-system -zcf - -C $filesystem . $lastcmd \
+	--exclude-from=$exf\" | pee \"cat > $TARFILE\" \"tar tzvf -\""
+ echo `date`" $computer: Running $TARCMD" >&2
+ eval $TARCMD > $DATE.idx
+
+ if [ -s $TARFILE ]; then
+  # File is >0 in size, we assume it worked.
+
+  # Move tar file in place
+  mv $TARFILE $DATE.tgz
+
+  # Update timestamp
+  echo $DATEs > ../.date && touch -t $DATE ../.date
 
-  #oppdatere datofil
-  echo $DATEs > ../.dato
-  touch -t $DATE ../.dato
+  # Make a sorted filelist as well
+  sort -k6 < $DATE.idx > $DATE.sdx
 
-  #fikse rettigheter
-  chmod 600 *tgz
-  chmod 644 *sdx *idx
+  # Fix permissions
+  chmod 600 *tgz                      #only for us
+  chmod 644 *sdx *idx 2>/dev/null     #everyone can read
 
  else
-  #Det gikk ikke bra
-  rm $DATE.tmp
-  echo "$backuplevel backup av $computer:$filesystem gikk ikke bra"
+  # Something wrong happened.
+  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
+
 }
 
-for computer in $computers
+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
+ 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."
+
+  # 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 $computer "cat /etc/fstab" \
+ # Check dump bit in fstab
+ filesystems=`ssh -n root@$computer "cat /etc/fstab" \
   | grep -v nfs \
-  | grep -v ^# \
+  | grep -v "^#" \
+  | grep -v "^$" \
   | awk '{ if ( $(NF-1) != "0" ) print $2}' `
 
- #for debugging
- #filesystems="/"
+ # 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;
+ 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;
+ 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
+   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 $computer"
+     R=1
+     continue
+   fi
+ fi
+
+ # Try to run preexec if it exist
+ if ! ssh root@$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
+ fi
+
 
  for filesystem in $filesystems
  do
-  #lager en variant uten tegnet "/"
-  sfilesystem=`echo $filesystem | tr '/' '_'`
+  # Remove / and $ (for Windows) for path names
+  sfilesystem=`echo $filesystem | tr '\/\$' '__'`
 
-  #
-  echo $filesystem > /backup/$computer/$sfilesystem.navn
-  chmod 644 /backup/$computer/$sfilesystem.navn
+  # Prepare storage area
+  mkdir -m 755 -p $storagedir/$computer/$sfilesystem/{full,daglig} 2>/dev/null
+  echo $filesystem > ${storagedir}/${computer}/.${sfilesystem}.name 
+  chmod 644 ${storagedir}/${computer}/.${sfilesystem}.name
 
-  #standard backuplevel. full og daglig er eneste muligheter
+  # Default backuplevel
   backuplevel=daglig
-  #backuplevel=full
-
-  #lager det som trengs av kataloger
-  mkdir -m 755 -p /backup/$computer/$sfilesystem/{full,daglig} 2>/dev/null
 
-  #Bør bare skje ved første gangs kjøring for dette fs
-  if [ ! -f /backup/$computer/$sfilesystem/.dato ]
+  if [ ! -f $storagedir/$computer/$sfilesystem/.date ]
   then
+   # Take the first full backup of this filesystem on this computer
    backuplevel=full
-   echo $DATEs > /backup/$computer/$sfilesystem/.dato
-   echo "$computer:$filesystem datofil laget"
-   echo "$computer:$filesystem trenger full pga ingen fra foer"
+   echo $DATEs > $storagedir/$computer/$sfilesystem/.date
   fi
 
-  #sjekker om det er på tide med en full
-  if 
- [ -z `find /backup/$computer/$sfilesystem/full/ -name \*tgz -mtime -$dagermellomfulle` ]
-  then
-   echo "$computer:$filesystem trenger full pga lenge siden sist"
+  # Check if we want a full backup - Debug statement that causes noise in cron mail:
+  # printf " $computer $filesystem: $confdir/fastfullbackupdag.$computer.$sfilesystem\n"
+  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
- 
-  #gå ned i rett katalog
-  cd /backup/$computer/$sfilesystem/$backuplevel
 
-  #selve backupen
+  # We want to be here
+  cd $storagedir/$computer/$sfilesystem/$backuplevel || die
+
+  # Perform the actual backup
   backup
 
-  echo "$computer:$filesystem sletter fulle slik at vi ikke har mer enn 2"
-  mf=$(($maksantallfulle+1))
-  for full in `ls -1t /backup/$computer/$sfilesystem/full/*tgz | tail +$mf`
+  # 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 $full|cut -d . -f 1`
-   echo "$computer:$filesystem sletter $prefix"
+   prefix=`echo $full | sed "s/\.[^.]*$//"`
+   echo `date`": $computer:$filesystem Deleting full backup $prefix" >&2
    rm $prefix*
   done
-  echo "$computer:$filesystem ferdig med slett av fulle"
 
-  echo "$computer:$filesystem slette daglige eldre enn eldste fulle"
-  eldste=`ls -t1 /backup/$computer/$sfilesystem/full/*tgz | tail -1`
+  # Delete incremental backups older than the oldest full backup
+  oldf=`ls -t1 $storagedir/$computer/$sfilesystem/full/*tgz | tail -1`
   find \
-     /backup/$computer/$sfilesystem/daglig \
-     \! -newer $eldste \
-     -print \
-     -exec rm {} \;
-  echo "$computer:$filesystem ferdig med slett av gamle daglige"
-
+     $storagedir/$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"
+	R=1
+ fi
+
+) &
 done
 
-echo -n "Ferdig... "
-date
+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