4 # Backup up Unix-like computers
8 # This script was first checked into RCS in 2000, and has since backed up a few
9 # servers almost every day, whenever not broken.
11 # It's a bit peculiar, but quite effective. The following is worth keeping in
14 # KISS. And by that I mean really simple. Only regular Unix commands, please
15 # (but GNU extensions are used).
17 # Because of the way we run this script from cron, informational logging goes
18 # to stderr and warnings and errors to stdout. This way, we get mail whenever
19 # stuff fails, and other output is redirected to the log file.
21 # The computers are backed up in parallell (but the filesystems on each
22 # computer in serial). This is implemented highly simplistic, with forking
23 # subshells. The script has no concept about threads, and stuff might go wrong,
24 # which usually requires manual cleanups.
26 # The script is halfway Norwegian and halfway English. Newer modifications are
27 # in English, we should probably stick to that.
31 # o $R has no proper meaning because of the forking subshell stuff. We need a
32 # more robust way of returning non-zero on errors, which we probably don't
34 # o The code around line 211 and downwards is messy.
38 LOCKFILE=/home/backup/backuprun.lockfile
39 confdir=/home/backup/conf/ # Configuration files
40 storagedir=/backup # Where we keep backups
41 maksantallfulle=3 # Number of full backups
42 dagermellomfulle=30 # Days between full backups
43 DATE=`date "+%Y%m%d%H%M"` #format: touch
44 DATEs=`date "+%Y-%m-%d %H:%M"` #format: tar
47 exclude=$confdir/exclude
48 [ ! -f $exclude ] && printf "tmp\ncore\n" > $exclude
50 # Initially, we consider everything to be just fine.
53 # Die more or less gracefully (at least notify user)
55 echo `date`": Something nasty happened, and since I fork a lot:"
56 echo `date`": I CANNOT CLEAN UP THE MESS MYSELF."
57 echo `date`": You need to get rid of lost process named stuff like $0, tar and ssh."
62 trap die SIGINT SIGTERM
64 # Don't start if we're already running
65 if [ -e $LOCKFILE ]; then
66 echo `date`": $LOCKFILE exists for pid `cat $LOCKFILE`, exiting."
67 echo `date`": $LOCKFILE exists for pid `cat $LOCKFILE`, exiting." >&2
73 #syntax of remotestat:
75 remotestatf=$confdir/remotestat
76 [ -f $remotestatf ] && remotestat=`cat $remotestatf`
79 PATH=/local/bin:$PATH:/store/bin
82 echo `date`": Backup run starting" >&2
86 # The computers we want to back up
87 unixcomputers=`cat $confdir/computers.unix \
91 # Backup only one computer, from command line?
98 # krever at noen variabler er satt
99 # krever at vi er i rett katalog
104 printf " $computer: $computer:$filesystem $backuplevel backup" >&2
106 if [ "$backuplevel" = "daglig" ] || [ "$backuplevel" = "incremental" ]
109 lastcmd="--newer='$lastd'"
114 #if this client has a special tar
115 #we need to find a better solution to this conf-issue
116 if [ -f $confdir/tar.$computer ] ; then
117 tar=`cat $confdir/tar.$computer`
122 # expand the exclude-path for use with tar
123 exf=`ssh root@$computer "ls ~/.backup/exclude"`
125 # We try to run tar on the remote computer
128 # C change to directory first
129 # . where to start taring (see C)
130 # $lastcmd only files newer than this
131 # --one-file-system don't traverse file systems
132 # --exclude-from file to get exclusion pattern from
133 # Pipe the stuff over ssh to ourselves, run pee to cat the contents to a
134 # file and build a file list, respectivly.
136 TARCMD="ssh root@$computer \"$tar --one-file-system -zcf - -C $filesystem . $lastcmd \
137 --exclude-from=$exf\" | pee \"cat > $TARFILE\" \"tar tzvf -\""
138 echo `date`" $computer: Running $TARCMD" >&2
139 eval $TARCMD > $DATE.idx
141 if [ -s $TARFILE ]; then
142 # File is >0 in size, we assume it worked.
144 # Move tar file in place
145 mv $TARFILE $DATE.tgz
148 echo $DATEs > ../.date && touch -t $DATE ../.date
150 # Make a sorted filelist as well
151 sort -k6 < $DATE.idx > $DATE.sdx
154 chmod 600 *tgz #only for us
155 chmod 644 *sdx *idx 2>/dev/null #everyone can read
158 # Something wrong happened.
160 echo `date`" $computer: $TARFILE empty. $backuplevel backup of $computer:$filesystem failed and deleted"
162 # We don't want to return 0
168 for computer in $unixcomputers
171 # Check that the target filesystem is mounted (actually check that it's not
172 # the root filesystem)
173 rootfilesystem=`df -P /`
174 targetfilesystem=`df -P "$storagedir/$computer"`
175 if [ "$rootfilesystem" == "$targetfilesystem" ]; then
176 echo `date`" $computer: Target filesystem ($storagedir/$computer) was mounted on /. Aborting"
180 echo `date`" $computer: Backing up $computer" >&2
182 # Try to SSH to the computer without entering a password.
183 if ! `ssh -n -o NumberOfPasswordPrompts=0 root@$computer /bin/true`; then
184 echo `date`" $computer: Could not use passwordless SSH."
186 # We don't want to return 0
194 # Check dump bit in fstab
195 filesystems=`ssh -n root@$computer "cat /etc/fstab" \
199 | awk '{ if ( $(NF-1) != "0" ) print $2}' `
201 # Clean up our dir at this client
202 if ! ssh root@$computer "rm -r ~/.backup ; mkdir -m 700 ~/.backup"; then
203 echo `date`" $computer: Could not create backup staging area at $computer:~/.backup - skipping backup of $computer"
208 # Try to copy $exclude to $computer
209 if ! scp $exclude root@$computer:~/.backup/exclude > /dev/null; then
210 echo `date`" $computer: Could not copy exclude.txt to $computer - skipping backup of $computer"
215 # Try to copy preeexec and postexec if they exist
216 # TODO: Gah, clean this mess!
217 [ -f $confdir/preexec.$computer ] && (
218 scp $confdir/preexec.$computer root@$computer:~/.backup/preexec >&2 ||
219 ( echo `date`" $computer: Could not copy preexec.$computer to $computer:~/.backup/preexec - skipping backup of $computer"
224 [ -f $confdir/postexec.$computer ] && (
225 scp $confdir/postexec.$computer root@$computer:~/.backup/postexec >&2 ||
226 ( echo `date`" $computer: Could not copy postexec.$computer to $computer:~/.backup/postexec - skipping backup of $computer"
232 # Try to run preexec if it exist
233 if ! ssh root@$computer "[ ! -f ~/.backup/preexec ] || /bin/bash -x ~/.backup/preexec" >&2; then
234 echo `date`" $computer: Could not run $computer:~/.backup/preexec - skipping backup of $computer"
240 for filesystem in $filesystems
242 # Remove / and $ (for Windows) for path names
243 sfilesystem=`echo $filesystem | tr '\/\$' '__'`
245 # Prepare storage area
246 mkdir -m 755 -p $storagedir/$computer/$sfilesystem/{full,daglig} 2>/dev/null
247 echo $filesystem > ${storagedir}/${computer}/.${sfilesystem}.name
248 chmod 644 ${storagedir}/${computer}/.${sfilesystem}.name
250 # Default backuplevel
253 if [ ! -f $storagedir/$computer/$sfilesystem/.date ]
255 # Take the first full backup of this filesystem on this computer
257 echo $DATEs > $storagedir/$computer/$sfilesystem/.date
260 # Check if we want a full backup
261 if [ -z "`find $storagedir/$computer/$sfilesystem/full/ -name \*tgz -mtime -$dagermellomfulle`" ]; then
266 cd $storagedir/$computer/$sfilesystem/$backuplevel || die
268 # Perform the actual backup
271 # Check if this box has a custom number of full backups
272 if [ -f $confdir/maksfulle.$computer ] ; then
273 mf=$((`cat $confdir/maksfulle.$computer`+1))
275 mf=$(($maksantallfulle+1))
278 # Delete old full backups
279 for full in `ls -1t $storagedir/$computer/$sfilesystem/full/*tgz | tail -n +$mf`
281 prefix=`echo $full | sed "s/\.[^.]*$//"`
282 echo `date`": $computer:$filesystem Deleting full backup $prefix" >&2
286 # Delete incremental backups older than the oldest full backup
287 oldf=`ls -t1 $storagedir/$computer/$sfilesystem/full/*tgz | tail -1`
289 $storagedir/$computer/$sfilesystem/daglig \
292 -printf "`date`: $computer: Deleting old incremental backup: %p\n" \
296 # Try to run postexec if it exist
297 if ! ssh root@$computer "[ ! -f ~/.backup/postexec ] || /bin/bash -x ~/.backup/postexec" >&2; then
298 echo `date`" $computer: Could not run $computer:~/.backup/postexec"
310 # Did anything go wrong?
311 # (IMPORTANT NOTE: The R-business does not work as expected, since this script
314 echo `date`": Backup run ended with errors, check logs."
317 echo `date`": Backup run ended" >&2