Intermittent Discovery of servers in a cluster.

[pczeus]

We are running embedded HorneqQ as part of a Spring Boot application. Our environments typically have 4-6 servers.

We consistently experience varying behaviors where in an environment, servers will form multiple clusters and not all join the same cluster.

I have seen where all servers do join the same cluster, and I am convinced they are 'capable' of discovering each other, yet it is very inconsistent.

Can anyone give me some suggestions on which settings can/should be adjusted to make the discover/broadcast more 'tolerant' and consistent?

[jbertram]

If the nodes do form a cluster as expected on occasion then my first thought is that the problem is environmental.  You could monitor the UDP traffic on your network to see if messages aren't being delivered as expected. 

The normal discovery mechanism uses UDP multicast which is unreliable by its nature (unlike, e.g., TCP).  You could switch to static clustering (which uses TCP), but that would add additional configuration overhead.