Re: Intermittent Routing Problem

From: David Kirk (davidrkirk.NOSPAM_at_hotmail.com)
Date: 04/05/04 

Date: Mon, 05 Apr 2004 15:54:59 +1200

On Sun, 04 Apr 2004 17:59:28 -0700, Jeff Liebermann
<jeffl@comix.santa-cruz.ca.us> wrote:

Jeff,

Sorry I haven't been very clear. See my comments inline.

>On Mon, 05 Apr 2004 11:02:11 +1200, David Kirk
><davidrkirk.NOSPAM@hotmail.com> wrote:
>
>>We are migrating our network from a 192.100.100.0 subnet to a
>>192.168.1.0 subnet and I am having problems with our SCO 5.0.4 boxes.
>
>Be advised that RFC1918 designates 192.168.xxx.xxx and others for use
>in private LAN's. Although you can probably make 192.100.100.xxx
>work, I would not recommend it. Changeing the 2nd octet from 100 to
>168 would not be a major project.

That is the whole point of this change. Our local LAN uses
192.100.100.0/24 which is owned by other people out there on the
Internet. We are moving to the 192.168.1.0/24 subnet.

Although we have a firewall which does NAT from our 192.100.100.x
addresses to our external address, the new subnet will comply with
RFC1918.

>Was this working properly *BEFORE* you twiddled with the IP addresses?

Yes, but we only had a single subnet, so we didn't even have a gateway
setup.

>>One box is configured to run an end-of-day procedure and print reports
>>at 04:30. The printer doesn't start printing until 07:00 or later
>>sometimes. Under normal circumstances it would start printing at
>>04:34.
>
>Ummm, what does the aformentioned have to do with a routing problem?

The printer is a network printer. It is not attached to the SCO box
directly. It has a JetDirect card in it. The end of day reports get
stuck in the print queue until something (I don't know what) allows
the server to connect to the remote printer.

>>This box has an IP address of 192.168.1.9 and the printer is
>>at 192.100.100.179.
>
>># netstat -rn
>>Routing tables
>>Destination Gateway Flags Refs Use Interface
>>default 192.168.1.251 UGS 5 227476 net1
>>127.0.0.1 127.0.0.1 UH 3 620393 lo0
>>192.168.1 192.168.1.9 UC 1 0 net1
>>192.168.1.9 127.0.0.1 UGHS 0 1630 lo0
>>224 192.168.1.9 UCS 0 0 net1
>
>Ok, packets to 192.168.100.179 are going out the default route to the
>internet. That's ok.
>
>>The router at the default gateway (192.168.1.251) is a Cisco 1721 set
>>up to pass all traffic between the two subnets.
>
>I'll take your word for it. Try using traceroute on the SCO box to
>see where the packets go between the SCO box and the remote printer.
>
>>The problem only seems to happen if the server hasn't been accessed by
>>a host on the other subnet (or tried to access a host on the other
>>subnet).
>
>What problem? You haven't described what does NOT work or what you're
>trying to achieve. My guess(tm) is that you can't print to the remote
>printer.

The problem is that hosts on the old subnet (including the printer)
cannot connect to the server and vice versa. Telnet, ping, printing,
etc don't work between subnets.

> Can you ping the remote printer?

No.

>If "accessing" (I hate that word, it's so meaningless) the SCO server
>by a local machine on the remote LAN, it could easily be problem with
>whatever you're using for an ethernet switch. A switch needs to see
>traffic in order to obtain the MAC address of the SCO box. One would
>think that the SCO box belches enough broadcasts to not require
>priming, but it's possible. Does pinging the remote SCO box give the
>same results as "accessing" the SCO box?

Yep. "Accessing" = telnet or ping

>>So far I have only noticed this happening with SCO boxes.
>
>You have more than one OSR5 box? Have all the boxes had their IP
>address's and/or names changed or just one?

I have another one that I am changing tomorrow morning. Hosts on the
new subnet can't ping or telnet to it.

Did I mention that the problem is intermittent. This morning when I
came in to the office, I couldn't ping the server on the other subnet.
Sometime during my troubleshooting, it started working and has done
all day. The problem seems to only occur when no hosts on the other
network have communicated for some undetermined amount of time.

>>My Windows
>>XP PC is on the new subnet and I don't have any problems connecting to
>>any hosts on either network (except SCO boxes).
>
>OK, so the undefined problem is unique to the SCO box. The first step
>to solving a problem is blame someone (or something). The SCO box
>will do.

I am blaming the SCO boxes, but I haven't completely ruled out the
router yet either.

>>Does anyone know what might be causing this or have any ideas on how
>>to fix it?
>
>Sure.
>
>1. See:
> http://www.LearnByDestroying.com/sco/new_name.txt
>for a shopping list of files that have the IP address imbedded. You
>probably missed one somewhere. Methinks this is the most likely
>culprit.

Nice. Thanks for that. I will check those files out.

I've already found that the old broadcast address was still in
/etc/default/tcp.

>2. SCO has the nasty habit of running the "routed" daemon which is
>RIP1. If you have any routers or broken Microsoft impersonations of a
>server that advertises RIP1 routes, SCO 3.2v5.0.4 will happily send
>the default route off to who knows where. I advise that you kill the
>"routed" daemon, and edit the file /etc/tcp to comment out the lines
>that start routed. This was sorta fixed in 3.2v5.0.6. I suggest you
>do this even if you find the solution elsewhere.

Done. Hopefully it will help.

>3. Flakey NWAY negotiation on ethernet switch together with buggy
>3Scum ethernet cards that result in the card set to full-duplex, while
>the switch lands in half-duplex. The result is moderate packet loss.
>It will still work but act erratic with occassional hangs. Try file
>copies in both directions between one of the local machines and the
>SCO box. If the bits/sec numbers in both directions look reasonable,
>you don't have this problem. If it hangs, acts erratic, or is
>asymetrical, you have a problem.

Both NIC and switch are forced to 100-FD.

Thanks for your help.

Later

David Kirk

Relevant Pages

  • Calculating subnet hosts
    ... Your company has been assigned a network ID of 197.132.240.0. ... The max number of hosts per subnet is 25. ... Class A addresses have the highest level bit turned ...
    (microsoft.public.cert.exam.mcse)
  • Re: IP Address problems
    ... Accidently mailed this to DLP instead of posting it. ... change your network to /24 or move your router and gateway up into your network. ... mask subnets valid hosts per total hosts ... so can't be used for hosts, so there are 2 fewer hosts per subnet ...
    (comp.os.vms)
  • Re: Can I use swticher instead of router in this case?
    ... Suppose you are designing a campus network with 200 host spead over four ... want each of these four buildings to include 50 hosts. ... By extending the subnet mask and borrowing 2 bits from the host ID portion ... I wonder if I can use a switcher instead of a router to connect other 4 ...
    (microsoft.public.cert.exam.mcse)
  • Re: Subnetting Question
    ... You administer a network that contains 175 machines. ... the IP address 192.168.11.0 with the default subnet mask of 255.255.255.0. ... The network address 192.168.11.192 with a subnet mask of 255.255.255.224 is ... Subnet A because it supports up to 30 hosts. ...
    (microsoft.public.cert.exam.mcsa)
  • Re: Running out of IP addresses
    ... I have a network of 5 servers running Windows 2003 server on the same ... The subnet has 254 possible addresses,...there is no way you would run out. ... Ethernet is not supposed to have subnets bigger than 254 hosts. ...
    (microsoft.public.windows.server.networking)