On Sun, 05 Apr 2009 06:20:29 +0200, Kevin Keane <subscription AT kkeane DOT 
com>  
wrote:

> Close, but not quite. The junction, and anything underneath, is simply
> disregarded. As John Drescher mentioned, you probably have to recreate
> it manually.
>
> The data itself actually doesn't even sit under the junction in the
> first place. It's somewhere else in your file system. And it would get
> backed up *in that place*.

Unless it's excluded for some other reason. I would agree this is  
unlikely, but not impossible.

>> But what happens if the data has to be restored? Will the actual data be
>> restored together with the junction?
>>
> The actual data will be restored with the whatever directory it actually
> resides.

The important bit is, how do you handle the junction point? If Bacula  
doesn't understand it, like now, you can either back it up explicitly,  
which is redundant and quite possibly breaks any updates for apps that  
expect a junction point instead of a real directory (or update files in  
one location expecting the other to be in sync), or you can ignore it and  
probably break whatever is using it immediately after a restore, unless  
you know why something breaks and how to fix it.

Neither option is very good. You might argue that the latter option is  
better, since it's best to fix something immediately than to have a latent  
problem that's harder to troubleshoot later, however when you need a  
restore you tend to be under SLA, already sleep deprived etc. :)

Bacula should handle this itself intelligently: on backup note in the  
Catalog it's a junction point and then check whether the original location  
is backed up, if so do nothing else, if not, back up enough of the  
structure referenced by the junction point so that it can be restored and  
complain in warnings. Do the same in reverse for restores, i.e. restore  
junction point and if it's not in the fileset enough of the hierarchy it  
references so the structure underneath it can be restored (and possibly  
complain again).

If the intelligent bit is too much work you could possibly leave the  
junction dangling like a (soft) symlink without target in *IX, although  
I'm not familiar enough with Windows to say this will work without  
problems.


------------------------------------------------------------------------------
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users