Crossover cable for Heartbeat network

[Ra Jose]

Hello freinds,

I have 2 servers rx2620 that are on 11iv2 and Sg11.16. Each server has 3 lan interfaces.
All 3 interfaces are 1000Base-T. One lan interface is primary+heartbeat, second interface is for failover. Both these interfaces are connected the switch.

We have 3rd interface which is dedicated heartbeat and crossover cable runs between the 2 servers. Somehow, I see the ip/route config ok on both servers. But I cannot ping one another. Here is how config looks like -

node-1# ifconfig lan2
lan2: flags=1843
inet 192.168.1.4 netmask ffffff00 broadcast 192.168.1.255

node-1# netstat -f inet -rn (o/p partially shown)
IPv4 Routing tables:
Destination Gateway Flags Refs Interface Pmtu
192.168.1.4 192.168.1.4 UH 0 lan2 4136
192.168.1.0 192.168.1.4 U 2 lan2 1500


node-2# ifconfig lan2
lan2: flags=1843
inet 192.168.1.3 netmask ffffff00 broadcast 192.168.1.255

node-2# netstat -f inet -rn
IPv4 Routing tables:
Destination Gateway Flags Refs Interface Pmtu
192.168.1.3 192.168.1.3 UH 0 lan2 4136
192.168.1.0 192.168.1.3 U 2 lan2 1500

However, I cannot ping node-2 from node-1 and vice-versa.

Is there anything wrong I am doing in configuration. Or is it the cable could be bad?

Your help is appreciated

[John Poff]

At a glance, your network configuration looks ok. I would suspect your crossover cable. I've used crossover cables before for heartbeat LANs in a pinch, but I much prefer to buy a cheap hub and use regular cables. It makes it easier to manage.

I have a three node cluster here with a cheap hub, and I was able to troubleshoot a heartbeat problem earlier this week where one of the cables wasn't plugged in completely to one of the LAN ports on the server. No green light on the hub for that node, so I knew where my problem likely was.

JP

[Geoff Wild · Accepted solution]

Officialy, cross over cables are not supported in ServiceGuard - though technically you can use them.

Try doing a linkloop test.

linkloop -i 2 0xmacdest


the -i points to the ppa number of lan card to go out.

Enter the destination mac address.

If that fails, then you have bad cable.

Rgds...Geoff

[Ra Jose]

Thank you John Geoff. I ran Geoff's tests and
it succeeded. Looks like the cable may be bad. I am going to replace this cable and see if tests succeed. I will back after this test.

[Ra Jose]

Thank you Geoff for your expert insight.

It was the bad cable. Replaced the cable and it worked fine. Thank you again.

Rajose.

[Ra Jose]

Problem solved.

[rick jones]

BTW, the 1000Base-T spec includes AutoMDIX, which means that to connect two 1000Base-T stations back-to-back one does not actually need a cross-over cable. A "regular" cable will work just as well.

[John Bigg]

I know this thread is closed but I wanted to correct an earlier statement. It is untrue to say crossover cables are unsupported.

I would not recommend them, I would always suggest a small hub/switch over a crossover cable, but they are supported.

The reason for not recommending them is that if you power off one node, the other node sees a cable disconnect event when a crossover cable is used.

[Ra Jose]

Thank you Rick and John. I was under the impression that we should use only cross-over cables when connected back-to-back. Rick's research indicates that we can use regular cables as well for 1000base-T. I should try this and I will when I get the next opportunity. Thanks folks.

Rajose