Skip to main content

Replication factors in Aiven for OpenSearch®

The replication factor in Aiven for OpenSearch® determines the number of copies (replicas) of each index shard, which ensures data availability and prevents data loss.

The number_of_replicas is an index-level setting that defines how many replicas each primary shard has. By default, it is set to 1, meaning each shard has one replica for redundancy. You can configure this when creating or updating an index in the OpenSearch Dashboard.

Automatic replication factor adjustment​

Aiven for OpenSearch automatically adjusts replication factors in your indexes to maintain data availability.

  • The maximum number_of_replicas is the total number of nodes in your cluster minus one. This ensures that your data is replicated across all available nodes.

    Example: In a 3-node cluster, the maximum number_of_replicas is 2, which replicates all shards across the three nodes.

  • If you set number_of_replicas to 0 in a multi-node cluster, Aiven automatically increases it to 1. This ensures that your data remains available if one node fails.

Replication factor 0​

Setting the replication factor (number_of_replicas) to 0 means your data has no replicas. This reduces storage usage but significantly increases the risk of data loss if a node in the cluster fails.

note

Before enabling this configuration, consult with your account manager to discuss your use case and agree on the reduced SLA.

When to use replication factor 0​

Replication factor 0 is generally not recommended for most OpenSearch use cases. In specific cases, you might consider it for:

  • Non-critical environments: QA, testing, or development clusters where data loss is acceptable.
  • Temporary data: Scenarios where data can be recreated and storage costs need to be minimized.

Risks and considerations​

Setting the replication factor to 0 increases the risk of data loss. If a node failure occurs, Aiven for OpenSearch automatically restores data from the latest snapshot, but any data added after that snapshot will be lost. If no snapshot exists, an empty index is created, resulting in the loss of any unsaved data.

note
  • In a scenario where a node fails before the first snapshot is taken, the system cannot recover the data. Aiven for OpenSearch automatically recreates the missing index, but the recreated index is empty.
  • In a normal node recovery scenario, Aiven for OpenSearch loads the latest snapshot to restore your data.

Set replication factor 0​

Before enabling this option, consult with your account manager to discuss your use case and agree on the reduced SLA. Afterward, contact Aiven support to enable the option to set the replication factor (number_of_replicas) to 0.