NFS write() calls lead to read() calls?

[Ulrich Spoerlein]

On 3/28/07, Ivan Voras <ivoras at fer.hr> wrote:
> > The traffic flows like this:
> >
> >  ----->   ----->
> > A        B        C
> >           <-----
>
> Have you tested with a small C program or a script to see if it's the
> rewriting that's causing reads or is it some weird artifact of scp?

I'm not a big user of scp(1), the reason this problem came up is
because Samba has the same problem too. If our Windows Terminal server
(over)writes onto the Samba-Export (which in turn is NFS mounted from
the fileserver) then we get WRITE *and* READ calls. nfsstat -s shows
that the READ column is increasing just as fast/slow as the WRITE
column.

So while I could work around scp anytime (tar|ssh is better in almost
all cases) there is probably no way to work around Samba.

What kind of C program or script did you have in mind? My C-foo is very weak ...

Cheers,
Uli