Paul Bijnens <paul.bijnens AT xplanation DOT com> writes:

> Kevin Dalley wrote:
>
>> Oops.  I forgot to mention the version number
>> I am using a CVS release of amanda, which is named Amanda-2.4.5b1,
>
> should be fine, but...
>
>>>Kevin Dalley wrote:
>>>
>>>>Last night, my amanda run had a problem.  The partition on which
>>>>infofile is stored ran out of space.  With infofile set to
>>>>/var/lib/amanda/normal/curinfo, the file:
>>>>/var/lib/amanda/normal/curinfo/condor/__redbird_l$
>>>>had 0 length, which meant that all history for this partition was
>>>>lost.  I restored the file from backup, and things look better.
>>>>Is it possible to write the new file in a way that the old file is
>>>>not
>>>>deleted until the new file is known to exist?
>
> The problem here is that, if you would do that, the file does not
> contain the current info anymore.
> Does it make sense to keep the old file?
> On the other hand, does it make sense to have an empty file instead?
>
> Also the file is written to when the backup is already done.
> Whatever you decide, the information is not consistent with the reality
> anyway. Any good argument to prefer ancient over empty ?


I think that I prefer the old file to an empty file, but I'm not sure
that I understand everything this file does.  An empty curinfo file
gave me bad results when I typed:

      amadmin normal due

It said:

   new disk condor://redbird/l$ ignored.

Though:
        amadmin normal find condor //redbird/l\\$

works


The old curinfo file would at least have showed that there were some
backups of this partition, even if it didn't have the latest
information.  If amanda could pretend that the latest backup hadn't
happened, that would be better than having no backups.


-- 
Kevin Dalley
kevin AT kelphead DOT org