Re: SV: Different kind of access methods

[Mike Castle]

On Wed, Dec 13, 2000 at 08:21:15PM -0600, Richard Cobbe wrote:
> I should probably have been clearer---our repository is currently on a
> disk that we mount via NFS.  What are the issues involved with that
> specific situation?

Buggy NFS implementations.  Client and server.

Not to mention performance.

One particular item I've seen is Linux's buggy user space nfsd, especially
with dealing with nfsd exporting two partitions (in my case, /usr and
/usr/src were two partitions on the server, but exported simply as /usr).

I could reliably (that is, it would break in exactly the same spot each
time), by doing a rather large co (test case I used was a series of cvs
imports for the entire historical linux binutils releases).  What I was
seeing was a file turning into a directory, or vice-versa.  (I imagine it
had something to do with an inode in the cvs repository matching an inode
in something like /usr/bin, and the client getting all confused).

Granted, this was a Linux bug.  But it definitely shows the potential
problems.

And getting back to performance, the client (nfs/smb/nw) has to
open/read/write/close the entire archive file.  This sucks over a
networked file system.  Better to let the local machine do all the
accessing and send you just the pertinent parts.

mrc
-- 
       Mike Castle       Life is like a clock:  You can work constantly
  address@hidden  and be right all the time, or not work at all
www.netcom.com/~dalgoda/ and be right at least twice a day.  -- mrc
    We are all of us living in the shadow of Manhattan.  -- Watchmen