Primary VS Standby LANs

[Tom Jameson]

After reading through"Managing MC/ServiceGuard" , I'm confused as to whether the Primary and Standby LAN segments are differing subnets and how the relocatable addresses could swap to the Standby if it were a different subnet. Can someone offer any clarification?

Thanks in advance.

[Michael Elleby III_1]

Tom-

The way I understand it, the Primary and Standby LAN segments should be differing subnets based on a failure on the lan where the Primary Lan segment exists.

If a router goes down, or the network (subnet) becomes unavailable for any reason, then the Standby lan which is on a different subnet would then take command... In this way, you eliminate single point of failure of the network in your SG cluster..

It's funny though, since most of the environments that I have worked in have maintained both their primary and secondary lan on the same subnet, with only using a differing subnet or private lan for the heartbeat functionality..

Just my $.02 worth..

Mike-

[Christopher McCray_1 · Accepted solution]

Hello,

To have a standby LAN, it MUST be on the same subnet as the primary you wish to back up.

When you first issue the cmquerycl command for the cluster, the possible standby LANs will be identified and for which primary it will backup. If you do not have it on the same subnet, the output will display this in your cluster.ascii file and issue a warning for this.

Hope this helps

Chris

[Tom Jameson]

Many thanks for the swift replies!

Now I can go forward with my data center setup.

[Misa]

Same subnet, BUT it's better if the primary and secondary are on different network gear. Fewer single points of failure that way. I think that's the point they're trying to make in that illustration, which I agree is not the clearest.

For example, say that you'll be using lan0 and lan1 on each of your two cluster hosts as the primary and secondary.

Plug lan0 on each host into core router A, and plug lan1 on each host into core router B. That way, when someone accidentally yanks the power on core A (not that it would ever happen! ;), your networking isn't all knocked out and can fail to core B.

By the way, if you haven't taken the class, this is IMO a worthwhile one. Having play hardware seemed to crystallize the theoretical.

--M????a