Redundancy on Ethernet Ports

[Nicholas Noor_1]

We are seeking to use all 4 ports of the Blader Server. The first two ports on either side connect via datalink layer, but there is only 1 port connecting via network layer.

We have this connected via single cisco switch, but if this switch fails we have no connectivity to the server.

If we connect to another cisco switch using the 4th port, how should we handle redundancy while maintaining connectivity? Do we use a replicator of just connect to the 2nd cisco switch?

Now a single IP address for eth0 that is used will the switch know the connectivity that this NIC port failed and that it takes over the eth1 and the IP address doesn't change?

Eth0 and Eth0 cannot be bound with a single IP I supppose, what is the solution that the same IP remains or is there another way of doing this?
Please mail me.
Thanks

[rick jones]

_Which_ blade server, and under which OS? Putting links together for HA involves what is variously called trunking, teaming, bonding, aggregation depending on the OS and context. When multiple switches are involved, _generally_ the trunk/aggregate is active/passive rather than active/active.

When you say "cisco switch" is that the "3020" switch module in a blade chassis, or a "standalone" Cisco switch outside the chassis?

When trunked/bonded/aggregated, the two (or more) physical NICs look like a single interface to the rest of the stack.

The stone knives and bearskins way to do things would be to have something continuously checking the health of the primary NIC, and upon sensing it is down, "manually" (in the script) migrate the IP from the primary to the backup NIC. That is in essense what the trunking/teaming/aggregation software would be doing.

Switches in and of themselves care not about IP addresses. They only care about MAC (Media Access Control, aka "Ethernet") addresses. Other systems may care about the MAC address changing under the IP address - they may or may not update their ARP caches in a timely fashion. For this reason some if not all trunking/teaming/aggregatation solutions cause the MAC address to remain the same through the switchover.

[Nicholas Noor_1]

This is a Blp40 server (RH linux) which has 4 Ethernet ports. The 2 are now connected via a 2950 Cisco switch. Another Cisco 2950 will connect to the other 2 ethernet ports.

If one of the switches dies, the other should still be able to function. Now, my network admin says that we want to have one IP, I understand the ARP cache will not update immediately.

But with teaming/bonding the software should do it. I also found modprobe.conf on a linux system does not exist, maybe drivers have to be downloaded?

We just want a remote user not to lose connection using the systems we have. How do we go about in a simple way?

Thanks for all replies to this question.