> **But the nfs messages into the syslog persists. What can we do
> against it? If one gets this error, NFS seems to be very
> slow. And sometimes one gets this kind of error with amd, which
> should be fast enough to fulfill the requests. **
The slowdown is 'normal' (well at least partially). The whole problem
is lack of bandwidth. Unfortunately, the SunRPC code in linux-2.2
handles this pretty badly:- it keeps pumping data at the socket,
irrespectively of whether the socket is able to cache more data or
not.
A solution has been implemented in linux-2.3 (and in the linux-2.2 +
NFSv3 patches) whereby the SunRPC code takes into account the state of
the socket buffer, and will cause requests to sleep if it is full. In
addition, you have the option of using NFS over TCP (if your server
supports it) which further improves performance over noisy connections.
Cheers,
Trond
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/