X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=backup.sh;h=c4330309ea2d685208766d08a3aa3d99ad6adbe2;hb=22aae2730e071334116c61dc67d9f00ea960b364;hp=adeea18272ce1860731430b7fb3e4b6434c2497e;hpb=6d730569affcb76153e16c99c568663932c6dbd1;p=backup.sh diff --git a/backup.sh b/backup.sh index adeea18..c433030 100755 --- a/backup.sh +++ b/backup.sh @@ -1,5 +1,29 @@ #!/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 $LOCKFILE + #dirs confdir=/home/backup/conf/ #configuration files storagedir=/backup #mountpoint of huge disc @@ -20,12 +44,7 @@ PATH=/local/bin:$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 @@ -34,11 +53,10 @@ unixcomputers=`cat $confdir/computers.unix \ | grep -v "^#" \ | grep -v "^$" ` - +# Backup only one computer, from command line? if [ $1 ] ; then unixcomputers=$1 fi; -### #disse bør kunne varieres fra fs til fs? maksantallfulle=3 #hvor mange fulle vi tar vare på @@ -57,12 +75,13 @@ DATEs=`date "+%Y-%m-%d %H:%M"` #format: tar backup() { - printf "%-${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 @@ -79,17 +98,23 @@ backup() exf=`ssh $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 --one-file-system -cfC - $filesystem . $lastcmd --exclude-from=$exf | gzip" - if `ssh $computer "$tar --one-file-system -cfC - $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 $computer \"$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 && @@ -105,9 +130,11 @@ backup() 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 } @@ -116,13 +143,15 @@ backup() 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 $computer /bin/true`; then - echo "Passwordless SSH to $computer works." - else - echo "Could not use passwordless SSH to $computer. Exiting." - exit 1; + if ! `ssh -n -o NumberOfPasswordPrompts=0 $computer /bin/true`; then + echo `date`" $computer: Could not use passwordless SSH." + + # We don't want to return 0 + R=1 + break; fi #tømmer variabelen for sikkerhets skyld @@ -136,17 +165,37 @@ do | 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 $computer "rm -r ~/.backup ; mkdir -m 700 ~/.backup"; then + echo `date`" $computer: Could not create backup staging area at $computer:~/.backup" + # We don't want to return 0 + R=1 + break; + fi #try to copy $exclude to $computer - scp $exclude $computer:~/.backup/exclude || - echo "could not copy exclude.txt to $computer :(" + if ! scp $exclude $computer:~/.backup/exclude > /dev/null; then + echo `date`" $computer: Could not copy exclude.txt to $computer" + # We don't want to return 0 + R=1 + break; + 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 $computer:~/.backup/preexec || + ( echo `date`" $computer: Could not copy preexec.$computer to $computer:~/.backup/preexec"; + R=1 + break + ) + ) + [ -f $confdir/postexec.$computer ] && ( + scp $confdir/postexec.$computer $computer:~/.backup/postexec || + ( echo `date`" $computer: Could not copy postexec.$computer to $computer:~/.backup/postexec" + break + R=1 + ) + ) #try to run preexec if it exist ssh $computer "[ -f ~/.backup/preexec ] && /bin/bash -x ~/.backup/preexec" @@ -159,7 +208,6 @@ 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 @@ -174,15 +222,13 @@ 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, 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 @@ -198,7 +244,7 @@ do for full in `ls -1t $storagedir/$computer/$sfilesystem/full/*tgz | tail +$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 @@ -209,8 +255,6 @@ 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 @@ -221,13 +265,13 @@ done & wait -#create report -# (Hasn't been working for years -berge) -#/home/backup/report.sh $storagedir > /tmp/report.txt -#[ ! -z $remotestat ] && scp /tmp/report.txt $remotestat -### - -# print diskusage to logfile -df -k +# Remove lockfile +rm $LOCKFILE -date +# Did anything go wrong? +if [ $R != 0 ]; then + echo `date`": Backup run ended with errors, check logs." + exit 1 +else + echo `date`": Backup run ended" >&2 +fi