Hazelcast Store

The HazelcastStore is a distributed implementation of the Store interface backed by Hazelcast’s IMap. Each session receives its own dedicated distributed map keyed by its sessionId, enabling session data to be shared across nodes in a Hazelcast cluster. This makes the store suitable for horizontally scaled deployments, where per-session metadata must remain consistent across multiple JVM instances.

Key characteristics

• Distributed per-session storage — session data is stored in Hazelcast’s partitioned, replicated map

• Cluster-wide visibility — values can be accessed and modified from any node in the Hazelcast cluster

• Resilient to node failure — depends on Hazelcast configuration for partition redundancy

• No external broker required — ideal when Hazelcast is already used for clustering or data grids

• Composable with any EventStore — event distribution backend may differ (e.g. Kafka, Redis Streams, etc.)

How it works

• set(key, val) updates the distributed map entry for this session

• get(key) retrieves the value from the cluster

• has(key) checks key existence

• del(key) removes a session key from distributed storage

• destroy() removes the entire IMap for the session, releasing all associated data

Advantages

👍 Transparent data sharing across the cluster 👍 Works naturally with Hazelcast-backed adapters or distributed state systems 👍 Supports hybrid deployments when paired with external EventStores 👍 Provides higher durability and resiliency compared to pure in-memory storage (e.g. HazelcastStore + KafkaEventStore, HazelcastStore + RedisStreamEventStore)

---

Summary

HazelcastStore provides distributed, session-scoped storage backed by Hazelcast IMap, making per-session metadata visible and resilient across nodes in a clustered deployment.