Originally quoted from www.digriz.org.uk (but the page no longer exists)…because the rsync support for backupninja (or at least the Debian one) is a bit naff…or I just simply do not understand it. Very true if a little surprising since all the other backupninja handlers have always worked very well for me. Backupninja is wonderful for creating a backup system, run these shell commands, hotcopy those databases, rdiff-backup the result and rsync or duplicity all that somewhere else. Wonderful stuff now lets fix rsync.
P.S: if you do know how to make the supplied rsync handler work then please post a comment.
The backupninja supplied rsync handler might be able to do rotations when all I want is a mirror…but as far as actual backup even the simplest config causes trouble.
[general] mountpoint = /mnt/1 days = 30 [source] from = local include = /tunes
Note: example taken from an official backupninja bug, I have also tried and failed with many other combinations.
All I get is this:
Info: >>>> starting action /etc/backup.d/10.rsync (because of --now) Debug: yes Info: Rotating /mnt/1///tunes... Info: Syncing /tunes on /mnt/1///tunes/tunes.0/... Debug: nice -n 0 /usr/bin/rsync -av --delete --recursive --numeric-ids //tunes/ /mnt/1///tunes/tunes.0/ rsync: -av --delete --recursive --numeric-ids: unknown option rsync error: syntax or usage error (code 1) at main.c(1443) [client=3.0.7] Info: <<<< finished action /etc/backup.d/10.rsync: SUCCESS Debug: send report to root Info: FINISHED: 1 actions run. 0 fatal. 0 error. 0 warning.
So even though the backup clearly failed there are no errors reported! What the hell is up with all the extra ////’s! The strange thing is that manually running the rsync command that debug outputs works just fine. Smells bad, in every example that I found people are using the .sh handler to rsync just as I was, at least this is simple and works but often suffers from not reporting rsync errors. Seems there is a need for a .rsyncsimple handler.
Turns out that the .sh handler is smarter than explicitly mentioned in the docs, it can report errors!
# This script can run most any command normally run on the command line. # Add anything usefull that you would like to do during a backup. # Handler output functions may be used: # * debug $msg -- log debugging info # * info $msg -- log general information. # * warning $msg -- log a warning # * error $msg -- log an error # * fatal $msg -- log a fatal error and exit # (1) a list of all the packages installed and removed. # this file can be used to restore the state of installed packages # by running "dpkg --set-selections < dpkg-selections.txt" BAK_GET_SELECTIONS="ssh root@host 'dpkg --get-selections > /var/backups/dpkg-selections.txt'" eval $BAK_GET_SELECTIONS || fatal "$BAK_GET_SELECTIONS" # (2) used with along with /etc/fstab and /var/backups/partitions.hdx.txt # to decide on how to partition a new disk during a recovery. # possibly by running "sfdisk /dev/sdx < partitions.hdx.txt" # (MAKE SURE YOU PARTITION THE CORRECT DISK!!!) BAK_DF="ssh root@host 'df > /var/backups/df.txt'" eval $BAK_DF || fatal "$BAK_DF"
Turns out Chris Lamb felt the same way and posted a Debian bug report with a GLP’d .rsync_ssh handler. Thanks Chris!
Sadly I’ve just upgraded to Debian 6.0 “Squeeze” and even with the new ‘mirror’ option I don’t get what I want. So adapting .rsync_ssh I think we can get .rsyncsimple, retain Chris’s ssh functionality and put it on gist.github.com! More tomorrow 01:30 and bed time…
I’ve noticed the same problem with the .rsync handler. I seems unusual that others aren’t hitting this problem, and the issue solved.
I actually have started writing a handler, trying to do it ‘the right way’ I started the tests script first and ran out of time…so my backups haven’t been done lately. Soon I will have to use the .sh handler (but at least with improved error reporting). I have the scripts if anyone would like to add to what I have done so far.