Strange problem about net communication

[Ryan Gu]

Hi,all

I find a strange problem about network communication.
Desc:
hostA: v11i 171.20.1.1 255.255.248.0
hostB: v11.0 171.20.1.2
255.255.248.0
hostC: v11.0 171.20.96.1
I cann't ping each other between hostA and hostB.But I can ping 171.20.1.200(a cisco switch) and hostC from hostA & hostB.

How can I resolve this problem?Please help me.
Thanks

[John Dvorchak]

You didn't mention what error is being generated. No repsonse or can't resolve the host name? Let's assume you have no response.
The only time I have seen that problem occur, one of the hosts had recently changed their IP address and/or NIC card and they were on different switches. Sometimes the Netjerk guys set their routers/switches to never dump the ARP cache mistakingly thinking that it helps performance by not broadcasting ARP requests every 3 minutes or so. If one of the hosts has changed NIC's recently and the ARP cache isn't dumped every few minutes you can see that one station cant "see" the other. Remember IP addresses are for us dumb humans and computers really talk via a MAC (Media Access Control) address, the "hardware" address of each LAN card. That is why when they both use the router, they work as the host C address is on a different subnet and they both use the router's MAC address as the path. The same situation will occur if one host is swapped out for another but using the same IP address as the original.

If you suspect that is what is happening, you can try changing the address of one or both of the hosts to a previously unused address and note the results. But only change one address at a time.

[U.SivaKumar_2]

Restart the Switch , some ARP cache issue

[Ryan Gu]

And another thing:
I found that hostA & hostB can communit for a while,but not at most time.

I have change tcp6 to tcp in /etc/rc.config.d/netconf.

Before I install 11i on hostA,everything is ok(hostA is running 11.0).

[Ron Kinner]

Perhaps a VLAN misconfig on the switch? A dropped trunk between switches? Incorrect static arp on a switch. Perhaps you have reused a MAC address by cloning a box?

Look up linkloop in the man and run linkloop tests with the verbose option turned on first to the cisco and hostC so you know what to expect then to the one that doesn't work. See what the test says. That may give you a clue.

Since they are supposedly in the same subnet there should be no networking issue but it never hurts to double check that you have the masks correct and that its routing table is set up the way you expect:

netstat -rnv

Ping both hosts from the cisco. See if one or both of them returns Q's instead of !'s. That's a common problem on the 11.0's and means the host is trying to send a source quench because it has buffer problems. There is a patch for that but you can simply tell ndd not to send source quench since the cause of the buffer problem is known and trivial.

Ron

[Steven E. Protter]

It looks to me what happens when there is a typo in /etc/rc.config.d/netconf

I've seen smilar behavior when two machines are trtying to use the same IP address.

I'm assuming you have checked into basic issues, like sticking your nose behind the server and making sure the link light on the NIC cards is still lit, and the hub or switch shows green lights. Right?

Try this
traceroute Hosta
traceroute Hostb
traceroute Hostc

from another unix box, change the command to tracert if you have to unix a microsoft box.

I'd be interested in results and error messages and a look at the following files:

/etc/resolv.conf
/etc/nsswitch.conf
/etc/rc.config.d/netconf

On all three unix hosts.

There may be an inconsistency there.


P

[Ryan Gu]

Thanks for all reply.
I check all the device and configuration.It seems that the duplex configuration cause this problem.
I changed the port of cisco switch then everything is ok.

[Ryan Gu]

Thanks for all reply.
I check all the device and configuration.It seems that the duplex configuration cause this problem.
I changed the port dup of cisco switch then everything is ok.