How do I get Azure Managed Redis to deploy with Replica nodes?

Question

How do I get Azure Managed Redis to deploy with Replica nodes?

Don Mitchell 0 Microsoft Employee

(Azure Managed Redis question but only tag I could find was Azure Cache?)

Context - Evaluating AMR for a Redis migration in an attempt to provide better scaling and reduce total number of connections to single node.

Specifically looking at utilizing SSubscribe coupled with connections to secondary nodes to provide horizontal scaling of Shard PubSub connections while limiting impact to primary nodes.

According to the AMR Architecture doc this should be possible:

"The OSS clustering protocol requires the client to make the correct shard connections. ... use the CLUSTER NODES command to determine the exact ports used for the primary and replica shards and make the shard connections for you."

but I cannot figure out how to deploy with replica shards, so when performing the CLUSTER NODES query all I see are primary nodes.

Am I just missing an obvious config somewhere or do I need to load a bunch of data into Redis first to get it to trigger the creation of replica shards or is it something else I am missing? (Sincere apologies if I'm just blind and missing a checkbox somewhere)

Don Mitchell 0 Reputation points Microsoft Employee

2026-04-07T19:08:00.7666667+00:00

Thanks Manoj, that reference document is the one I was quoting when I said I expected to see the replica nodes. We are using OSS clustering as we would rather calculate shards and ping the nodes directly than take the performance hit going through a proxy would involve.

That being said, I was also just told that replica nodes are never made visible, so I am getting conflicting information. Can you confirm that the documentation is correct and if so provide any guidance on what I need to change to get and use the replica nodes?
Manoj Kumar Boyini 13,930 Reputation points Microsoft External Staff Moderator

2026-04-07T23:32:05.9266667+00:00

Hi @Don Mitchell

Thanks for the follow-up. In Azure Managed Redis (AMR), replica shards do exist and are used for high availability and failover, but they aren’t exposed as independent, directly connectable endpoints for client workloads. That’s different from open-source Redis, where connecting to replicas is a common scaling pattern.

In AMR, clients connect to node-level endpoints, and the service routes traffic internally to the appropriate shard. Even in OSS clustering mode, while CLUSTER NODES can provide topology/port discovery, the endpoints exposed are node-level proxy endpoints rather than individual shard processes. As a result, replica shards aren’t available for direct use for read scaling.

You’re not missing any configuration, and loading more data won’t cause replicas to appear as connectable nodes. There isn’t a setting to expose replicas for direct client access in AMR today.

For scaling Pub/Sub in AMR, the recommended approach is to scale out by increasing the number of shards (more primaries) and distributing your Pub/Sub workload across those shards.

Hope this helps, Please let us know if you have any questions and concerns.
Manoj Kumar Boyini 13,930 Reputation points Microsoft External Staff Moderator

2026-04-08T19:25:30.8933333+00:00

Hi @Don Mitchell

I hope you had a chance to review the information shared earlier, and I hope this information has been helpful! If you still have questions, please let us know what is needed in the comments so the question can be answered.

2 answers

Your answer

Don Mitchell 0 Reputation points Microsoft Employee

2026-04-07T19:08:00.7666667+00:00

Thanks Manoj, that reference document is the one I was quoting when I said I expected to see the replica nodes. We are using OSS clustering as we would rather calculate shards and ping the nodes directly than take the performance hit going through a proxy would involve.

That being said, I was also just told that replica nodes are never made visible, so I am getting conflicting information. Can you confirm that the documentation is correct and if so provide any guidance on what I need to change to get and use the replica nodes?
Manoj Kumar Boyini 13,930 Reputation points Microsoft External Staff Moderator

2026-04-07T23:32:05.9266667+00:00

Hi @Don Mitchell

Thanks for the follow-up. In Azure Managed Redis (AMR), replica shards do exist and are used for high availability and failover, but they aren’t exposed as independent, directly connectable endpoints for client workloads. That’s different from open-source Redis, where connecting to replicas is a common scaling pattern.

In AMR, clients connect to node-level endpoints, and the service routes traffic internally to the appropriate shard. Even in OSS clustering mode, while CLUSTER NODES can provide topology/port discovery, the endpoints exposed are node-level proxy endpoints rather than individual shard processes. As a result, replica shards aren’t available for direct use for read scaling.

You’re not missing any configuration, and loading more data won’t cause replicas to appear as connectable nodes. There isn’t a setting to expose replicas for direct client access in AMR today.

For scaling Pub/Sub in AMR, the recommended approach is to scale out by increasing the number of shards (more primaries) and distributing your Pub/Sub workload across those shards.

Hope this helps, Please let us know if you have any questions and concerns.
Manoj Kumar Boyini 13,930 Reputation points Microsoft External Staff Moderator

2026-04-08T19:25:30.8933333+00:00

Hi @Don Mitchell

I hope you had a chance to review the information shared earlier, and I hope this information has been helpful! If you still have questions, please let us know what is needed in the comments so the question can be answered.

Answer 1

Hi Don Mitchell

The replica shards are not accessible as they are there as a backup for the primary. The only time the replica is available is via a failover where replica and primary switch. As far as connecting to individual "nodes" in a cache that has OSS clustering. You would use the particular port for each node. These ports start at 8500 and increment consecutively.

The language here is misleading. What it should explain is "determine the ports for primary and replica. And make the connections to the correct PRIMARY shard for you"

Answer 2

Hi Don Mitchell

The behavior you’re seeing is expected. In Azure Managed Redis (AMR), replica shards are not created dynamically and are not controlled through a user-configurable setting like in classic Azure Cache for Redis. Replication is part of the service’s high-availability architecture and is handled automatically by the platform.

The key factor in your scenario is the clustering policy.

Azure Managed Redis supports two clustering modes:

OSS Clustering
– Exposes full Redis Cluster topology
– CLUSTER NODES returns primary and replica shard details
– Clients can connect to individual shards
Enterprise (proxy-based) Clustering
– Uses a single endpoint
– Does not expose shard or replica topology
– CLUSTER NODES is not supported
– Replication still exists internally but is abstracted by the service

If your deployment is using Enterprise clustering, you will not see replica nodes via CLUSTER NODES, even though replicas exist internally.

Replica nodes are not created based on data volume. They are provisioned as part of the service architecture, and their visibility depends on the clustering mode selected during deployment.

Because the clustering policy cannot be changed after provisioning, if you require visibility into shard topology (for example, to connect to replicas for Pub/Sub scaling), you will need to deploy the cache using OSS clustering.

Helpful Reference:
https://learn.microsoft.com/en-us/azure/redis/architecture

Hope this helps, Please let us know if you have any questions and concerns.

Don Mitchell 0 Reputation points Microsoft Employee

2026-04-14T17:16:11.25+00:00

To be clear, my deployment is using OSS Clustering, NOT Enterprise.

Running CLUSTER NODES was NOT giving me the full Redis Cluster topology or any information about Replica Shards, nor could I connect to any nodes in the shard but the primary, which was what prompted this question.

What do I need to do so that I see behavior you are describing above?

Share via

How do I get Azure Managed Redis to deploy with Replica nodes?

2 answers

Your answer