#!/bin/sh
+# TODO(sesse, 2007-12-16):
+# * Fiks så parallelliseringen fungerer igjen
+# * Bruk tee i stedet for en separat tar ztvf-prosess
+# * Sjekk hvorfor .idx-filene til UD er blitt 0 for siste fullbackuper
+
+# Locking
+LOCKFILE=/home/backup/backuprun.lockfile
+
+# 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, exiting."
+ echo `date`": $LOCKFILE exists, exiting." >&2
+ exit 1
+fi;
+touch $LOCKFILE
+
#dirs
confdir=/home/backup/conf/ #configuration files
storagedir=/backup #mountpoint of huge disc
export PATH
#start the logfile
-date
-echo
-echo "Exclude pattern:"
-cat $exclude
-echo
-echo
+echo `date`": Backup run starting" >&2
umask 027
| grep -v "^#" \
| grep -v "^$" `
-
+# Backup only one computer, from command line?
if [ $1 ] ; then
unixcomputers=$1
fi;
-###
#disse bør kunne varieres fra fs til fs?
maksantallfulle=3 #hvor mange fulle vi tar vare på
backup()
{
- printf "%-${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
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
- echo "cmdline: $tar cfCl - $filesystem . $lastcmd --exclude-from=$exf | gzip"
- if `ssh -c blowfish $computer "$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
+ eval $TARCMD;
+ if [ -s $TARFILE ]; then
+ mv $TARFILE $DATE.tgz
#make a filelist.
#update the datefile if the filelist is ok.
tar tvfz $DATE.tgz > $DATE.idx 2>&1 &&
touch -t $DATE ../.date
#make a sortet filelist
- grep -v ^d $DATE.idx | sort -n -r +2 > $DATE.sdx
+ grep -v ^d $DATE.idx | sort -n -r -k 2 > $DATE.sdx
#fix perm
chmod 600 *tgz #only for us
else
#it did not work
- rm $DATE.tmp 2>/dev/null
- echo "$backuplevel backup of $computer:$filesystem failed and deleted"
- # TODO send mail!
+ 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 $unixcomputers
do
(
+ # Check that the target filesystem is mounted
+ # (actually check that it's not the root filesystem)
+ rootfilesystem=`df -P /`
+ targetfilesystem=`df -P "$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 $computer "cat /etc/fstab" \
+ filesystems=`ssh -n root@$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
- 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
#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
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, eller dø
# TODO bør sende mail om dette skjer!
- cd $storagedir/$computer/$sfilesystem/$backuplevel || exit 1
+ cd $storagedir/$computer/$sfilesystem/$backuplevel || die
#perform the actual backup
backup
fi
#delete complete backups
- for full in `ls -1t $storagedir/$computer/$sfilesystem/full/*tgz | tail +$mf`
+ for full in `ls -1t $storagedir/$computer/$sfilesystem/full/*tgz | tail -n +$mf`
do
prefix=`echo $full | sed "s/\.[^.]*$//"`
- echo "$computer:$filesystem sletter full $prefix (for mange)"
+ echo `date`": $computer:$filesystem sletter full $prefix (for mange)" >&2
rm $prefix*
done
-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
wait
-#create report
-/home/backup/report.sh $storagedir > /tmp/report.txt
-#[ ! -z $remotestat ] && scp /tmp/report.txt $remotestat
-###
-
-# print diskusage to logfile
-df -k
-
-date
+# 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
projects / backup.sh / blobdiff