Hazelcast IMDG

Supported Versions

Currently supported versions are Hazelcast 3.2 to 3.10.

Sensor (Data Collection)

Node-Level

Tracked Configuration

  • Node name¹
  • Node address
  • Node UUID
  • Group name
  • Cluster Id (starting with Hazelcast 3.5)
  • Version (starting with Hazelcast 3.2)
  • OpenSource / Enterprise version (starting with Hazelcast 3.2)
  • Lite Member Status (starting with Hazelcast 3.6)

Instana extracts the label for the Hazelcast instances by using the node address and the cluster name, e.g. [127.0.0.1:5701]@my-cluster-5c07d2.

Metrics

  • Operation Count (starting with Hazelcast 3.2)
  • MigrationQueue Size
  • EventQueue Size (starting with Hazelcast 3.2)
  • Distributed Objects (starting with Hazelcast 3.5)
  • ExecutionService Queue Size (starting with Hazelcast 3.2)
  • Connected Client Count (starting with Hazelcast 3.2)

Health Signatures

  • Local Member Status (starting with Hazelcast 3.3²)

Cluster-Level

Tracked Configuration

  • Cluster Id
  • Group name

Starting with Hazelcast 3.5 the cluster name is a combination of the configured group name and the first characters of the cluster Id, e.g. my-cluster-5c07d2. In older Hazelcast versions only the group name is used, e.g. my-cluster.

Metrics

  • Node count

Health Signatures

  • Cluster Status (starting with Hazelcast 3.3²)

¹ As returned by HazelcastInstance::getName().

² Starting with Hazelcast 3.3 the public methods HazelcastInstance::getPartitionService()::isLocalMemberSafe() is used. For older Hazelcast versions the health status is derived from an internal “has ongoing migrations” status on each local node.

The Hazelcast cluster health status is aggregated from each Hazelcast node. This is exactly what HazelcastInstance::getPartitionService()::isClusterSafe() does internally, but without creating additional overhead of calling this method.