Hazelcast: how data is partitioned when using Cluster Groups?

0
votes

In the Hazelcast docs it is stated about Cluster Groups:

You can create cluster groups. To do this, use the group configuration element.

By specifying a group name and group password, you can separate your clusters in a simple way. Example groupings can be by development, production, test, app, etc. <...> Each Hazelcast instance can only participate in one group. Each Hazelcast instance only joins to its own group and does not interact with other groups.

<...>

The cluster members (nodes) and clients having the same group configuration (i.e., the same group name and password) forms a private cluster.

Each cluster will have its own group and it will not interfere with other clusters.

But there are no details about data partitioning.

If I have 5 nodes and 2 cluster groups:

  • node1, node2 and node3 are members of GroupA
  • node4 and node5 are members of GroupB

does it means that no data from GroupA will be stored at nodes4 and node5?

1
hazelcast

1 Answers

3
votes

Yeah that's what it means. Those groups are independent clusters and have nothing in common (except maybe the network ;-)).

If you look for data partitioning, Hazelcast distributes information based on keys but you can have some kind of influence by utilizing data affinity (http://docs.hazelcast.org/docs/3.8/manual/html-single/index.html#data-affinity).

If you're looking for backup distribution you might be interested in partition groups (http://docs.hazelcast.org/docs/3.8/manual/html-single/index.html#partition-group-configuration).