Subject: Re: NFS/RPC and server clusters
To: der Mouse <mouse@Rodents.Montreal.QC.CA>
From: Matthias Drochner <>
List: tech-net
Date: 10/16/2003 17:50:22
mouse@Rodents.Montreal.QC.CA said:
> That's weird; there's no reason it should have to reconnect, since the
> peer's IP address is remaining constant. 

As Jason said, the internal state gets lost. (I'd imagine the failover
mechanism does more or less an "ifconfig ifX delete service1" on one
box and "ifconfig ifX service1 alias" on the other.)

>     -C      For UDP mount points, do a connect(2).

This wouldn't help, the kernel only knows the wrong server
address. And if i fixed the mount argument, nothing would work because
replies from the wrong address get ignored.

>     -Q      Explicitly specify the port number to be used for NFS traffic,
>             rather than querying the portmapper on the target host.

This would help probably, but i really think there is something wrong
with the RPC code. I'll try to set up some tests next week.

best regards