[Nottingham] Live system backup pitfalls / problems

[Martin]

Thanks for the comments...

Robert Postill wrote:
[...]
> 1) Make sure that your boot disk boots your system (I use Linux Bootable
> Business Card http://www.lnx-bbc.net quite a lot, the form factor really
> is great).  Particularly if you're going to store the tar image on NFS
> check that you can bring the network card up on whatever you want to
> restore to.

Yep.


> 2) Checksum your image, even if it's a tar image.

I rely on a full readback verify at the moment, though additionally 
using checksums is a good idea.


> 3) Try to make sure you get the right files (i.e. /dev is more useful
> than you think and /proc is a positive hindrance).  Also think about
> hidden file (like ~/.profile).

Ooops... Been relying on the startup scripts to sort out the /dev mess. 
Capturing that also would be good for diagnostics at least...

Good reminder... Note that there's a couple of .files in / even.


> 4) Try and test the image out a couple of times.  Particularly if you
> have SCSI/RAID, think about what you have in your initrd, it's a

!!! Yep! (:-))

Using lilo in any case.

Already tested on my raid machine here... (Also very nearly retested on 
a rushed upgrade just before Christmas... Perl 5.6 -> Perl 5.8... (:-O))


I was looking for comments on such as not backing up tmp directories and 
caches. Assuming a cold boot up they shouldn't be a problem unless you 
want them for diagnostics.


However, what of active stuff in such as /var and for window managers. 
Manually clear out the 'lock' files or wait for the bootup hiccups?

Any other 'system' stuff?

(Tempted to script a tidy-up for whenever a rapid restore is ever 
required...)


So far, I've been healthyly paranoid enough not to have had one of my 
systems clobbered, so far...

And don't forget to backup your partition tables!


Further thoughts?

Cheers,
Martin


-- 
----------------
Martin Lomas
martin at ml1.co.uk
----------------