Cross-over cable for heartbeat?

[Igor Streharski]

Hello,

I wonder whether it is possible to have heartbeat signaling over a cross-over UTP cable connecting LAN cards (D-class, core i/o 10Mbps LANICs). If yes, is such a configuration supported?

[Leila Maria Rebel]

Hello, Igor !

Yes, you can use your heartbeat like that ! I have exactly this configuration in my site. I have 2 D-Class with a cross cable as a heartbeat network.

If you need more details, you can ask me.

[melvyn burnard · Accepted solution]

Yes this is possible, and MC/SG supports it, but I have seen many problems caused by this, when a small 4 port 10BaseT hub costs very little.
Simply put, if one of the nodes is down for any reason, no signals will go out of the port on the down node, thus the other node assumes a network cable/card failure has occurred on it's OWN network side. In effect, an unterminated connection. Depending on the configuration, patching level and circumstances, I have seen strange symptoms, form minor irritations in the syslog, to the 2nd node performing a TOC.
Save the headaches, buy a cheap hub.

[John Waller]

Igor,
As Leila Maria said it does work if you use a network crossover cable and until recently I believed this was HP's recommended solution. I have said "until recently" because I had a problem with 2 N class servers in a ServiceGuard cluster and the person at the reponse centre said that HP recommend using a stand alone mini hub. The reason, is that if you reboot one of the servers the running server detects a LAN Disaster and this appeared to cause a problem with swagentd. By connecting 2 seperate standard network cables to a standalone 4 port mini hub, when you reboot one server the running server will not detect a loss of voltage (signal) so will not report a LAN disaster. I don't know about where you come from but in the UK you can buy these hubs very cheap so the minimal cost spend now may mean less problems in the future.

[Ovidiu D. Raita]

If you chose the crossover or the hub solution you still have a Single Point of Failure.

Ideally you should have a dedicated heartbeat network using a hub AND at least one of the other data networks set up as a heartbeat network.

It is recommended also that you increase the NODE_TIMEOUT to at least 10 secs. I use as a rule of thumb 20 secs. This way you just delay 10 more secs the cluster reformation.

HEARTBEAT_INTERVAL = 1000000
NODE_TIMEOUT = 20000000

Ovidiu

[Alexander M. Ermes]

Hi there.
We have a cluster of two N4000 running.
We have decided to go for the crossover cable
as the main heartbeat connection and the data
network connection as a backup line for the heartbeat. You then should set the parameters as one of my precedessors has said.
Rgds
Alexander M. Ermes

[Dan Hetzel]

Hi Igor,

As Melvyn and John said, with a cross-over cable, if you reboot one server the other running server will detect the signal loss and assume a network/card failure.
It's really worth adding a mini (4 ports) hub as they are really inexpensive.
That's how we configured heartbeat network here and it's running perfectly.

Best regards,

Dan