On Tue, 5 May 2009 05:15:31 -0700, Jeff Schroeder
<jeffschroed(a)gmail.com>
wrote:
> On Tue, May 5, 2009 at 3:05 AM, Tom Brown <tom(a)ng23.net> wrote:
>
>> At the moment i use the keep_ssh_host_keys snippet to make sure the host
>> key does not change accross rebuilds to avoid fingerprint trouble, and
>> this works fine.
>>
>> I have a situation where i have to preserve a certain directory /var/log
>> and this is NOT always on a seperate partition so i wonder would it be
>> possible to use a similar trick to preserve this directory? I am not
>> certain as this directory can be quite large.
>>
>> any ideas ?
>>
> Using the same method as keep_ssh_host_keys is not suggested as it
> puts the contents in memory. Since the contents of /var isn't going to
> really mess with the system if there are some stray directories or
> files you should try the --noformat option. In the case that /var is
> on the same volume as /, you really should format it. Just make sure
> to copy the data off first. I've heard of people installing overtop of
> an existing install without formatting, but that is black magic and
> should absolutely be avoided. It could lead to problems that are very
> difficult or impossible to reproduce (and may "void" your support
> agreements for those systems).
>
> Take a look at:
>
http://fedoraproject.org/wiki/Anaconda/Kickstart
>
> Something like this might work for you (not tested):
> part /var --label=var --recommended --noformat --ondisk
>
> If doing the above, make sure to not do "clearpart --all" and instead
> specify the partitions you want to clear or manually do it with some
> parted magic. Maybe this will give you some ideas?
>
Another option would be to mount up a remote share via NFS and rsync the
contents there in %pre, then in %post remount it and sync the contents
back. In order to do this, you'd have to know in which partition the
folder you want sits, and would have to mount that as well in %pre (and be
sure to unmount otherwise installation will probably fail).
This seems to me to imply a workflow that might not have a good backup
procedure.
I'd look at modelling this more as a backup/restore type of scenario.
If the data has to be "preserved", it seems to imply it can't be quickly
recreated from scratch?
--Michael