topic Re: ServiceGuard Configuration in Operating System - HP-UX

ServiceGuard Configuration

Raj_112345 — Sun, 13 Apr 2008 19:12:32 GMT

Hi,

While setting up ServiceGuard for Clustering someone said both the nodes should be on the same subnet.

Wouldn't this be an issue when this subnet experience network glitch and both the nodes in the cluster become unavailable as they exist in the same subnet ?

Thanks,
Raj

Re: ServiceGuard Configuration

Svetoslav Gyurov — Sun, 13 Apr 2008 20:58:19 GMT

Hi,

For heartbeat it is mandatory have both nodes on the same subnet. Not sure for the stationary address although it doesn't make sense to be at different subnets.

You say it's an issue, but how can you say which one of the two nodes has failed in case of heartbeat subnet failure when you have only two nodes ? Well in most cases the cluster will arbitrate and thats why people use APA or put standby LANs to avoid failure of the heartbeat subnet.

--
sve

Re: ServiceGuard Configuration

Duncan Edmonstone — Mon, 14 Apr 2008 06:03:47 GMT

Raj,

Well genrally people operate a highly available network so this isn't a problem - don't interperet 'on the same subnet' as meaning 'attached to the same switch', as that definitely isn't the case.

Anyway, with a fully patched copy of the latest version of Serviceguard (11.18), the nodes can be in different subnets.

HTH

Duncan

Re: ServiceGuard Configuration

Sajjad Sahir — Sat, 19 Apr 2008 08:33:15 GMT

Dear Raj

it is not compulsory to belongs all nodes on the same subnet

thanks and regards
Sajjad

Re: ServiceGuard Configuration

Aashique — Sat, 19 Apr 2008 08:44:59 GMT

Hi,
Useful link for service guard

http://docs.hp.com/en/ha.html#Serviceguard

Thanks & Regards

Aashique

Re: ServiceGuard Configuration

Jeeshan — Tue, 22 Apr 2008 05:17:57 GMT

Hi Raj

The speed of cluster re-formation is partially dependent on the type of
hearbeat network that is used. Ethernet results in a slower failover time
than the other types. If two or more heartbeat subnets are used, the one
with the fastest failover time is used.

May be this can help tou for understanding.