Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!cs.utexas.edu!ginosko!usc!apple!vsi1!frame!bugeater.UUCP!tvb
From: tvb@bugeater.UUCP (Terry Bush)
Newsgroups: comp.sys.apollo
Subject: Re: Apollo NFS V2.0
Message-ID: <3601@frame.UUCP>
Date: 23 Oct 89 23:27:16 GMT
References: <758@idacom.UUCP> <239@ticipa.ti.com>
Sender: news@frame.UUCP
Organization: Frame Technology, 1010 Rincon Cir., San Jose, CA 95131
Lines: 53
In-reply-to: clif@ticipa.ti.com's message of 20 Oct 89 19:10:09 GMT
Phone.......: (408) 954-3900 (Office) "tvb@frame.com (email)"


> From article <758@idacom.UUCP>, by danny@idacom.UUCP (Danny Wilson):
> >
> > To realize this, we run the NFS servers on a node one the ring _different_
> > from the gateway node to the Ethernet (which connects to some Suns and a
> > Vax)
>
> I am assuming that you mean the your gateway node is not running as a
> NFS server.  Running the gateway node as a NFS server won't cause you
> any problems.  In fact it may be part of your problem.  We had a problem
> quite similar to this.  We mounted files from our microvax to our
> gateway node which had NFS running.  Some of our nodes didn't have
> NFS running, these nodes could see the NFS mounted files but could not
> access them properly, they timed out trying to access them.
>
> Just because you have NFS running on some of your nodes doesn't
> mean all your nodes can use the NFS mounted files.  Those nodes that
> don't have NFS installed probably won't work correctly.
>
> I suppect that if your gateway node isn't running NFS it may be
> having trouble handling the NFS requests.
>
> I run NFS on all the nodes in my network and I don't have any NFS problems.

This is not entirely true.  Your problems may be one of the following:

1. TCP/IP is not configured such that nodes on one side of the gateway
   or the other are not aware of the other network.  Run
   '/usr/ucb/netstat -r' to determine if the hosts know about the
   gateway.  To solve this you can run an active router on the gateway
   node and passive routers on non-router nodes (On Apollos run
   '/etc/routed -q -h' for the passive routers.)  -OR- Run '/etc/route'
   to add knowledge of the gateways on all nodes in question.


2. NFS type managers are not installed.  To determine this, run '/com/ld
   -a <mount_point>' (fill in <mount_point>) on a node that is having
   trouble.  If the file type should be 'nfs_gate' for the mount point
   and 'nfs_dir' or 'nfs_file' inside the tree.  To fix this, install
   NFS on all hosts as a 'link' to another node.  This will install the
   type server, etc. on each host, but will not install the files in
   /etc that take up so much disk space.

I run NFS routinely on my Apollos and always have to rsh to the gateway,
then to the remote host in order to run /etc/route so they all know
about each other.  Maybe other vendors than Apollo will supply the '-h'
option to routed that makes it go away after the routing tables
stablize.  This is very nice of Apollo.


	Peace,
	Terry V. Bush
	The Veritable Bugeater