Re: CVS and NFS

[Larry Jones]

Steve Ramage writes:
>  
> The short question:
> I guess the short question to this post is the discussion in
> http://mongers.org/cvs#cvs_nfs <blocked::http://mongers.org/cvs#cvs_nfs>
> still mostly relevant to modern day NFS implementations and NAS storage
> devices, and do we risk NFS corruption. 

The short answer: yes, it's still relevant; and yes, you're risking NFS
corruption.  However, you're in the very low risk category right now --
your file server is a dedicated NAS appliance that should have been
throughly tested with most popular NFS implementations and your clients
are all using the same NFS implementation (Solaris's).  You're probably
at no more risk of repository corruption than someone with a repository
on a directly connected ATA disk.  Nonetheless, I'd advise running the
contrib/validate_repo (or check_cvs, depending on which release of CVS
you have) script periodically so that if you do suffer from corruption,
it won't go unnoticed for so long that it's difficult to recover from.

Adding Windows clients using NFS will greatly increase your risk and I
strongly encourage you not to do so.  Designate one (or more) of your
Solaris machines as a CVS server for the Windows machines and use
client/server mode on the Windows machines.

NFS mounting working directories is generally OK.  I don't believe I've
ever seen a report of NFS corruption in a working directory and, if it
did occur, it would likely be noticed very quickly and be very easy to
recover from by simply refetching the file from the repository.

-Larry Jones

OK, what's the NEXT amendment say?  I know it's in here someplace. -- Calvin