Routing issue or Sendmail issue?

Routing issue or Sendmail issue?

Post by Ronald O. Christia » Sat, 22 Feb 2003 07:11:51



Hey folks, I am trying to administer a pair of Cobalt web server
appliances remotely and would welcome your advice.

One's a Raq 3 and the other is a Cube.  The Raq 3 appears to be the
main server.  The main problem appears to be that although the Raq 3
and the Cube can both get to the outside world, they can't talk to
each other.  (Error is "no route to host".)

Both machines have two interfaces, one on a local LAN using unroutable
class C addresses (192.168.x.x) and a connection to the outside world
with "real" addresses.

The "real" address for the Raq is the default gateway of the network
on which the "real" addresses reside.

The machines can contact each other on the unroutable interface but
not via the "real" IP addresses.  I suspect that this is a feature.

So, I put entries in the hosts files on both machines so that their
names would resolve to the local unroutable IPs, and most things
(telnet, et al) started working again.

The holdout is sendmail.  Sendmail apparently doesn't consult
/etc/hosts, and skimming the sendmail.cf, it looks like neither
HostsFile nor ServiceSwitchFile m4 options are set.

So, now I'm looking at finding the original m4 file, (or writing one),
adding HostsFile and ServiceSwitchFile, and regenerating sendmail.cf.

But, I'm beginning to think that I'll continue to find communication
issues between the machines, and perhaps I should instead try to get
the "real" IPs talking to each other.

My idea is to tell both machines that the other machine's "real" IP
address is on the local network.  As so:

IP routing table for machine 1:
Destination             Gateway
machine2_real_ip                0.0.0.0

IP routing table for machine 2:
Destination             Gateway
machine1_real_ip                0.0.0.0

Would this work, or should I be doing something else?  For instance,
should the gateway for machine 2's "real" IP be machine 2's unroutable
IP?  This goes a bit beyond my understanding of this stuff.

All assistance gratefully accepted.

Thanks.

                Ron
-
"Save the whales.  Collect the whole set."
http://roc85.home.attbi.com

 
 
 

Routing issue or Sendmail issue?

Post by John Oliv » Sat, 22 Feb 2003 07:18:59



> One's a Raq 3 and the other is a Cube.  The Raq 3 appears to be the
> main server.  The main problem appears to be that although the Raq 3
> and the Cube can both get to the outside world, they can't talk to
> each other.  (Error is "no route to host".)

What do you mean by "main server"?

Quote:> Both machines have two interfaces, one on a local LAN using unroutable
> class C addresses (192.168.x.x) and a connection to the outside world
> with "real" addresses.

> The "real" address for the Raq is the default gateway of the network
> on which the "real" addresses reside.

Then I'm amazed that either can connect to anything *besides* each
other.  The "default gateway" on any given subnet is the way *out* of
that subnet.  If a device is its' own default gateway, it won't be able
to reach anything.

Quote:> The machines can contact each other on the unroutable interface but
> not via the "real" IP addresses.  I suspect that this is a feature.

No... it's probably a PEBKAC error :-)

To fix... let's pretend your "real" subnet is 172.16.1.0/28  That would
mean that 172.16.1.1 through 172.16.1.14 are valid IP addresses.  One of
those, most likely 172.16.1.1 or maybe 172.16.1.14, would be the router,
which would also be the default gateway.  Make your Cobalts 172.16.1.2
and 172.16.1.3 and give them the correct subnet mask and gateway, and
you ought to be set.

Quote:> So, I put entries in the hosts files on both machines so that their
> names would resolve to the local unroutable IPs, and most things
> (telnet, et al) started working again.

It's a band-aid, not a fix.

Post the actual routing tables of these machines if you need more help.
Also, how is the physical cabling laid out?  That might be important
too.

--
John Oliver, CCNA                            http://www.john-oliver.net/
Linux/UNIX/network consulting         http://www.john-oliver.net/resume/
***               sendmail, Apache, ftp, DNS, spam filtering         ***
****                Colocation, T1s, web/email/ftp hosting          ****

 
 
 

1. 4500 issue with veritas/clarrion issue

Hello all,

I've got an Enterprise 4500(solaris 2.6) with 500GB of mirrored Clarrion
drives thats having a few issues. Yesterday we noticed by running
"netstat" that there were 9435 connections to the machine in a
"CLOSE_WAIT" state. We were not able to stop any applications on the box
(ie: Oracle) so I just issued the "reboot" commad. When the machine
boots backup, it goes through POST with takes about 10 minutes, then it
tries to check all of the filesystems and just hangs. According to
sunsolv, it will take up to 40 minutes per 100GB of disk space to run
the check. The problem is that the check never finishes. Has anybody
else had experience with the same setup and experinced the same sort of
issues? If so, what was your fix? Thanks in advance.
--
Brian
http://www.eviscerated.net

----BEGIN GEEK CODE BLOCK----

GMC/E$ d--- s:>: a-- C++++ UBLIS++++$ P++ L++ E--- W++(-) N+ o+ K- w---
O- M- V-PS+ PE+
Y+ PGP++ t- 5--(--) X++ R tv+ b+ DI++ D+ G e*(++) h**** r+++ y++++ z**

----END GEEK CODE BLOCK----

2. Email and Masquerade

3. The "NT vs. Linux" thread not the issue...real issue elsewhere

4. almost complete newbie tries to install xfree86 4.1.0 and...

5. Changing 'issue' & 'issue.net' with Mandrake?

6. Tricky question about '_init' function of a shared lib

7. KDE3/RH 7.2 paths issue - icons for apps issues etc.

8. LILO boot Win98 on 2nd drive

9. The issue file issue

10. issue/issue.net & escape sequences

11. No /etc/issue issued on remote connections...

12. Apache, cgi forms, cgi programs - issues - restart resolves issue??

13. Lilo Issue and Library Issue on a Redhat 6.2 System [Newbie]