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 $R has no proper meaning because of the forking subshell stuff. We need a
32 # more robust way of returning non-zero on errors, which we probably don't
37 LOCKFILE=/home/backup/backuprun.lockfile
38 confdir=/home/backup/conf/ # Configuration files
39 storagedir=/backup # Where we keep backups
40 maksantallfulle=3 # Number of full backups
41 dagermellomfulle=30 # Days between full backups
42 dagermellomfulleforfast=32 # Days between full backups for machines with fixed
43 # full-backup date (in case the fixed day gets missed)
44 DAY_OF_MONTH=`date "+%d" | sed s/^0//` # from 1 to 31
45 DATE=`date "+%Y%m%d%H%M"` #format: touch
46 DATEs=`date "+%Y-%m-%d %H:%M"` #format: tar
49 exclude=$confdir/exclude
50 [ ! -f $exclude ] && printf "tmp\ncore\n" > $exclude
52 # Initially, we consider everything to be just fine.
55 # Die more or less gracefully (at least notify user)
57 echo `date`": Something nasty happened, and since I fork a lot:"
58 echo `date`": I CANNOT CLEAN UP THE MESS MYSELF."
59 echo `date`": You need to get rid of lost process named stuff like $0, tar and ssh."
64 trap die SIGINT SIGTERM
66 # Don't start if we're already running
67 if [ -e $LOCKFILE ]; then
68 echo `date`": $LOCKFILE exists for pid `cat $LOCKFILE`, exiting."
69 echo `date`": $LOCKFILE exists for pid `cat $LOCKFILE`, exiting." >&2
75 #syntax of remotestat:
77 remotestatf=$confdir/remotestat
78 [ -f $remotestatf ] && remotestat=`cat $remotestatf`
81 PATH=/local/bin:$PATH:/store/bin
84 echo `date`": Backup run starting" >&2
88 # The computers we want to back up
89 unixcomputers=`cat $confdir/computers.unix \
93 # Backup only one computer, from command line?
100 # krever at noen variabler er satt
101 # krever at vi er i rett katalog
106 printf " $computer: $computer:$filesystem $backuplevel backup\n" >&2
108 if [ "$backuplevel" = "daglig" ] || [ "$backuplevel" = "incremental" ]
111 lastcmd="--newer='$lastd'"
116 #if this client has a special tar
117 #we need to find a better solution to this conf-issue
118 if [ -f $confdir/tar.$computer ] ; then
119 tar=`cat $confdir/tar.$computer`
124 # expand the exclude-path for use with tar
125 exf=`ssh root@$computer "ls ~/.backup/exclude"`
127 # We try to run tar on the remote computer
130 # C change to directory first
131 # . where to start taring (see C)
132 # $lastcmd only files newer than this
133 # --one-file-system don't traverse file systems
134 # --exclude-from file to get exclusion pattern from
135 # Pipe the stuff over ssh to ourselves, run pee to cat the contents to a
136 # file and build a file list, respectivly.
138 TARCMD="ssh root@$computer \"$tar --one-file-system -zcf - -C $filesystem . $lastcmd \
139 --exclude-from=$exf\" | pee \"cat > $TARFILE\" \"tar tzvf -\""
140 echo `date`" $computer: Running $TARCMD" >&2
141 eval $TARCMD > $DATE.idx
143 if [ -s $TARFILE ]; then
144 # File is >0 in size, we assume it worked.
146 # Move tar file in place
147 mv $TARFILE $DATE.tgz
150 echo $DATEs > ../.date && touch -t $DATE ../.date
152 # Make a sorted filelist as well
153 sort -k6 < $DATE.idx > $DATE.sdx
156 chmod 600 *tgz #only for us
157 chmod 644 *sdx *idx 2>/dev/null #everyone can read
160 # Something wrong happened.
162 echo `date`" $computer: $TARFILE empty. $backuplevel backup of $computer:$filesystem failed and deleted"
164 # We don't want to return 0
170 for computer in $unixcomputers
173 # Check that the target filesystem is mounted (actually check that it's not
174 # the root filesystem)
175 rootfilesystem=`df -P /`
176 targetfilesystem=`df -P "$storagedir/$computer"`
177 if [ "$rootfilesystem" == "$targetfilesystem" ]; then
178 echo `date`" $computer: Target filesystem ($storagedir/$computer) was mounted on /. Aborting"
182 echo `date`" $computer: Backing up $computer" >&2
184 # Try to SSH to the computer without entering a password.
185 if ! `ssh -n -o NumberOfPasswordPrompts=0 root@$computer /bin/true`; then
186 echo `date`" $computer: Could not use passwordless SSH."
188 # We don't want to return 0
196 # Check dump bit in fstab
197 filesystems=`ssh -n root@$computer "cat /etc/fstab" \
201 | awk '{ if ( $(NF-1) != "0" ) print $2}' `
203 # Clean up our dir at this client
204 if ! ssh root@$computer "rm -r ~/.backup ; mkdir -m 700 ~/.backup"; then
205 echo `date`" $computer: Could not create backup staging area at $computer:~/.backup - skipping backup of $computer"
210 # Try to copy $exclude to $computer
211 if ! scp $exclude root@$computer:~/.backup/exclude > /dev/null; then
212 echo `date`" $computer: Could not copy exclude.txt to $computer - skipping backup of $computer"
217 # Try to copy preeexec and postexec if they exist
218 if [ -f $confdir/preexec.$computer ]; then
219 if ! scp $confdir/preexec.$computer root@$computer:~/.backup/preexec >&2; then
220 echo `date`" $computer: Could not copy preexec.$computer to $computer:~/.backup/preexec - skipping backup of $computer"
225 if [ -f $confdir/postexec.$computer ]; then
226 if ! scp $confdir/postexec.$computer root@$computer:~/.backup/postexec >&2; then
227 echo `date`" $computer: Could not copy postexec.$computer to $computer:~/.backup/postexec - skipping backup of $computer"
233 # Try to run preexec if it exist
234 if ! ssh root@$computer "[ ! -f ~/.backup/preexec ] || /bin/bash -x ~/.backup/preexec" >&2; then
235 echo `date`" $computer: Could not run $computer:~/.backup/preexec - skipping backup of $computer"
241 for filesystem in $filesystems
243 # Remove / and $ (for Windows) for path names
244 sfilesystem=`echo $filesystem | tr '\/\$' '__'`
246 # Prepare storage area
247 mkdir -m 755 -p $storagedir/$computer/$sfilesystem/{full,daglig} 2>/dev/null
248 echo $filesystem > ${storagedir}/${computer}/.${sfilesystem}.name
249 chmod 644 ${storagedir}/${computer}/.${sfilesystem}.name
251 # Default backuplevel
254 if [ ! -f $storagedir/$computer/$sfilesystem/.date ]
256 # Take the first full backup of this filesystem on this computer
258 echo $DATEs > $storagedir/$computer/$sfilesystem/.date
261 # Check if we want a full backup
262 printf " $computer $filesystem: $confdir/fastfullbackupdag.$computer.$sfilesystem\n"
263 if [ -f $confdir/fastfullbackupdag.$computer.$sfilesystem ]; then
264 fullbackup_min_for_this_machine=$dagermellomfulleforfast
265 if [ "$DAY_OF_MONTH" = "`cat $confdir/fastfullbackupdag.$computer.$sfilesystem`" ]; then
268 elif [ -f $confdir/fastfullbackupdag.$computer ]; then
269 fullbackup_min_for_this_machine=$dagermellomfulleforfast
270 if [ "$DAY_OF_MONTH" = "`cat $confdir/fastfullbackupdag.$computer`" ]; then
274 fullbackup_min_for_this_machine=$dagermellomfulle
277 if [ -z "`find $storagedir/$computer/$sfilesystem/full/ -name \*tgz -mtime -$fullbackup_min_for_this_machine`" ]; then
282 cd $storagedir/$computer/$sfilesystem/$backuplevel || die
284 # Perform the actual backup
287 # Check if this box has a custom number of full backups
288 if [ -f $confdir/maksfulle.$computer ] ; then
289 mf=$((`cat $confdir/maksfulle.$computer`+1))
291 mf=$(($maksantallfulle+1))
294 # Delete old full backups
295 for full in `ls -1t $storagedir/$computer/$sfilesystem/full/*tgz | tail -n +$mf`
297 prefix=`echo $full | sed "s/\.[^.]*$//"`
298 echo `date`": $computer:$filesystem Deleting full backup $prefix" >&2
302 # Delete incremental backups older than the oldest full backup
303 oldf=`ls -t1 $storagedir/$computer/$sfilesystem/full/*tgz | tail -1`
305 $storagedir/$computer/$sfilesystem/daglig \
308 -printf "`date`: $computer: Deleting old incremental backup: %p\n" \
312 # Try to run postexec if it exist
313 if ! ssh root@$computer "[ ! -f ~/.backup/postexec ] || /bin/bash -x ~/.backup/postexec" >&2; then
314 echo `date`" $computer: Could not run $computer:~/.backup/postexec"
326 # Did anything go wrong?
327 # (IMPORTANT NOTE: The R-business does not work as expected, since this script
330 echo `date`": Backup run ended with errors, check logs."
333 echo `date`": Backup run ended" >&2