howard@ntmtv.UUCP (Howard Hart) (11/29/90)
What happens when a directory is NFS hard mounted (rw) over a router
CISCO in this case, and the server machine connection is lost? I know
the client hangs until the route is reestablished, (the server won't
care one way or the other since NFS is "stateless", at least on the
server side) but will this also potentially hang the intermediate
CISCO(s) due to the immmense amount of UDP traffic the client is
shipping out trying to reestablish the connection? If one hard mount
won't do it, what happens if I have 20-30 hard-mounts across the
route? (this will get very common with automount).
--
Howard Hart UUCP:{ames,pyramid!amdahl,hplabs}!ntmtv!howard
System Administrator INTERNET: ntmtv!howard@ames.arc.nasa.gov
Northern Telecom PHONE: (415) 940-2680
Mt. View, CABILLW@mathom.cisco.com (WilliamChops Westfield) (11/30/90)
What happens when a directory is NFS hard mounted over a router CISCO in
this case, and the server machine connection is lost? I know the client
hangs until the route is reestablished, but will this also potentially
hang the intermediate CISCO(s) due to the immmense amount of UDP traffic
the client is shipping out trying to reestablish the connection?
If NFS in fact generates "immense amounts" of traffic when it is unable
to talk to a remote host, then it is very broken. The correct behavior
would be to exponentially back off retransmissions in the absence of
ACKs, and to essentially ignore anything like a "network unreachable"
that is received. I'd be very surprised if NFS would generate more
traffic in the face of network problems - this would be considered very
anti-social in the internet community.
Bill Westfield
cisco Systems.
-------