Master HPE Morpheus Software HA deployments with our deep dive into three-node and full distributed architectures for zero-downtime multicloud orchestration.

In the world of enterprise cloud orchestration, availability is not just a feature—it is a requirement. As organizations rely on HPE Morpheus Software to manage multicloud environments, the underlying appliance must be resilient enough to handle hardware failures and software updates without interrupting service. This technical guide explores the full high availability (HA) architecture, best practices, and the recommended sequence for a successful resilient deployment.

By following this structured approach, your HPE Morpheus HA deployment provides the high-performance, resilient orchestration engine required for modern digital business.

Figure 1. Resilient architecture to deploy HPE Morpheus

Reference architecture for a resilient deployment of HPE Morpheus

HPE Morpheus is composed of four primary service tiers that must be coordinated for a successful HA installation:

1. Application tier: Runs stateless services like NGNIX and Tomcat
2. Transactional database tier: A MySQL-compatible logistical datastore
3. Non-transactional database tier: An OpenSearch cluster used for log aggregation, stats, and metrics
4. Messaging tier: A RabbitMQ cluster using Advanced Message Queuing Protocol (AMQP) and Streaming Text Oriented Messaging Protocol (STOMP) for internal communication and agent connectivity

While a standard three-node HA architecture (clustering OpenSearch and RabbitMQ on the app nodes while externalizing MySQL) is common, a full distributed HA deployment goes a step further. In a full HA scenario, every tier is hosted on its own specialized cluster or supported platform-as-a-service (PaaS) offering, such as Amazon Aurora for the database or Amazon MQ for messaging. This provides greater horizontal scalability and allows teams to offload the management of individual components to cloud providers.

Figure 2. Factors to consider when planning and deploying an HA architecture for HPE Morpheus

Pros and cons: Is full HA right for you?

Deploying a distributed HA architecture is a strategic decision that balances complexity against reliability.

Advantages

• Zero downtime upgrades: Software updates can be performed as rolling upgrades, keeping the UI available throughout the process.
• No single point of failure: Every component is redundant, allowing the system to tolerate the loss of individual nodes.
• Scalability: Each tier can be scaled independently, either vertically by adding resources or horizontally by adding nodes.
• PaaS integration: Leverage managed services like AWS RDS or Azure Storage Accounts to reduce administrative overhead.

Disadvantages

• Increased complexity: Requires more specialized knowledge for external database and messaging cluster maintenance.
• Infrastructure overhead: Higher initial resource requirements and networking complexity compared to all-in-one (AIO) setups.
• Extended setup time: The orchestration of multiple clusters and shared storage increases the initial installation timeline.

Core infrastructure prerequisites

Before beginning the deployment, the following infrastructure must be ready:

• External database: A MySQL v8.0 cluster (minimum v8.0.72) with at least three nodes for redundancy. It must have network connectivity to the application nodes on port 3306.
• Shared storage: A high-performance NFS share or object storage (Amazon S3/OpenStack Swift) is mandatory. This must be mounted at /var/opt/morpheus/morpheus-ui on all application nodes to store virtual images, plug-ins, and deployment archives.
• Load balancer: A front-end load balancer is required to distribute traffic on port 443. It must support WebSockets for agent communication and should ideally be configured for SSL bridging.
• Networking: All nodes must resolve each other's short names (not FQDNs) through DNS or /etc/hosts. Specific ports (e.g., 9200/9300 for OpenSearch and 5672/15672 for RabbitMQ) must be open for intercluster communication.

Recommended design and deploy sequence

To help ensure a smooth installation, follow this step-by-step deployment sequence:

1. Prepare the external tiers

Start by setting up your external clusters (MySQL, OpenSearch, and RabbitMQ if not using embedded versions). Log in to your MySQL cluster to create the morpheus database and a dedicated user with full privileges and system-level permissions like SELECT, PROCESS, and RELOAD.

2. Software installation

Install the HPE Morpheus appliance package on all target application nodes using rpm or dpkg (depending on the actual distro in use). Important: Do not run the reconfigure command yet.

3. Critical configuration: The morpheus.rb file

Edit /etc/morpheus/morpheus.rb on every node. Consistency is vital here. You must manually define identical UIDs and GIDs for service users (e.g., user, opensearch, rabbitmq) across all nodes. Failure to do so will result in permission sync errors on the shared storage. Ensure mysql['enable'] is set to false and point the configuration to your external database endpoints.

4. Mount shared storage

Ensure the NFS or shared volume is correctly mounted at the required path on every node before proceeding.

5. Initial reconfigure and clustering

Run sudo morpheus-ctl reconfigure on all nodes. OpenSearch is designed to auto-cluster during this process once nodes can reach each other on the transport ports.

6. Manual RabbitMQ finalization

Unlike OpenSearch, the embedded RabbitMQ service requires manual clustering steps after the initial reconfigure. This involves stopping the service on secondary nodes, syncing the .erlang.cookie and secrets from the source of truth node, and using the rabbitmqctl join_cluster command [Conversation History].

Operational best practices

• Rolling upgrades: Always upgrade and reconfigure one node at a time to maintain availability. Verify the UI is back online before moving to the next node.
• Health monitoring: Configure the load balancer to use an HTTPS health monitor pointing to /ping. A healthy node will respond with a 200 status and the body MORPHEUS PING.
• Storage management: Monitor partition space closely; if the OpenSearch partition reaches 85% capacity, the cluster enters an unhealthy state and ceases to function properly.
• Scaling: When adding new application nodes, copy the morpheus.rb from an existing node to help ensure consistent cluster settings, and then add the new node to the load balancer pool after reconfiguration.

Requirements at a glance

Requirement category	Feature	Method	Port	Source/destination	Details
HA specific storage	Shared storage (NFS)	NFS mount		/var/opt/morpheus/morpheus-ui	Required for HA installations to store virtual images, plug-ins, and shared files across all nodes. 50 GB minimum recommended.
Network port	OpenSearch cluster	Transport	9300	node to node	Required for internode communication and cluster formation in HA.
Hardware	Storage				200 GB minimum; 400 GB recommended per node for three-node HA deployments.
Hardware	CPU				Recommended 4-core, 1.4 GHz (or better), 64-bit CPU per node for HA/distributed installations.
Hardware	Memory (RAM)				16 GB recommended per node; 8 GB minimum required.
Operating system	OS support				Amazon Linux 2, Debian 11/12, RHEL 8.x/9.x, OEL 8.x/9.x, SLES 15, Ubuntu 20.04/22.04/24.04, Rocky 8/9.
Storage path (default)	Datastore			/var/opt/morpheus	User, application, and services data (OpenSearch, RabbitMQ, DB data).
Storage Path (default)	App files			/opt/morpheus	HPE Morpheus application and services files.
Network port	Agent communication	HTTPS/WSS	443	Node to appliance	Outbound connection from managed node to appliance URL.
Network port	Hypervisor console	HTTPS	443	Appliance to hypervisor	Host names must be resolvable by HPE Morpheus appliance.
Network port	RabbitMQ cluster	AMQP	5672	Appliance to tier	Non-TLS messaging connection.
Network port	Agent installation	YUM/APT	80	Node to appliance	Used for appliance yum and apt repos (specifically older Ubuntu 14.04).
Network port	SSH access	SSH	22	Appliance to node	Used for Linux agent installation and remote console.
Network port	WinRM access	WinRM	5985	Appliance to node	Used for Windows agent installation.

Summary

By implementing a robust HA architecture, organizations can transform their HPE Morpheus Software deployment from a single management point into a resilient, enterprise-grade orchestration engine. Whether opting for the recommended three-node HA or a fully distributed model, the primary goal remains to mitigate single points of failure and facilitate zero-downtime operations through rolling upgrades. Adhering to the prescribed setup sequence—from externalizing the transactional database to manual messaging tier clustering—helps ensure that the core infrastructure remains stable and consistent under load. Ultimately, this investment in redundant tiers and high-performance shared storage provides the foundation necessary for modern digital businesses to manage complex multicloud environments with absolute confidence.

For more information, please visit hpe.com/us/en/software/cloud-ops-suite.html

Meet the author:

Fernando Escobar, Sr. Technical Product Manager, HPE Morpheus Enterprise, HPE