# Mule runtime clusters versus server groups

*MuleSoft Platform Integration Architect Study Guide · Designing for the runtime plane technology architecture · Last updated November 13, 2026 · https://www.certifyforce.com/study-guide/mulesoft-platform-integration-architect/designing-for-the-runtime-plane-technology-architecture/mule-runtime-clusters-versus-server-groups*

How an on-premises Mule 4 cluster shares state through its memory grid, runs primary-only sources and hands work between nodes, and how that differs from a server group. Also covers store profiles, node prerequisites and the two ways to manage a cluster.

- Certification: Salesforce MuleSoft Platform Integration Architect
- Exam section: Designing for the runtime plane technology architecture
- Study guide: https://www.certifyforce.com/study-guide/mulesoft-platform-integration-architect

**Official exam objective:**

> Analyze the mode of operation of a Mule runtime cluster that differentiates it from other deployment options.

## What the exam expects

- Explain the active-active mode of an on-premises Mule runtime cluster and what the primary node does differently.
- Distinguish a cluster, which shares state, from a server group, whose application instances run in isolation.
- Design flows so later processing stages can move to another node, and recognize where cluster coordination stops.
- Choose between the performance and reliable store profiles for a container or a single application.
- Prepare nodes for a cluster: version alignment, discovery settings, ports and a single management method.

**Cluster:** Mule runtime engines acting as one virtual server over a shared memory grid. **Server group:** servers used as one deployment target whose instances run in isolation. **Primary node:** the node that runs schedulers and primary-only sources. **Store profile:** the setting that decides whether queues and object stores are distributed and backed up ([HA cluster overview](https://docs.mulesoft.com/on-premises/4.6/mule-high-availability-ha-clusters)).

## Key facts

| Fact | Detail | Source |
|---|---|---|
| Cluster model | Active-active; all nodes process messages | [HA cluster overview](https://docs.mulesoft.com/on-premises/4.6/mule-high-availability-ha-clusters) |
| Grid contents | Object stores, VM queues and the LockFactory are replicated | [HA cluster overview](https://docs.mulesoft.com/on-premises/4.6/mule-high-availability-ha-clusters) |
| Server group | Application instances run in isolation | [Server Groups](https://docs.mulesoft.com/runtime-manager/server-group-about) |
| Scheduler | Runs only on the primary polling node | [HA cluster overview](https://docs.mulesoft.com/on-premises/4.6/mule-high-availability-ha-clusters) |
| Flow Reference | Runs on the node where the event began | [HA cluster overview](https://docs.mulesoft.com/on-premises/4.6/mule-high-availability-ha-clusters) |
| Resource locking | Applies only to nodes inside the cluster | [HA cluster overview](https://docs.mulesoft.com/on-premises/4.6/mule-high-availability-ha-clusters) |
| Profile precedence | Application-level store profile overrides the container | [HA cluster overview](https://docs.mulesoft.com/on-premises/4.6/mule-high-availability-ha-clusters) |
| Node versions | Same Mule (and cumulative patch), agent and Java | [Manual cluster](https://docs.mulesoft.com/on-premises/4.6/creating-and-managing-a-cluster-manually) |
| Unicast file setting | `mule.cluster.nodes` plus `mule.cluster.multicastenabled=false` | [Manual cluster](https://docs.mulesoft.com/on-premises/4.6/creating-and-managing-a-cluster-manually) |
| Management | Manual files are not synced; platform changes override them | [Manual cluster](https://docs.mulesoft.com/on-premises/4.6/creating-and-managing-a-cluster-manually) |
| Size | Up to eight servers | [Clusters](https://docs.mulesoft.com/runtime-manager/cluster-about) |
| Default ports | TCP 5701, 5702, 5703 | [HA cluster overview](https://docs.mulesoft.com/on-premises/4.6/mule-high-availability-ha-clusters) |

## Check yourself

- Why can't the servers in a unicast Runtime Manager cluster take DHCP-assigned addresses?
- When a cluster is configured through Runtime Manager on its default ports, which TCP ports must stay open between nodes?
- Which discovery type lets a server be configured as a node before it is Running?
- Why does an on-premises cluster still need a load balancer or client-side failover for HTTP traffic?

## Sources

- [On-Premises Mule Runtime High Availability Cluster Overview](https://docs.mulesoft.com/on-premises/4.6/mule-high-availability-ha-clusters)
- [Create and Manage an On-Premises Mule Runtime Cluster Manually](https://docs.mulesoft.com/on-premises/4.6/creating-and-managing-a-cluster-manually)
- [Server Groups](https://docs.mulesoft.com/runtime-manager/server-group-about)
- [Clusters](https://docs.mulesoft.com/runtime-manager/cluster-about)
- [Deployment Options](https://docs.mulesoft.com/runtime-manager/deployment-strategies)

## About the full unit

The "Core concepts", "Exam traps" and "Worked scenario" sections of this unit, its 9 linked practice questions, and unlimited timed mock exams are included with the MuleSoft Platform Integration Architect pack (USD $19): https://www.certifyforce.com/practice-exams