Re: [rdiff-backup-users] critical irresponsible behaviour

[Dave Kempe]

mortee wrote:
> Hi,
>
> I just had rdiff-backup erase *all* my backups. That is just plain wrong.

Hi mortee,
I have had a similar issue which actually erased my / directory, (well a 
staff member did it). This was because he used --force with a restore 
and ended up erasing his root (/) directories. Made a pretty bad mess of 
things. Basically, I now caution all users of rdiff-backup to restore 
only to a fresh target directory, then move the files into place. This 
way you avoid using --force at all really.
If you would like to submit a patch that prevents --force from really 
forcing and behaving differently, I am sure one of the active 
maintainers would consider it. I am sorry you lost all your data in this 
way, however you can see that if something is designed to force 
something, and makes you use --force to do it, it expects you know what 
you are doing. I agree it could be more considerate, and I have had just 
as bad an experience as you (perhaps).

Anyway, as Warren Guy has requested, I know generally try and backup by 
backups to another location if at all possible. This accounts for not 
only software and filesystem problems, but loss of backup server etc.
It relies on the layout like so:
/storage/backups-repo/servername/jobname/
where jobname is the rdiff-backup repository.

Here is a somewhat sanitized script that does just that, ssh into a 
target server, figure out which backups have worked, and backup them up 
offsite, keeping 3 good backups of the backups.

It relies on the layout like so:
/storage/backups-repo/servername/jobname/
where jobname is the rdiff-backup repository.
-----------------------------
#!/bin/bash
# script to do the offsite backup

# assumes you have keyed ssh, root access to the backuphost



BACKUPHOST=ernie
BACKUPPATH=/storage/nearline-backups
BACKUPDEST=/storage/offsite-backups/


# first find the repos to backup and check them.
# create a list of good repos and back them up.
# email the list of bad repos to  support

LISTFILE=/storage/tmp/current-list-of-backups.txt
GOODREPOLIST=/storage/tmp/current-list-of-completed-backups.txt
GOODHOSTNAMES=/storage/tmp/current-list-backup-sources.txt

# first create the list of all repos on the backuphost
ssh $BACKUPHOST find $BACKUPPATH -type d -maxdepth 3 |grep 
rdiff-backup-data | grep -v "/archive/"  > $LISTFILE

touch $GOODREPOLIST
rm $GOODREPOLIST
touch $GOODREPOLIST

# ok got the list now check it for broken backups
for REPO in `cat $LISTFILE`
       do MARKERCOUNT=`ssh $BACKUPHOST ls $REPO/current_mirror*|wc -w`
               if [ $MARKERCOUNT -ne 1 ]
                       then echo $REPO BAD
                       else echo $REPO >> $GOODREPOLIST
               fi
done

# sort out the unique hostames from the list and add any missing/special 
ones with an echo
cat $GOODREPOLIST |cut -f4 -d "/"|sort|uniq > $GOODHOSTNAMES
#echo scripts >> $GOODHOSTNAMES
echo pictures >> $GOODHOSTNAMES

# then back each one of the hosts up individually
for HOST in `cat $GOODHOSTNAMES`
   do echo "Backing up backup of $HOST from $BACKUPHOST!"
       /usr/bin/time -o /storage/tmp/$HOST.time -f "%E total  %P CPU" 
rdiff-backup -v5 --print-statistics $1 $2 
$BACKUPHOST::/$BACKUPPATH/$HOST $BACKUPDEST/$HOST
   echo "removing increments for $HOST"
   rdiff-backup --force --remove-older-than 3B $BACKUPDEST/$HOST
   echo "done $HOST"
done

# This is a ugly hack for script that I think should be cleaned up later 
but I'm difficult that way - Matt
echo "Backing up backup of scripts from $BACKUPHOST!"
/usr/bin/time -o /storage/tmp/scripts.time -f "%E total  %P CPU" 
rdiff-backup -v5 --print-statistics $1 $2 $BACKUPHOST::/storage/scripts 
$BACKUPDEST/scripts
echo "removing increments for scripts"
rdiff-backup --force --remove-older-than 3B $BACKUPDEST/scripts
echo "done scripts"

# now backing up the stats for future reference
/usr/bin/rdiff-backup /storage/tmp /storage/backup-stats

--------------------------
I should probably put that on the wiki - someone else can if they want. 
(I might have time some other time).

thanks

dave