Ella Core’s mission is to provide a simple, reliable, and secure 5G core for private networks. From day one, we knew Ella Core needed a failover story for critical deployment — but not at the cost of high complexity. Today we’re shipping it. You can deploy Ella Core v1.10 as a highly available cluster.
Each node is active and can accept 5G radios and subscriber traffic.
Raft as the consensus algorithm for fault tolerance
High Availability (HA) in Ella Core is designed around the Raft Consensus Algorithm: at any time one node is the leader, it is the only node that accepts writes, and every write replicates to a majority of nodes before it is considered committed. Nodes communicate together via mTLS to share changes.
We chose Raft over external alternatives to minimize HA-related complexity for our users. Ella Core remains a single self-contained binary: no extra services to deploy, monitor, or upgrade.
Deploy three or five nodes. Three nodes tolerate one failure; five nodes tolerate two.
All persistent resources are replicated across the cluster, so if a node dies, the others have the same subscribers, policies, and operator configuration. The cluster automatically elects a new leader and keeps accepting operator changes with no manual intervention.
Runtime state tied to a specific connection or session does not replicate. This includes 5G radio associations, subscriber sessions and their User Plane state, GTP-U tunnels, and active BGP adjacencies. If a node dies, the UE must reconnect. Observability is also per-node.
Deployment Scenarios
The HA cluster is the same regardless of how radios connect to it; the radio side determines how much HA reaches individual UEs.
Radios Connected to Every Node
When a Core dies, the radio reselects within the cluster automatically; affected UEs re-register on a surviving node without operator action.
Radios Pinned to Specific Nodes
Useful for partitioned deployments. The cluster still replicates operator state across all nodes, so changes made anywhere are visible everywhere — but if a Core dies, its paired radios and subscribers lose connectivity.
How to deploy a Highly Available cluster
Deploy a HA cluster in 5 steps:
- Install Ella Core v1.10 on three hosts
- Start Ella Core on node 1 with a cluster definition in its configuration file
- Open the Cluster page on node 1, click Add Node and copy the token.
- Start Ella Core on node 2 with a cluster definition in its configuration file with the token.
- Repeat steps 3-4 for node 3
View the complete guide here.
Conclusion
High Availability is the most important feature we added to Ella Core since we launched. It’s a significant improvement that enables critical deployments in the manufacturing, transport, and military domains. While we did everything we could to keep it minimal, running a HA cluster requires more operational effort than a standalone setup. There’s no way around it. Performing a rolling upgrade on a HA cluster takes more time than upgrading a single-node deployment, but it involves no downtime. It’s a trade-off and depending on your deployment you may choose one or the other.