| nfs3: rfs3_lookup: failed for server <server>
RPC: Can't decode result
I wrote the NFS over TCP code, but have no good explanation for this.
There are a couple glitches, but I don't think they explain this. Are
client or server SMP?
Tcpdump traces would be helpful, (see note 2328), but tcpdump will often
lose synchronization with NFS over TCP reads and writes as it only looks
for NFS requests at the beginning of packets.
While TCP is reliable and UDP unreliable, RPC provides most of the reliability
in UDP. The NFS over TCP is better in congested networks (TCP implementations
can do better congestion control) and TCP's fast retransmit algorithm often
provides better performance in lossy/errorful environments.
Improved reliability was not a major design goal, as the NFS community has a
decades experience with coexisting with UDP. Better behaviour of WANs was the
primary goal.
N.B. The Notes/news gateway for dec.notes.basesys.unix appears broken - Email
will get answers faster.
[Posted by WWW Notes gateway]
|