Linux Redhat 7.2 2 NIC's problem with RIP configuration

[Johan Leroy]

Hello readers,

We have a configuration which consists of a Cisco 2611 router, 2 Cisco Catalyst 1924 switches and 1 IBM X232 server, running Redhat Linux 7.2.
The server has 2 NIC's, through which he is connected to the router (via the switches).

WAN
|
|
Cisco router
| |
| |
switch 1 switch 2
| |
| |
IBM Linux server

We (try to) use RIP between the router and the server to pass the IP address of the default gateway from the router to the server.
When we pull out the patchcord of the Ethernet interface on the router, which is currently the default gateway, we see that the router distributes via RIP that the default gateway is now the IP address of his other Ethernet interface. However, the server does not react to this notification, and does not change his default gateway. Only after 3 minutes the server seems to detect that his default gateway is no longer valid, and changes the default gateway to the second alternative.

We have a similar configuration with a Cisco 2611 router, 2 Cisco Catalyst 1924 switches and 2 HP9000 servers, running HP-UX 11.00.
Here we also use RIP to distribute the default gateway to the servers. Everything works fine.

The RIP configuration on the Cisco router is identical for both configurations (Linux and HP-UX):

!
router rip
version 2
redistribute static
redistribute eigrp 100 metric 1
passive-interface Serial0/0
passive-interface BRI0/0
passive-interface Serial0/1
offset-list 25 out 1 Ethernet0/0
network 10.0.0.0
distribute-list 31 out Ethernet0/0
distribute-list 31 out Ethernet0/1
distance 180
!
access-list 25 permit any
access-list 31 permit 0.0.0.0
!

Explanation:
- The router uses RIP version 2.
- The router only distributes RIP information
over its interfaces Ethernet0/0 and
Ethernet0/1. These are the connections to
the 2 switches.
- Only the default gateway is distributed via
RIP (access-list 31 takes care of that).
- Interface Ethernet0/0 gets an "offset" of 1.
This means that the gateway via interface
Ethernet0/0 is 1 hop-count further than the
gateway via interface Ethernet0/1. Therefore
the router will prefer interface Ethernet0/1
as default gateway. Interface Ethernet0/0
will be his second choise as default gateway.


The content of the /etc/gated.conf file on the IBM Linux server:

traceoptions "/var/log/gated.log" general ;

interfaces {
interface all passive ;
} ;

rip yes {
interface eth0 version 2 ;
interface eth1 version 2 ;
} ;

kernel {
options noflushatexit ;
} ;

static {
default gateway 10.60.105.251 preference 200 retain ;
default gateway 10.60.104.251 preference 210 retain ;
} ;


Possibly important remark is that "IP multicasting" is disabled in the Linux kernel, whereas RIP version 2 uses multicast addresses to distribute information.

Any suggestions?

Johan Leroy
Network Engineer
Infoco n.v./s.a.
Edingensesteenweg 196, 1500 Halle, Belgium

tel: +32 2 360 10 40
fax: +32 2 360 02 07
email: johan.leroy@infoco.be
http: http://www.infoco.be

[U.SivaKumar_2]

hi.
RIP Version 2 can use multicast or broadcast also for distribution . it depends on the configuration.
Anyway try enabling the multicast in your ethernet interface.
after that ,
#ifconfig eth0
MULTICAST- this line should be there in the output
#netstat -rn
does it shows a entry for 224.0.0.9 in the
routing table.
I think what has happened in your case is simple dead gateway detection by linux. But
not because of RIP updates from router.
Can you ping to 224.0.0.1 from the server ?.
There should be no issue with switches for
multicast because they handle multicast in
the same way they handle broadcast.
So try enabling multicast in ethernet interface.

regards,
U.SivaKumar

[Ron Kinner]

Seems to me that if the gateway changes after 3 minutes that either the router is not sending updates (however you said an HPUX box could see the changes so that's out) or there is something in your Linux code which keeps it from reacting to unscheduled updates. It must be getting the regularly scheduled updates since it does change when the 3 minute timer expires. Have you checked to make sure you have the latest version of gated?

Also check to see what the RIP timers are set to by default. They could be set real high and cause this slow response.

I believe gated has a debug option. That might give you a clue about what is going on.

On the Cisco you can run the debug to watch the updates:
debug ip rip
or
debug ip rip events

If you are logged on by telnet you need to do
term mon
to see the debug output on the Cisco.

Also on the Cisco
sh ip protocol

should show you something like this:

Routing Protocol is "rip"
Sending updates every 30 seconds, next due in 1 seconds
Invalid after 180 seconds, hold down 180, flushed after 240
Outgoing update filter list for all interfaces is not set
Incoming update filter list for all interfaces is not set
Redistributing: rip

Now that I think about it I think the RIP is working correctly. Just HPUX has this gateway health detection trick where they ping the gateway every so often to make sure it's alive. Don't think Linux does that. Try turning off rpdp in netconf on the hp and see if it slows down.

Ron

[Fabi]

You may try to tcpdump from your Linux box and see whats happening, tryt with normal NIC operation, then try again in promiscous mode to see if the RIP messages are getting to your Linux box,

cheers