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.
29 # TODO(sesse, 2007-12-16):
30 # * Sjekk hvorfor .idx-filene til UD er blitt 0 for siste fullbackuper
31 # * Rydde uleselig kode fra linje 211 og litt nedover
34 LOCKFILE=/home/backup/backuprun.lockfile
35 confdir=/home/backup/conf/ # Configuration files
36 storagedir=/backup # Where we keep backups
37 maksantallfulle=3 # Number of full backups
38 dagermellomfulle=30 # Days between full backups
39 DATE=`date "+%Y%m%d%H%M"` #format: touch
40 DATEs=`date "+%Y-%m-%d %H:%M"` #format: tar
43 exclude=$confdir/exclude
44 [ ! -f $exclude ] && printf "tmp\ncore\n" > $exclude
46 # Initially, we consider everything to be just fine.
49 # Die more or less gracefully (at least notify user)
51 echo `date`": Something nasty happened, and since I fork a lot:"
52 echo `date`": I CANNOT CLEAN UP THE MESS MYSELF."
53 echo `date`": You need to get rid of lost process named stuff like $0, tar and ssh."
58 trap die SIGINT SIGTERM
60 # Don't start if we're already running
61 if [ -e $LOCKFILE ]; then
62 echo `date`": $LOCKFILE exists, exiting."
63 echo `date`": $LOCKFILE exists, exiting." >&2
68 #syntax of remotestat:
70 remotestatf=$confdir/remotestat
71 [ -f $remotestatf ] && remotestat=`cat $remotestatf`
74 PATH=/local/bin:$PATH:/store/bin
77 echo `date`": Backup run starting" >&2
81 # The computers we want to back up
82 unixcomputers=`cat $confdir/computers.unix \
86 # Backup only one computer, from command line?
93 # krever at noen variabler er satt
94 # krever at vi er i rett katalog
99 printf " $computer: $computer:$filesystem $backuplevel backup" >&2
101 if [ "$backuplevel" = "daglig" ] || [ "$backuplevel" = "incremental" ]
104 lastcmd="--newer='$lastd'"
109 #if this client has a special tar
110 #we need to find a better solution to this conf-issue
111 if [ -f $confdir/tar.$computer ] ; then
112 tar=`cat $confdir/tar.$computer`
117 # expand the exclude-path for use with tar
118 exf=`ssh root@$computer "ls ~/.backup/exclude"`
120 # We try to run tar on the remote computer
123 # C change to directory first
124 # . where to start taring (see C)
125 # $lastcmd only files newer than this
126 # --one-file-system don't traverse file systems
127 # --exclude-from file to get exclusion pattern from
128 # Pipe the stuff over ssh to ourselves, run pee to cat the contents to a
129 # file and build a file list, respectivly.
131 TARCMD="ssh root@$computer \"$tar --one-file-system -zcf - -C $filesystem . $lastcmd \
132 --exclude-from=$exf\" | pee \"cat > $TARFILE\" \"tar tzvf -\""
133 echo `date`" $computer: Running $TARCMD" >&2
134 eval $TARCMD > $DATE.idx
136 if [ -s $TARFILE ]; then
137 # File is >0 in size, we assume it worked.
139 # Move tar file in place
140 mv $TARFILE $DATE.tgz
143 echo $DATEs > ../.date && touch -t $DATE ../.date
145 # Make a sorted filelist
146 grep -v ^d $DATE.idx | sort -n -r -k 2 > $DATE.sdx
149 chmod 600 *tgz #only for us
150 chmod 644 *sdx *idx 2>/dev/null #everyone can read
153 # Something wrong happened.
155 echo `date`" $computer: $TARFILE empty. $backuplevel backup of $computer:$filesystem failed and deleted"
157 # We don't want to return 0
163 for computer in $unixcomputers
166 # Check that the target filesystem is mounted (actually check that it's not
167 # the root filesystem)
168 rootfilesystem=`df -P /`
169 targetfilesystem=`df -P "$storagedir/$computer"`
170 if [ "$rootfilesystem" == "$targetfilesystem" ]; then
171 echo `date`" $computer: Target filesystem ($storagedir/$computer) was mounted on /. Aborting"
175 echo `date`" $computer: Backing up $computer" >&2
177 # Try to SSH to the computer without entering a password.
178 if ! `ssh -n -o NumberOfPasswordPrompts=0 root@$computer /bin/true`; then
179 echo `date`" $computer: Could not use passwordless SSH."
181 # We don't want to return 0
189 # Check dump bit in fstab
190 filesystems=`ssh -n root@$computer "cat /etc/fstab" \
194 | awk '{ if ( $(NF-1) != "0" ) print $2}' `
196 # Clean up our dir at this client
197 if ! ssh root@$computer "rm -r ~/.backup ; mkdir -m 700 ~/.backup"; then
198 echo `date`" $computer: Could not create backup staging area at $computer:~/.backup - skipping backup of $computer"
203 # Try to copy $exclude to $computer
204 if ! scp $exclude root@$computer:~/.backup/exclude > /dev/null; then
205 echo `date`" $computer: Could not copy exclude.txt to $computer - skipping backup of $computer"
210 # Try to copy preeexec and postexec if they exist
211 # TODO: Gah, clean this mess!
212 [ -f $confdir/preexec.$computer ] && (
213 scp $confdir/preexec.$computer root@$computer:~/.backup/preexec >&2 ||
214 ( echo `date`" $computer: Could not copy preexec.$computer to $computer:~/.backup/preexec - skipping backup of $computer"
219 [ -f $confdir/postexec.$computer ] && (
220 scp $confdir/postexec.$computer root@$computer:~/.backup/postexec >&2 ||
221 ( echo `date`" $computer: Could not copy postexec.$computer to $computer:~/.backup/postexec - skipping backup of $computer"
227 # Try to run preexec if it exist
228 if ! ssh root@$computer "[ ! -f ~/.backup/preexec ] || /bin/bash -x ~/.backup/preexec" >&2; then
229 echo `date`" $computer: Could not run $computer:~/.backup/preexec - skipping backup of $computer"
235 for filesystem in $filesystems
237 # Remove / and $ (for Windows) for path names
238 sfilesystem=`echo $filesystem | tr '\/\$' '__'`
240 # Prepare storage area
241 mkdir -m 755 -p $storagedir/$computer/$sfilesystem/{full,daglig} 2>/dev/null
242 echo $filesystem > ${storagedir}/${computer}/.${sfilesystem}.name
243 chmod 644 ${storagedir}/${computer}/.${sfilesystem}.name
245 # Default backuplevel
248 if [ ! -f $storagedir/$computer/$sfilesystem/.date ]
250 # Take the first full backup of this filesystem on this computer
252 echo $DATEs > $storagedir/$computer/$sfilesystem/.date
255 # Check if we want a full backup
256 if [ -z "`find $storagedir/$computer/$sfilesystem/full/ -name \*tgz -mtime -$dagermellomfulle`" ]; then
261 cd $storagedir/$computer/$sfilesystem/$backuplevel || die
263 # Perform the actual backup
266 # Check if this box has a custom number of full backups
267 if [ -f $confdir/maksfulle.$computer ] ; then
268 mf=$((`cat $confdir/maksfulle.$computer`+1))
270 mf=$(($maksantallfulle+1))
273 # Delete old full backups
274 for full in `ls -1t $storagedir/$computer/$sfilesystem/full/*tgz | tail -n +$mf`
276 prefix=`echo $full | sed "s/\.[^.]*$//"`
277 echo `date`": $computer:$filesystem Deleting full backup $prefix" >&2
281 # Delete incremental backups older than the oldest full backup
282 oldf=`ls -t1 $storagedir/$computer/$sfilesystem/full/*tgz | tail -1`
284 $storagedir/$computer/$sfilesystem/daglig \
287 -printf "`date`: $computer: Deleting old incremental backup: %p\n" \
291 # Try to run postexec if it exist
292 if ! ssh root@$computer "[ ! -f ~/.backup/postexec ] || /bin/bash -x ~/.backup/postexec" >&2; then
293 echo `date`" $computer: Could not run $computer:~/.backup/postexec"
305 # Did anything go wrong?
306 # (IMPORTANT NOTE: The R-business does not work as expected, since this script
309 echo `date`": Backup run ended with errors, check logs."
312 echo `date`": Backup run ended" >&2