network problem

[Muhammad Shamroze]

Im using Redhat ES 3.0 Linux and on the follwing OS two LAN configured one eht0 and other is eth1 and working fine also ping other machines but after some hours when u use to ping other m/c ip it say host unreachable but ping itself and when i use to type "service network restart" it again resolve the issue but again in after some hours of time it agian do the same .....when i see the /var/logs/messages i found that it is showing the following error that is:

Kernel:Flow control is off for TX and off for RX

Kindly help me out is this regards that what should be done so that the other machines can also ping the following server and also operate through telnet or ssh

Regards

[Matti_Kurkela]

This message is not an error.

It is a status message from the network card driver, reporting the state of flow control options.

What type of network cards are you using?
When the host is unreachable, try to use "ethtool" or "mii-tool" on that host to check the state of the network interfaces. Is the card detecting a link while the problem is occurring?

Also check the statistics reported by "ifconfig": are there significant numbers of errors?

Pinging the host itself will only tell that the TCP/IP networking subsystem still exists on the host: it will cause *no* traffic on the network, because the TCP/IP subsystem routes any traffic addressed to the hosts own interfaces through a shortcut, bypassing the network cards altogether.

Pinging another host on the same network will produce actual traffic on the wire, and may trigger ARP queries to be sent if necessary. If a switch is thinking your host is in a different port than it actually is, ARP queries offer it a chance of detecting its error. But if a switch repeatedly gets wrong ideas like this, there may be another host with the same MAC address on the same segment.

Another possible reason: if you are using DHCP and have iptables filtering configured, your filter may be so tight it blocks DHCP address renewal messages. When the host doesn't yet have an IP address, it must send DHCP queries as broadcasts; but when it has got a lease on an IP address and tries to renew it to continue operation, the renewal messages are sent as non-broadcast to the DHCP server.

The "service network restart" resets the DHCP mechanism back to the initial state: if DHCP broadcasts are allowed but non-broadcasts are not, the host can initially get a lease but cannot renew it. So this might also produce a pattern like you're seeing.

[timmy2006]

1. If its available in your conditions - setup your card on 10Mb Half Duplex.
1.1. Do the same on the other side (switch\router).
1.2. Check the behaviour of link.
3. Use another port in router\switch your machine is connected to (its better to use another group of ports - usually there are 4 ports serviced by 1 chip on the circuit, so for example if you are using port N3 - connect cable into port N5.
4. Use another cable. Sometimes visually they are looks fine but while testing with LANtester shows problems.
5. Use tcpdump -ni eth0/1 while pinging other hosts. Try other packet size. try to telnet to some service.
But in short i guess you need to replace your NICs or switch\router. (from my personal expirience)