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 $R has no proper meaning because of the forking subshell stuff. We need a
32 # more robust way of returning non-zero on errors, which we probably don't
37 LOCKFILE=/home/backup/backuprun.lockfile
38 confdir=/home/backup/conf/ # Configuration files
39 storagedir=/backup # Where we keep backups
40 maksantallfulle=3 # Number of full backups
41 dagermellomfulle=30 # Days between full backups
42 DATE=`date "+%Y%m%d%H%M"` #format: touch
43 DATEs=`date "+%Y-%m-%d %H:%M"` #format: tar
46 exclude=$confdir/exclude
47 [ ! -f $exclude ] && printf "tmp\ncore\n" > $exclude
49 # Initially, we consider everything to be just fine.
52 # Die more or less gracefully (at least notify user)
54 echo `date`": Something nasty happened, and since I fork a lot:"
55 echo `date`": I CANNOT CLEAN UP THE MESS MYSELF."
56 echo `date`": You need to get rid of lost process named stuff like $0, tar and ssh."
61 trap die SIGINT SIGTERM
63 # Don't start if we're already running
64 if [ -e $LOCKFILE ]; then
65 echo `date`": $LOCKFILE exists for pid `cat $LOCKFILE`, exiting."
66 echo `date`": $LOCKFILE exists for pid `cat $LOCKFILE`, exiting." >&2
72 #syntax of remotestat:
74 remotestatf=$confdir/remotestat
75 [ -f $remotestatf ] && remotestat=`cat $remotestatf`
78 PATH=/local/bin:$PATH:/store/bin
81 echo `date`": Backup run starting" >&2
85 # The computers we want to back up
86 unixcomputers=`cat $confdir/computers.unix \
90 # Backup only one computer, from command line?
97 # krever at noen variabler er satt
98 # krever at vi er i rett katalog
103 printf " $computer: $computer:$filesystem $backuplevel backup" >&2
105 if [ "$backuplevel" = "daglig" ] || [ "$backuplevel" = "incremental" ]
108 lastcmd="--newer='$lastd'"
113 #if this client has a special tar
114 #we need to find a better solution to this conf-issue
115 if [ -f $confdir/tar.$computer ] ; then
116 tar=`cat $confdir/tar.$computer`
121 # expand the exclude-path for use with tar
122 exf=`ssh root@$computer "ls ~/.backup/exclude"`
124 # We try to run tar on the remote computer
127 # C change to directory first
128 # . where to start taring (see C)
129 # $lastcmd only files newer than this
130 # --one-file-system don't traverse file systems
131 # --exclude-from file to get exclusion pattern from
132 # Pipe the stuff over ssh to ourselves, run pee to cat the contents to a
133 # file and build a file list, respectivly.
135 TARCMD="ssh root@$computer \"$tar --one-file-system -zcf - -C $filesystem . $lastcmd \
136 --exclude-from=$exf\" | pee \"cat > $TARFILE\" \"tar tzvf -\""
137 echo `date`" $computer: Running $TARCMD" >&2
138 eval $TARCMD > $DATE.idx
140 if [ -s $TARFILE ]; then
141 # File is >0 in size, we assume it worked.
143 # Move tar file in place
144 mv $TARFILE $DATE.tgz
147 echo $DATEs > ../.date && touch -t $DATE ../.date
149 # Make a sorted filelist as well
150 sort -k6 < $DATE.idx > $DATE.sdx
153 chmod 600 *tgz #only for us
154 chmod 644 *sdx *idx 2>/dev/null #everyone can read
157 # Something wrong happened.
159 echo `date`" $computer: $TARFILE empty. $backuplevel backup of $computer:$filesystem failed and deleted"
161 # We don't want to return 0
167 for computer in $unixcomputers
170 # Check that the target filesystem is mounted (actually check that it's not
171 # the root filesystem)
172 rootfilesystem=`df -P /`
173 targetfilesystem=`df -P "$storagedir/$computer"`
174 if [ "$rootfilesystem" == "$targetfilesystem" ]; then
175 echo `date`" $computer: Target filesystem ($storagedir/$computer) was mounted on /. Aborting"
179 echo `date`" $computer: Backing up $computer" >&2
181 # Try to SSH to the computer without entering a password.
182 if ! `ssh -n -o NumberOfPasswordPrompts=0 root@$computer /bin/true`; then
183 echo `date`" $computer: Could not use passwordless SSH."
185 # We don't want to return 0
193 # Check dump bit in fstab
194 filesystems=`ssh -n root@$computer "cat /etc/fstab" \
198 | awk '{ if ( $(NF-1) != "0" ) print $2}' `
200 # Clean up our dir at this client
201 if ! ssh root@$computer "rm -r ~/.backup ; mkdir -m 700 ~/.backup"; then
202 echo `date`" $computer: Could not create backup staging area at $computer:~/.backup - skipping backup of $computer"
207 # Try to copy $exclude to $computer
208 if ! scp $exclude root@$computer:~/.backup/exclude > /dev/null; then
209 echo `date`" $computer: Could not copy exclude.txt to $computer - skipping backup of $computer"
214 # Try to copy preeexec and postexec if they exist
215 if [ -f $confdir/preexec.$computer ]; then
216 if ! scp $confdir/preexec.$computer root@$computer:~/.backup/preexec >&2; then
217 echo `date`" $computer: Could not copy preexec.$computer to $computer:~/.backup/preexec - skipping backup of $computer"
222 if [ -f $confdir/postexec.$computer ]; then
223 if ! scp $confdir/postexec.$computer root@$computer:~/.backup/postexec >&2; then
224 echo `date`" $computer: Could not copy postexec.$computer to $computer:~/.backup/postexec - skipping backup of $computer"
230 # Try to run preexec if it exist
231 if ! ssh root@$computer "[ ! -f ~/.backup/preexec ] || /bin/bash -x ~/.backup/preexec" >&2; then
232 echo `date`" $computer: Could not run $computer:~/.backup/preexec - skipping backup of $computer"
238 for filesystem in $filesystems
240 # Remove / and $ (for Windows) for path names
241 sfilesystem=`echo $filesystem | tr '\/\$' '__'`
243 # Prepare storage area
244 mkdir -m 755 -p $storagedir/$computer/$sfilesystem/{full,daglig} 2>/dev/null
245 echo $filesystem > ${storagedir}/${computer}/.${sfilesystem}.name
246 chmod 644 ${storagedir}/${computer}/.${sfilesystem}.name
248 # Default backuplevel
251 if [ ! -f $storagedir/$computer/$sfilesystem/.date ]
253 # Take the first full backup of this filesystem on this computer
255 echo $DATEs > $storagedir/$computer/$sfilesystem/.date
258 # Check if we want a full backup
259 if [ -z "`find $storagedir/$computer/$sfilesystem/full/ -name \*tgz -mtime -$dagermellomfulle`" ]; then
264 cd $storagedir/$computer/$sfilesystem/$backuplevel || die
266 # Perform the actual backup
269 # Check if this box has a custom number of full backups
270 if [ -f $confdir/maksfulle.$computer ] ; then
271 mf=$((`cat $confdir/maksfulle.$computer`+1))
273 mf=$(($maksantallfulle+1))
276 # Delete old full backups
277 for full in `ls -1t $storagedir/$computer/$sfilesystem/full/*tgz | tail -n +$mf`
279 prefix=`echo $full | sed "s/\.[^.]*$//"`
280 echo `date`": $computer:$filesystem Deleting full backup $prefix" >&2
284 # Delete incremental backups older than the oldest full backup
285 oldf=`ls -t1 $storagedir/$computer/$sfilesystem/full/*tgz | tail -1`
287 $storagedir/$computer/$sfilesystem/daglig \
290 -printf "`date`: $computer: Deleting old incremental backup: %p\n" \
294 # Try to run postexec if it exist
295 if ! ssh root@$computer "[ ! -f ~/.backup/postexec ] || /bin/bash -x ~/.backup/postexec" >&2; then
296 echo `date`" $computer: Could not run $computer:~/.backup/postexec"
308 # Did anything go wrong?
309 # (IMPORTANT NOTE: The R-business does not work as expected, since this script
312 echo `date`": Backup run ended with errors, check logs."
315 echo `date`": Backup run ended" >&2