[Info-vax] tcpip gateway question

[John Santos]

In article <mailman.12.1252413824.17612.info-vax_rbnsn.com at rbnsn.com>, 
mexas at bristol.ac.uk says...> 
> On Tue, Sep 08, 2009 at 11:36:48AM +0000, John Santos wrote:
> > In article <mailman.10.1252405909.17612.info-vax_rbnsn.com at rbnsn.com>, 
> > mexas at bristol.ac.uk says...> 
> > > I've a VMS cluster on a local 10.10.10.0/24 network.
> > > 
> > > I'm trying to set up one of the VMS nodes
> > > to also sit on the University network 137.222.0.0/16.
> > > 
> > > So I used tcpip$config and configured the two interfaces as:
> > > 
> > >    1  -  WE0 Menu (EWA0: TwistedPair 100mbps)
> > >    2  -  137.222.187.238/16  mech-cluster238       Configured,Active          
> > > 
> > >    3  -  WE1 Menu (EWB0: TwistedPair 100mbps)
> > >    4  -  10.10.10.1/24       vav                   Configured,Active          
> > >  
> > > 
> > > I added the default University gateway, 137.222.187.250, and
> > > name servers 137.222.10.36 and 137.222.10.39 with tcpip$config
> > > options
> > >                  3  -  Routing               
> > >                  4  -  BIND Resolver         
> > > 
> > > I've ssh server and client enabled on this node.
> > > 
> > > My problem is that I cannot even ping the gateway.
> > > 
> > > Does this look reasonable:
> > > 
> > > $ tcpip show route
> > >   
> > >                              DYNAMIC
> > >   
> > > Type           Destination                           Gateway
> > >   
> > > AN    0.0.0.0                               137.222.187.250
> > > AN    10.10.10.0/24                         10.10.10.1
> > > AH    10.10.10.1                            10.10.10.1
> > > AH    127.0.0.1                             127.0.0.1
> > > AN    137.222.0.0/16                        137.222.187.238
> > > AH    137.222.187.238                       137.222.187.238
> > > $ 
> > >  
> > > 
> > > many thanks for any advice or a link to a relevant manual.
> > 
> > Your net mask is almost certainly wrong.  No one has a class B on an
> > ethernet segment anymore!  My guess is you can't see the name servers
> > because your VMS system is trying to send directly to them and it
> > needs to go through your router.
> > 
> > Can you ping the router (137.222.187.250) by address?  If not,
> > your LAN might be subnetted to smaller than a /24 and you should
> > have a local router to get to the rest of it.  But most likely,
> > your netmask should be 255.255.255.0 (/24).  Ask your network
> > people...
> 
> John, thank you.
> 
> Yes, the netmask should be /24, I just confirmed this with my
> networks administrator. I changed the configuration, but the
> result is still the same.


In another followup, you posted the output of ipconfig and it
looked like the netmask had been fixed, but there was still
something funny about the broadcast addresses.

Both interfaces had a broadcast address of 137.222.255.255,
IIRC.  You had noticed that this seemed bogus for the 10.10.10.0
network (should be 10.10.10.255, since it's a /24), but
also it is wrong for the 137.222.187.0 network (it should be
137.222.187.255 since it's also a /24.)  But I don't know
what effect this would have.  Anyone know if it could possible
be breaking ARP?

I have noticed in the past that if you muck around trying to
fix IP configuration stuff in UCX (aka "Hewlitt-Packard
Transmission Control Protocol/Internet Protocol Services for
OpenVMS", no wonder everyone calls it by the obsolete "UCX"
name :-) :-) :-)) it doesn't always reset everything correctly,
but if you totally shut down and reboot, it does.  So you can
think you've got everything set up correctly (and you're right),
but some little difference bites you back.  I've always found
these things (like the broadcast mask being wrong) can be fixed
without rebooting, but many times the problem isn't obvious!

> 
> I'm not sure I understand this properly, but is it allowed
> or advisable to set different gateways for different nodes
> in a VMS cluster? I think, that's what I'm trying to do. 
> 

I think that should work fine.  But I suppose it is possible
some of the net config that should be node-specific is being
kept in common.  But that would cause problems on your *other*
nodes...

> Perhaps in a setup with common cluster environment tcpip
> setting are supposed to be the same for all nodes? I'm not
> sure if the gateway settings are stored under sys$common
> or under sys$specific. But on the other hand,
> TCPIP SHOW ROUTE show that different nodes have different
> routers.
> 
> I probably sound terribly confused, sorry about that.

It's always darkest just before the dawn ;-)

Good luck!


-- 
John Santos
Evans Griffiths & Hart, Inc.