[OpenAFS] AFS suggestions

Christopher D. Clausen cclausen@acm.org
Wed, 27 Jul 2005 13:09:04 -0500


ERIC K. CHEU <ekcheu@uncg.edu> wrote:
> Any suggestions or tips or how other people do this
> best?  (I'm not looking on how best to move volumes, but how poeple
> arrange their volumes for example.)

In my case, the most important data (user home directories, websites,
cvs, svn) are on the most "trusted" server (the one we have the most
faith that it won't fail, currently a Sun e450.)  Additional servers
don't always have RAIDed storage and a single drive failure can cause
data loss.  But then, we operate with donated hardware and whatever we
can scrounge.

I also use the different vice paritions on the servers to determine what
needs to be backed up.  Currently, the vicepa paritions on our three
primary AFS servers is backed up, with some additional exclutions and
inclusions in specific cases.  But then I have scripts that
automatically look at the volume list and add / remove volumes from the
backup list:
\\AFS\acm.uiuc.edu\admin\scripts\tsm.pl
/afs/acm.uiuc.edu/admin/scripts/tsm.pl

This is pretty crude and I doubt that doing backups on a per-partition
basis is a good idea for you.  But, it works for us, wasn't hard to
setup and requires little effort to backup more volumes via vos moving
the volume to the correct parition and server.

-----

In general, one of the benefits of AFS is that the data is specifically
NOT tied to a particular server and you can arrange volumes any way you
want.  There are several tools available to load balance volumes for
best performance of your cell.

I'm sure other sites have data organized by the entity that owns the
data or by the network topology (volumes closer network wise to the
users.)

Basically, AFS will let you do whatever you want.

<<CDC
Christopher D. Clausen
ACM@UIUC SysAdmin