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 and moreutils has to be installed).
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 parallel (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?
99 # krever at noen variabler er satt
100 # krever at vi er i rett katalog
105 printf " $computer: $computer:$filesystem $backuplevel backup\n" >&2
107 if [ "$backuplevel" = "daglig" ] || [ "$backuplevel" = "incremental" ]
110 lastcmd="--newer='$lastd'"
115 # expand the exclude-path for use with tar
116 exf=`ssh root@$computer "ls ~/.backup/exclude"`
118 # We try to run tar on the remote computer
121 # C change to directory first
122 # . where to start taring (see C)
123 # $lastcmd only files newer than this
124 # --one-file-system don't traverse file systems
125 # --exclude-from file to get exclusion pattern from
126 # Pipe the stuff over ssh to ourselves, run pee to cat the contents to a
127 # file and build a file list, respectivly.
129 TARCMD="ssh root@$computer \"nice -n 19 ionice -c3 tar --one-file-system -zcf - -C $filesystem . $lastcmd \
130 --exclude-from=$exf\" | pee \"cat > $TARFILE\" \"tar tzvf -\""
131 echo `date`" $computer: Running $TARCMD" >&2
132 eval $TARCMD > $DATE.idx
134 if [ -s $TARFILE ]; then
135 # File is >0 in size, we assume it worked.
137 # Move tar file in place
138 mv $TARFILE $DATE.tgz
141 echo $DATEs > ../.date && touch -t $DATE ../.date
143 # Make a sorted filelist as well
144 sort -k6 < $DATE.idx > $DATE.sdx
147 chmod 600 *tgz #only for us
148 chmod 644 *sdx *idx 2>/dev/null #everyone can read
150 # Let the remote computer know that we ran a successful backup (for nagios)
151 ssh root@$computer touch $filesystem/.lastbackup
154 # Something wrong happened.
156 echo `date`" $computer: $TARFILE empty. $backuplevel backup of $computer:$filesystem failed and deleted"
158 # We don't want to return 0
164 for computer in $unixcomputers
167 # Check that the target filesystem is mounted (actually check that it's not
168 # the root filesystem)
169 rootfilesystem=`df -P /`
170 targetfilesystem=`df -P "$storagedir/$computer"`
171 if [ "$rootfilesystem" == "$targetfilesystem" ]; then
172 echo `date`" $computer: Target filesystem ($storagedir/$computer) was mounted on /. Aborting"
176 echo `date`" $computer: Backing up $computer" >&2
178 # Try to SSH to the computer without entering a password.
179 if ! `ssh -n -o NumberOfPasswordPrompts=0 root@$computer /bin/true`; then
180 echo `date`" $computer: Could not use passwordless SSH."
182 # We don't want to return 0
190 # Check dump bit in fstab
191 filesystems=`ssh -n root@$computer "cat /etc/fstab" \
195 | awk '{ if ( $(NF-1) != "0" ) print $2}' `
197 # Clean up our dir at this client
198 if ! ssh root@$computer "rm -r ~/.backup ; mkdir -m 700 ~/.backup"; then
199 echo `date`" $computer: Could not create backup staging area at $computer:~/.backup - skipping backup of $computer"
204 # Try to copy $exclude to $computer
205 if ! scp $exclude root@$computer:~/.backup/exclude > /dev/null; then
206 echo `date`" $computer: Could not copy exclude.txt to $computer - skipping backup of $computer"
211 # Try to copy preeexec and postexec if they exist
212 if [ -f $confdir/preexec.$computer ]; then
213 if ! scp $confdir/preexec.$computer root@$computer:~/.backup/preexec >&2; then
214 echo `date`" $computer: Could not copy preexec.$computer to $computer:~/.backup/preexec - skipping backup of $computer"
219 if [ -f $confdir/postexec.$computer ]; then
220 if ! scp $confdir/postexec.$computer root@$computer:~/.backup/postexec >&2; then
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 2>/dev/null
242 mkdir -m 755 -p $storagedir/$computer/$sfilesystem/daglig 2>/dev/null
243 echo $filesystem > ${storagedir}/${computer}/.${sfilesystem}.name
244 chmod 644 ${storagedir}/${computer}/.${sfilesystem}.name
246 # Default backuplevel
249 if [ ! -f $storagedir/$computer/$sfilesystem/.date ]
251 # Take the first full backup of this filesystem on this computer
253 echo $DATEs > $storagedir/$computer/$sfilesystem/.date
256 # Check if we want a full backup - Debug statement that causes noise in cron mail:
257 # printf " $computer $filesystem: $confdir/fastfullbackupdag.$computer.$sfilesystem\n"
258 if [ -f $confdir/fastfullbackupdag.$computer.$sfilesystem ]; then
259 fullbackup_min_for_this_machine=$dagermellomfulleforfast
260 if [ "$DAY_OF_MONTH" = "`cat $confdir/fastfullbackupdag.$computer.$sfilesystem`" ]; then
263 elif [ -f $confdir/fastfullbackupdag.$computer ]; then
264 fullbackup_min_for_this_machine=$dagermellomfulleforfast
265 if [ "$DAY_OF_MONTH" = "`cat $confdir/fastfullbackupdag.$computer`" ]; then
269 fullbackup_min_for_this_machine=$dagermellomfulle
272 if [ -z "`find $storagedir/$computer/$sfilesystem/full/ -name \*tgz -mtime -$fullbackup_min_for_this_machine`" ]; then
277 cd $storagedir/$computer/$sfilesystem/$backuplevel || die
279 # Perform the actual backup
282 # Check if this box has a custom number of full backups
283 if [ -f $confdir/maksfulle.$computer ] ; then
284 mf=$((`cat $confdir/maksfulle.$computer`+1))
286 mf=$(($maksantallfulle+1))
289 # Delete old full backups
290 for full in `ls -1t $storagedir/$computer/$sfilesystem/full/*tgz | tail -n +$mf`
292 prefix=`echo $full | sed "s/\.[^.]*$//"`
293 echo `date`": $computer:$filesystem Deleting full backup $prefix" >&2
297 # Delete incremental backups older than the oldest full backup
298 oldf=`ls -t1 $storagedir/$computer/$sfilesystem/full/*tgz | tail -1`
300 $storagedir/$computer/$sfilesystem/daglig \
303 -printf "`date`: $computer: Deleting old incremental backup: %p\n" \
307 # Try to run postexec if it exist
308 if ! ssh root@$computer "[ ! -f ~/.backup/postexec ] || /bin/bash -x ~/.backup/postexec" >&2; then
309 echo `date`" $computer: Could not run $computer:~/.backup/postexec"
321 # Did anything go wrong?
322 # (IMPORTANT NOTE: The R-business does not work as expected, since this script
325 echo `date`": Backup run ended with errors, check logs."
328 echo `date`": Backup run ended" >&2