[OpenAFS-devel] AFS + GFS = crazy?

Mitch Collinsworth mitch@ccmr.cornell.edu
Wed, 2 Aug 2006 12:28:21 -0400 (EDT)


>>>> and yet people still backup raid filesystems.  there are certain
>>>> critical failures that cannot be solved with this solution.
>
>>> And which critical failures would that be?
>
>> They're not geographically redundant, for one thing.
>   (but I said they that they were...)

Sorry, I meant raid filesystems are not geographically redundant, that's
why we still back them up.  I'm trying to argue in support of your GFS
idea here.  :-)


>   As I mentioned in another message, they do lazy delete, and you
>   schedule when space is reclaimed.  It might actually be an
>   interesting idea to extend the lazy-delete idea to simply
>   making a new copy of a chunk everytime it is written -- giving
>   you the ability to go back to any instant in time (until you
>   cleanup the old versions).

Yes.  It could also give sites that keep periodic archives "forever" the
means to just tape the stuff that's so old we don't want to keep it on
disk any longer.  The trick then is having the needed meta-data onhand and
in the right format to be able to *find* any file someone might happen to
ask for.

-Mitch