X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=backup.sh;h=75c030841688ebc6792e64875142a11f81aabdcf;hb=035fbd1eb0d7eedee962e733517e20c9fcbb0273;hp=10aa460d28641d62f9c2a72137e96711bff5e351;hpb=3a2bc78ce197ae29491aa24ead7e82f891297270;p=backup.sh diff --git a/backup.sh b/backup.sh index 10aa460..75c0308 100755 --- a/backup.sh +++ b/backup.sh @@ -25,19 +25,25 @@ # # The script is halfway Norwegian and halfway English. Newer modifications are # in English, we should probably stick to that. - -# TODO(sesse, 2007-12-16): -# * Bruk tee i stedet for en separat tar ztvf-prosess -# * Sjekk hvorfor .idx-filene til UD er blitt 0 for siste fullbackuper +# +# Known issues: +# +# o $R has no proper meaning because of the forking subshell stuff. We need a +# more robust way of returning non-zero on errors, which we probably don't +# right now. +# # Configuration LOCKFILE=/home/backup/backuprun.lockfile -confdir=/home/backup/conf/ # Configuration files -storagedir=/backup # Where we keep backups -maksantallfulle=3 # Number of full backups -dagermellomfulle=30 # Days between full backups -DATE=`date "+%Y%m%d%H%M"` #format: touch -DATEs=`date "+%Y-%m-%d %H:%M"` #format: tar +confdir=/home/backup/conf/ # Configuration files +storagedir=/backup # Where we keep backups +maksantallfulle=3 # Number of full backups +dagermellomfulle=30 # Days between full backups +dagermellomfulleforfast=32 # Days between full backups for machines with fixed + # full-backup date (in case the fixed day gets missed) +DAY_OF_MONTH=`date "+%d" | sed s/^0//` # from 1 to 31 +DATE=`date "+%Y%m%d%H%M"` #format: touch +DATEs=`date "+%Y-%m-%d %H:%M"` #format: tar # Exclude-pattern exclude=$confdir/exclude @@ -59,11 +65,12 @@ 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 + 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/ @@ -96,7 +103,7 @@ backup() { echo -n `date` >&2 - printf " $computer: $computer:$filesystem $backuplevel backup" >&2 + printf " $computer: $computer:$filesystem $backuplevel backup\n" >&2 if [ "$backuplevel" = "daglig" ] || [ "$backuplevel" = "incremental" ] then @@ -106,18 +113,10 @@ backup() lastcmd="" fi - #if this client has a special tar - #we need to find a better solution to this conf-issue - if [ -f $confdir/tar.$computer ] ; then - tar=`cat $confdir/tar.$computer` - else - tar=tar - fi - - #expand the exclude-path for use with tar + # expand the exclude-path for use with tar exf=`ssh root@$computer "ls ~/.backup/exclude"` - #We try to run tar on the remote computer + # We try to run tar on the remote computer # z gzip it # c create archive # C change to directory first @@ -125,29 +124,32 @@ backup() # $lastcmd only files newer than this # --one-file-system don't traverse file systems # --exclude-from file to get exclusion pattern from - # And at last, redirect stdout to stderr, to get output logged. + # Pipe the stuff over ssh to ourselves, run pee to cat the contents to a + # file and build a file list, respectivly. TARFILE=$DATE.tmp - TARCMD="ssh root@$computer \"$tar --one-file-system -zcf - -C $filesystem . $lastcmd \ - --exclude-from=$exf\" > $TARFILE" + TARCMD="ssh root@$computer \"tar --one-file-system -zcf - -C $filesystem . $lastcmd \ + --exclude-from=$exf\" | pee \"cat > $TARFILE\" \"tar tzvf -\"" echo `date`" $computer: Running $TARCMD" >&2 - eval $TARCMD; + 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 - #make a filelist. - #update the datefile if the filelist is ok. - tar tvfz $DATE.tgz > $DATE.idx 2>&1 && - echo $DATEs > ../.date && - touch -t $DATE ../.date - #make a sortet filelist - grep -v ^d $DATE.idx | sort -n -r -k 2 > $DATE.sdx + # Update timestamp + echo $DATEs > ../.date && touch -t $DATE ../.date - #fix perm + # Make a sorted filelist as well + sort -k6 < $DATE.idx > $DATE.sdx + + # Fix permissions chmod 600 *tgz #only for us chmod 644 *sdx *idx 2>/dev/null #everyone can read else - #it did not work + # Something wrong happened. rm $TARFILE echo `date`" $computer: $TARFILE empty. $backuplevel backup of $computer:$filesystem failed and deleted" @@ -157,12 +159,11 @@ backup() } -#Løper gjennom listen av unixmaskiner som vi skal ta backup av for computer in $unixcomputers do ( - # Check that the target filesystem is mounted - # (actually check that it's not the root filesystem) + # Check that the target filesystem is mounted (actually check that it's not + # the root filesystem) rootfilesystem=`df -P /` targetfilesystem=`df -P "$storagedir/$computer"` if [ "$rootfilesystem" == "$targetfilesystem" ]; then @@ -182,50 +183,46 @@ do continue; fi - #tømmer variabelen for sikkerhets skyld filesystems="" - #Sjekker nest siste felt i fstab. Om det er 0 tar vi ikke backup + # Check dump bit in fstab filesystems=`ssh -n root@$computer "cat /etc/fstab" \ | grep -v nfs \ | grep -v "^#" \ | grep -v "^$" \ | awk '{ if ( $(NF-1) != "0" ) print $2}' ` - #clean up our dir at this client + # Clean up our dir at this client if ! ssh root@$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" - # We don't want to return 0 R=1 continue; fi - #try to copy $exclude to $computer + # 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" - # We don't want to return 0 R=1 continue; fi - #try to copy preeexec and postexec if they exist -# 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 + # 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 @@ -235,45 +232,58 @@ do for filesystem in $filesystems do - #lager en variant uten tegnet "/" eller $ (gjelder NT) + # Remove / and $ (for Windows) for path names sfilesystem=`echo $filesystem | tr '\/\$' '__'` - #lager det som trengs av kataloger + # 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 - #set default backuplevel + # Default backuplevel backuplevel=daglig if [ ! -f $storagedir/$computer/$sfilesystem/.date ] then - #take the first full backup of this filesystem on this computer + # Take the first full backup of this filesystem on this computer backuplevel=full echo $DATEs > $storagedir/$computer/$sfilesystem/.date fi - #sjekker om det er på tide med en full - if [ -z "`find $storagedir/$computer/$sfilesystem/full/ -name \*tgz -mtime -$dagermellomfulle`" ]; then + # 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, eller dø - # TODO bør sende mail om dette skjer! + + # We want to be here cd $storagedir/$computer/$sfilesystem/$backuplevel || die - #perform the actual backup + # Perform the actual backup backup - # Sjekk om det skal være et annet antall fulle backuper av en boks + # 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 complete backups + # Delete old full backups for full in `ls -1t $storagedir/$computer/$sfilesystem/full/*tgz | tail -n +$mf` do prefix=`echo $full | sed "s/\.[^.]*$//"` @@ -281,7 +291,7 @@ do rm $prefix* done - #delete incremental backups older than the oldest complete backup + # Delete incremental backups older than the oldest full backup oldf=`ls -t1 $storagedir/$computer/$sfilesystem/full/*tgz | tail -1` find \ $storagedir/$computer/$sfilesystem/daglig \ @@ -291,7 +301,7 @@ do -exec rm {} \; >&2 done - #try to run postexec if it exist + # Try to run postexec if it exist if ! ssh root@$computer "[ ! -f ~/.backup/postexec ] || /bin/bash -x ~/.backup/postexec" >&2; then echo `date`" $computer: Could not run $computer:~/.backup/postexec" R=1 @@ -307,7 +317,7 @@ rm $LOCKFILE # Did anything go wrong? # (IMPORTANT NOTE: The R-business does not work as expected, since this script -# forks itself. +# forks itself). if [ $R != 0 ]; then echo `date`": Backup run ended with errors, check logs." exit 1