Re: speed: pserver vs mount of repository

[Mark D. Baushke]

Richard Pfeiffer <address@hidden> writes:

> Is it just the ADD/COMMIT over an NFS that would
> cause the corruption, and, presuming an as-of-yet
> uncorrupted repository, doing a CHECK-OUT over an
> NFS mounted repository would NOT cause
> corruption;

A 'cvs add' should not be a huge problem as only directory creation may
go awry which will be easily discovered when a user goes to add and
commit files later.

The dangerous operations are 'cvs commit', 'cvs tag', and 'cvs rtag'
which cause the entire ,v file to be rewritten.

To a lessor extent, the 'cvs edit' and 'cvs watch' commands may corrupt
the repository CVS/fileattr file which may lead to notifications not
being properly transmitted.

With NFS there is also a greater possibility of stale locks in the
repository due to 'lost' delete requests of lock files that may arise
with flaky network connectivity.

Also, a crash during a commit may leave ,file, temporary RCS files in
the repository.

> or 
> would/could a CHECK-OUT over an as-of-yet
> uncorrupted NFS mounted repository also cause a
> revision to become corrupt?

No, modulo the possibility of stale locks arising.
 
> Note:  I suggested to the users we lose the NFS
> mount.  I just want to answer all their (and
> my)questions.

I understand.

> I originally noticed mention of this NFS issue in
> the CVS Manual, but since then have not been able
> to relocate it.  Is it in the Repository chapter
> or elsewhere?  Perhaps someone has the exact
> link?

Look in the cvs-1.11.6/FAQ or cvs-1.12.1/FAQ file for various
problems that have arisen in the past.

See a recent thread on the info-cvs archive:

http://mail.gnu.org/archive/html/info-cvs/2002-05/msg00108.html

http://www.google.com/search?hl=en&q=cvs+NFS+site%3Amail.gnu.org
http://www.google.com/search?hl=en&q=cvs+NFS+site%3Acvshome.org
http://www.cvshome.org/cyclic/cvs/dev-port.html

        Enjoy!
        -- Mark