We have 2 x45xx thumpers that do sync's over a backend, point-to-point
network connection. Both systems running Solaris 10.

Since this is a point-to-point, the default route is set to the
interface.   Unfortunately, when the system reboots, Solaris removes
this and calculates a 10/8 route which effectively obliterates all
network traffic (we use part of the 10 space elsewhere).

The table should look like this:

Routing Table: IPv4
  Destination           Gateway           Flags  Ref     Use
Interface
-------------------- -------------------- ----- ----- ----------
---------
default              10.250.0.1           UG        1
4919
10.245.137.0         10.245.137.20        U         1        670
e1000g2
10.250.0.0           10.250.0.2           U         1       3771
e1000g0
10.251.0.0           10.251.0.2           U         1       5624
e1000g1
224.0.0.0            10.250.0.2           U         1          0
e1000g0
127.0.0.1            127.0.0.1            UH        5        937 lo0

The point-to-point is the 10.245.137 net.   You an see where a route
for 10.0.0.0 would become a bit of a problem here.

So my question is, ultimately, how can I convine Solaris to keep this
route intact after a reboot?


Thanks.

"forrie@gmail.com" <forrie@gmail.com> writes:
>We have 2 x45xx thumpers that do sync's over a backend, point-to-point
>network connection. Both systems running Solaris 10.

>Since this is a point-to-point, the default route is set to the
>interface.   Unfortunately, when the system reboots, Solaris removes
>this and calculates a 10/8 route which effectively obliterates all
>network traffic (we use part of the 10 space elsewhere).

Sounds like two things. 

First, you probably have to edit /etc/netmasks to instruct Solaris
that you are using the 10.x.x space not as ancient Classful A space. 
With entries for each of your network segments. 

Second, I believe if you are new enough patch level/update your route
command could have a '-p' flag to make the route persistant across
reboots. This wasn't in the first many Update revs, it is pretty new
in 10 IIRC. 

If it does not, then you'll have to write an rc script to be launched
on boot setting up your static route entries. 

forrie@gmail.com <forrie@gmail.com> wrote:
> We have 2 x45xx thumpers that do sync's over a backend, point-to-point
> network connection. Both systems running Solaris 10.
> 
> Since this is a point-to-point, the default route is set to the
> interface.   Unfortunately, when the system reboots, Solaris removes
> this and calculates a 10/8 route which effectively obliterates all
> network traffic (we use part of the 10 space elsewhere).
> 
> The table should look like this:
> 
> Routing Table: IPv4
>  Destination           Gateway           Flags  Ref     Use
> Interface
> -------------------- -------------------- ----- ----- ----------
> ---------
> default              10.250.0.1           UG        1
> 4919
> 10.245.137.0         10.245.137.20        U         1        670
> e1000g2
> 10.250.0.0           10.250.0.2           U         1       3771
> e1000g0
> 10.251.0.0           10.251.0.2           U         1       5624
> e1000g1
> 224.0.0.0            10.250.0.2           U         1          0
> e1000g0
> 127.0.0.1            127.0.0.1            UH        5        937 lo0
> 
> The point-to-point is the 10.245.137 net.   You an see where a route
> for 10.0.0.0 would become a bit of a problem here.
> 
> So my question is, ultimately, how can I convine Solaris to keep this
> route intact after a reboot?

use persistent routes.

route -p add blah blah blah

they live past a reboot.

"forrie@gmail.com" <forrie@gmail.com> writes:
> We have 2 x45xx thumpers that do sync's over a backend, point-to-point
> network connection. Both systems running Solaris 10.
>
> Since this is a point-to-point, the default route is set to the
> interface.

What does "set to the interface" mean here?

In general, if you have a route (default or otherwise) configured for
a point-to-point interface, the next hop address on that route MUST be
the remote address on the point-to-point interface.

If you could show an example of what the system looks like with the
link "up," that would probably help.  In the "netstat -nr" output you
posted, the default route was present, and was pointing to the e1000g0
interface because the address was 10.250.0.1, which is reachable over
that interface.  No point-to-point interface was present in that
output.

It's unclear if that's what you wanted.

>   Unfortunately, when the system reboots, Solaris removes
> this and calculates a 10/8 route which effectively obliterates all
> network traffic (we use part of the 10 space elsewhere).

At a guess, that's because you've configured your Ethernet interfaces
with IP addresses only and without a mask.

You can either use /etc/netmasks to set up a mask for that network
(yuck!), or do the modern CIDR configuration with a mask per
interface.

In other words, instead of this:

   /etc/hostname.e1000g0
        10.250.0.2

you should use this:

   /etc/hostname.e1000g0
        10.250.0.2/24

where the "/24" means that you're using a prefix length of 24 bits,
which is a netmask of 255.255.255.0.  (If you insist, you can also use
"netmask 255.255.255.0" in that file instead of the CIDR "/24"
notation.)

Or if you want to use /etc/netmasks (not what I would recommend, as it
doesn't work too well with CIDR, but apparently liked by other users):

        10.0.0.0        255.255.255.0

> 10.245.137.0         10.245.137.20        U         1        670
> e1000g2
[...]
> The point-to-point is the 10.245.137 net.   You an see where a route
> for 10.0.0.0 would become a bit of a problem here.

I'm confused.  10.245.137.0/24 is on e1000g2, which is an Ethernet
interface.  Ethernet is a broadcast-type medium, regardless of
topology.  It's not an IP "point-to-point" interface.

-- 
James Carlson         42.703N 71.076W         <carlsonj@workingcode.com>

On Nov 18, 2:09=A0pm, "for...@gmail.com" <for...@gmail.com> wrote:
> We have 2 x45xx thumpers that do sync's over a backend, point-to-point
> network connection. Both systems running Solaris 10.
>
> Since this is a point-to-point, the default route is set to the
> interface. =A0 Unfortunately, when the system reboots, Solaris removes
> this and calculates a 10/8 route which effectively obliterates all
> network traffic (we use part of the 10 space elsewhere).
>


how are you configuring this "point-to-point" connection?

you "should" (in sol 10 or earlier) be using /etcc/
hostname.interfaceBLAH

it also seems like you are using .0, for interface IPs
DONT DO THAT. That's probably what's messing things up.
And even if it isnt, it makes the netstat output look misleading.

If you had, in contrast, something like

/etc/hostname.e1000g2
10.250.0.1  10.250.0.2

it might work out okay on reboot. But if not, you can further extend
the contents of the file, thus:

/etc/hostname.e1000g2
10.250.0.1  10.250.0.2  netmask 255.255.255.255

The applicable magic here,  is that the system basically calls

ifconfig $INT  `cat /etc/hostname.$INT`

now, for sol11 on the other hand, Idunnoitswierd.