Can customers deploy active-active clusters?
Yes, as of 2026, you can absolutely deploy active-active clusters on OCI. Unlike a basic failover setup (where one server sits idle), an active-active configuration allows all "rented" servers in the cluster to handle traffic simultaneously, maximizing your ROI and providing near-zero downtime.
Here is how active-active works across the OCI stack:
This is the most common active-active setup. You deploy identical servers across different Fault Domains or Availability Domains.
The Orchestrator: You use the OCI Load Balancer.
How it works: The Load Balancer uses a "Round Robin" or "Least Connections" policy to distribute incoming requests across all active servers.
Health Checks: If one server in the cluster fails, the Load Balancer detects it in milliseconds and redirects that server's share of the traffic to the remaining healthy nodes.
Oracle is famous for its Real Application Clusters (RAC), which is the industry's most mature active-active database technology.
True Active-Active: In OCI, you can rent a RAC DB System. Multiple database instances run on different physical servers but access the same underlying data.
Scalability: If your app hits a massive traffic spike, you can scale by adding more nodes to the active cluster without taking the database offline.
2026 Innovation: The Globally Distributed Autonomous Database now allows for active-active writes across different geographic regions by "sharding" data, ensuring users in London and users in New York both get local-speed performance.
For global applications, you can run active clusters in two different countries simultaneously.
DNS Steering: You use OCI Traffic Management Steering Policies.
Geolocation Routing: A user in Europe is directed to your "Active" cluster in Frankfurt, while a user in the US goes to your "Active" cluster in Ashburn.
Data Sync: To keep these clusters "in sync," 2026 architects typically use Oracle GoldenGate, which provides bi-directional data replication so that a write in one region is reflected in the other almost instantly.
| Feature | Active-Passive (Failover) | Active-Active (Cluster) |
| Resource Usage | 50% (Standby is idle) | 100% (All nodes working) |
| Recovery Time | 1–5 Minutes | Instant (Zero-downtime) |
| Complexity | Lower | Higher (Requires state sync) |
| Best For | Internal tools / Back office | E-commerce / Global APIs |
To make an active-active cluster successful, you must address two technical hurdles:
Session Persistence: If a user logs into Server A, you must ensure the Load Balancer keeps them there (Sticky Sessions) or use a shared "State Store" (like OCI Cache with Redis) so Server B knows who they are.
Latency: For "Stretch Clusters" (active-active across two physical sites), OCI recommends a maximum latency of 10ms to ensure that data synchronization doesn't slow down the user experience.
One of the best reasons to "rent" an active-active cluster is for Rolling Patching. In 2026, you can take one node of your cluster offline, update its software, and bring it back online while the other nodes continue to handle 100% of your customer traffic. You can update your entire infrastructure in the middle of a business day without a single user noticing.