heartbeat network

[alishokry]

Hello ,

I have a service guard cluster version A.11.18 configured on two nodes and I configured a three lans
on node#1 : Lan0 heartbeat Ip:10.0.0.1 netmask 255.0.0.0
Lan1 stationary IP :10.11.20.1 netmask 255.255.255.0
Lan2 is standby
On node#2 : Lan0 heartbeat Ip:10.0.0.2 netmask 255.0.0.0
Lan1 stationary IP :10.11.20.2 netmask 255.255.255.0
And I used a cross cable link between two heartbeat.
The cluster/package is working fine.
The question is a client can not access the nodes of the cluster.

Please advise
Thanks

[Duncan Edmonstone]

This looks like an invalid subnet configuration to me... if you're HB subnet lan0 has IP of 10.0.0.1 and a subnet of 255.0.0.0, then that means _all_ IP addresses in the 10.x range are out the lan0 card (which clearly they aren't as you have a crossover cable only connected to the other system).

Stop your cluster, set your subnet mask for the HB network to 255.255.255.0 on both nodes - and then restart your cluster and everything should start working OK...

HTH

Duncan

[alishokry]

Dear Duncan ,

Thanks for your support ,

But I have a question , the two heartbeat are
connected through a cross over cable , and they appear isolated from other network,

why they affecting on the network ?

Please , can you explain in details.

Thank you very much

[Duncan Edmonstone]

You need a lesson in IP subnetting...

an IP of 10.0.0.1 with a subnet mask of 255.0.0.0 means that the subnet is any IP in the 10.x range... so that's every IP from 10.0.0.1 to 10.254.254.254.

When the routing table looks at where to send your packets it will look at that subnet and correctly think that _any_ packet bound for a 10.x address should be sent out of lan0 - that includes the packets that _you_ think should be going out of the lan1 NIC.

So your current network configuration is "illegal" - this is nothing to do with HP-UX or Serviceguard, its a basic IP thing.

As lan0 is a private HB network and in fact is just a crossovewr cable, you could just as well change the IPs completely to some other private network space (like 192.x.x.x) as well and that would also fix the problem.

HTH

Duncan

[Michael Steele_2]

Hi

So find the default gateway using 'netstat -rn' and see if both your client and the server can ping the default gateway. Then get back to me.