Is it bad practice to use a default name (e.g. “my-product-name”) for a Hazelcast group if using a non-multicast member discovery method?
My worries are founded on the following scenario… If you have multiple Hazelcast groups on the same network but with different members what does this mean? Will they function correctly as separate groups or will there be problems such as a message being sent to a member of the wrong group?
For example if you had one group named “dev” that contained nodes A and B discovered through TCP/IP configuration (i.e. the members are explicitly defined) running on a laptop and someone starts a new group also named “dev” on a different PC on the same network and has explicitly defined members C and D – will these clusters/groups function perfectly well in isolation of each other?
Furthermore if the first group explicitly defined its members but the second group used multicast discovery would that change anything? Would the multicast group accidentally include members from the other group?
Even though it would be better to use different group names for separate clusters, creating two TCP-IP clusters with explicitly defined members or one TCP-IP and one multicast cluster also should work without any issue.
In explicitly defined TCP-IP join, Hazelcast only opens connections to defined members.
In multicast join, Hazelcast opens a multicast socket and listens to multicast packets. Because TCP-IP cluster won’t listen to multicast packets, creating one TCP-Ip and one multicast cluster will be safe.
Even you can create multiple multicast clusters using different multicast groups.