RE: Load balancing setup for CVS repositories

[Arthur Barrett]

Ramesh,

> I am more interested in the ability of the 
> two CVS servers to coexist and work in parallel. 

I think I answered that already.  Nothing needs to be done to CVS just
like nothing needs to be done to Apache - the configuration is in the
cluster or the router/switch.

> They will be using the same NAS filesystem as 
> their repository area. Is CVS designed to work like that?

I definitely I answered that already.  You'll then need to store your
repository on a SAN.  Storing the repo on an NFS or Samba share (which
is what a NAS does) is a good way to ensure that it gets corrupted - use
a real SAN.  

Apart from anything else a NAS would become a bottleneck far before your
CVS servers would.  What is the point in setting up load balanced CVS
servers if they spend all their time just waiting for the NAS?  CVS is
almost always IO bound, not CPU bound so I think you are probably
investing the time and effort on the wrong thing.  If you are concerned
with price then an iSCSI SAN is almost as cheap as a NAS - but has the
same bottleneck.

Maybe if you describe the business problem that you are trying to solve
then someone may have a better solution than setting up load balanced
CVS repositories?

Regards,


Arthur Barrett