Remote Read Replicas

This feature requires an Enterprise license. To upgrade, contact Redpanda sales.

A Remote Read Replica topic is a read-only topic that mirrors a topic on a different cluster. Remote Read Replicas work with both Tiered Storage and archival storage.

When a topic has object storage enabled, you can create a separate remote cluster just for consumers of this topic, and populate its topics from remote storage. A read-only topic on a remote cluster can serve any consumer, without increasing the load on the origin cluster. Use cases for Remote Read Replicas include data analytics, offline model training, and development clusters.

You can create Remote Read Replica topics in a Redpanda cluster that directly accesses data stored in cloud object storage. Because these read-only topics access data directly from cloud object storage instead of the topics' origin cluster, there’s no impact to the performance of the cluster. Topic data can be consumed within a region of your choice, regardless of the region where it was produced.

  • The Remote Read Replica cluster must run on the same version of Redpanda as the origin cluster, or just one feature release ahead of the origin cluster. For example, if the origin cluster is version 23.1, the Remote Read Replica cluster can be 23.2, but not 23.4. It cannot skip feature releases.

  • When upgrading, upgrade the Remote Read Replica cluster before upgrading the origin cluster.

  • When upgrading to Redpanda 23.2, metadata from object storage is not synchronized until all brokers in the cluster are upgraded. If you need to force a mixed-version cluster to sync read replicas, move partition leadership to brokers running the original version.

To create a Remote Read Replica topic in another region, consider using a multi-region bucket to simplify deployment and optimize performance.

Prerequisites

You need the following:

  • An origin cluster with Tiered Storage set up.

  • A topic on the origin cluster, which you can use as a Remote Read Replica topic on the remote cluster.

  • A separate remote cluster in the same region as the bucket or container used for the origin cluster.

    • If you use a multi-region bucket/container, you can create the read replica cluster in any region that has that bucket/container.

    • If you use a single-region bucket/container, the remote cluster must be in the same region as the bucket/container.

This feature requires an Enterprise license. To upgrade, contact Redpanda sales.

To check if you already have a license key applied to your cluster:

rpk cluster license info

Configure object storage for the remote cluster

You must configure access to the same object storage as the origin cluster.

To set up a Remote Read Replica topic on a separate remote cluster:

  1. Create a remote cluster for the Remote Read Replica topic.

    • If that’s a multi-region bucket/container, you can create the read replica cluster in any region that has that bucket/container.

    • If that’s a single-region bucket/container, the remote cluster must be in the same region as the bucket/container.

  2. Run rpk cluster config edit, and then specify properties specific to your object storage provider (your cluster will require a restart after any changes to these properties):

    Property Description

    cloud_storage_enabled

    Must be set to true to enable object storage.

    cloud_storage_bucket: "none"

    No AWS or GCS bucket is needed for the remote cluster.

    cloud_storage_access_key

    AWS or GCS access key.
    Required for AWS and GCS authentication with access keys.

    cloud_storage_secret_key

    AWS or GCS secret key.
    Required for AWS and GCS authentication with access keys.

    cloud_storage_region

    Object storage region of the remote cluster.
    Required for AWS and GCS.

    cloud_storage_api_endpoint

    AWS or GCS API endpoint.
    - For AWS, this can be left blank. It’s generated automatically using the region and bucket.
    - For GCS, use storage.googleapis.com.

    cloud_storage_azure_container

    Azure container name.
    Required for ABS.

    cloud_storage_azure_storage_account

    Azure account name.
    Required for ABS.

    cloud_storage_azure_shared_key

    Azure shared key.
    Required for ABS.

Create a Remote Read Replica topic

To create the Remote Read Replica topic, run:

rpk topic create <topic_name> -c redpanda.remote.readreplica=<bucket_name>
  • For <topic_name>, use the same name as the original topic.

  • For <bucket_name>, use the bucket/container specified in the cloud_storage_bucket or cloud_storage_azure_container properties for the origin cluster.

  • The Remote Read Replica cluster must run on the same version of Redpanda as the origin cluster, or just one feature release ahead of the origin cluster. For example, if the origin cluster is version 23.1, the Remote Read Replica cluster can be 23.2, but not 23.4. It cannot skip feature releases.

  • During upgrades, upgrade the Remote Read Replica cluster before upgrading the origin cluster.

  • Do not use redpanda.remote.read or redpanda.remote.write with redpanda.remote.readreplica. Redpanda ignores the values for remote read and remote write properties on read replica topics.

Reduce lag in data availability

When object storage is enabled on a topic, Redpanda copies closed log segments to the configured object store. Log segments are closed when the value of the segment size has been reached. A topic’s object store thus lags behind the local copy by the log_segment_size or, if set, by the topic’s segment.bytes value. To reduce this lag in the data availability for the Remote Read Replica:

  • You can lower the value of segment.bytes. This lets Redpanda archive smaller log segments more frequently, at the cost of increasing I/O and file count.

  • Self-hosted implementations can set an idle timeout with cloud_storage_segment_max_upload_interval_sec to force Redpanda to periodically archive the contents of open log segments to object storage. This is useful if a topic’s write rate is low and log segments are kept open for long periods of time. The appropriate interval may depend on your total partition count: a system with less partitions can handle a higher number of segments per partition.