Hi, Imiao,

this "Clusters up to 32 nodes are supported" is only the maximum number of nodes you can configure in the High Availability Engine. It depends on the hardware and your network bandwith, up to which number of nodes your LLT is stable.

In my projects sometimes I was forced to use more Links between the nodes, in one installation we had 5 interlink 1 Gbit connections between the nodes.

Yes, the more nodes you have in the cluster, the more unstable the cluster becomes, but only, when the bandwith is too small. Imagine the huge amount of packages, when each node is trying to update all other nodes with the newest ressource stati. And the other thing is the total number of ressources in the cluster. Because all nodes have to be up to date with the informations and stati of all ressources on all nodes these two things, nodes and ressources, are the killer parameters for the interlink.

In most of the cases you can change the landscape of your applications to configure more than one but smaller clusters. And for those smaller things you have better control. You have to find a balance between costs, administration effort and stability.

But think about bandwith. Helps in most cases.

And think about updating the cluster version. Also LLT works better in version 5.

You can also change the parameters for timeouts aus hangups in LLT and LLT/GAB/had handshaking as well, but pls do this only if all other fails. It makes the cluster slower in detecting and reacting to crashes and node failures.

Sry that I have no better messages