View a markdown version of this page

Managing reserved memory for Valkey and Redis OSS - Amazon ElastiCache
How much memory to reserveConfiguring reserved memory
Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Managing reserved memory for Valkey and Redis OSS

Each host type has an amount of memory dedicated for Valkey and Redis OSS. This is the advertised memory. The reserved-memory-percent parameter sets aside part of this memory for nondata uses:

  • Client output buffers on replicas

  • Memory lost to fragmentation

  • Copy-on-write memory used by forked bgsave

The maxmemory value equals the advertised memory minus the reserve. For example, with 13.5 GB of advertised memory and a 25% reserve, the maxmemory is about 10.1 GB.

Important

The reserved-memory-percent defaults to 25%. Do not reduce this value. A lower reserve can cause data loss, especially during upgrades or other changes to your cluster.

How much memory to reserve

The reserved-memory-percent parameter defaults to 25%. Do not reduce this value. However, there are situations where you can benefit from increasing the memory reserve.

Micro and small instances

The micro and small instances are sized for light activity without the burden of replication. To improve operational reliability of these instances for production workloads, increase the memory reserve to 30% on small instances and 50% on micro instances.

Forcing a forked bgsave or full sync

During a traditional forked bgsave or full sync, Valkey and Redis OSS memory usage can double (or more), depending on the traffic load and pattern. If FreeableMemory is insufficient to run a traditional forked operation, Amazon ElastiCache automatically switches to a forkless save mechanism. This forkless mechanism uses much less memory, but can add client latencies — especially with very large key/value items. If these latencies are impactful for your use case, increase the memory reserve to ensure enough memory for a traditional forked save.

If instructed by Amazon Support

Unusual traffic patterns or requirements can result in unique memory needs. Increase reserved-memory-percent if directed by Amazon Support.

Configuring reserved memory

The reserved-memory-percent parameter is specific to ElastiCache. It isn't part of the Valkey or Redis OSS open source code. To change this value, use a custom parameter group. You can't modify default parameter groups.

If 25% works for you, no changes are needed. To use a different value, follow these steps.

To change the reserved-memory-percent value

  1. Run the following Amazon CLI command to create a custom parameter group. Set the family to match your engine version. For more information about creating parameter groups, see Creating an ElastiCache parameter group.

    aws elasticache create-cache-parameter-group \
   --cache-parameter-group-name my-parameter-group \
   --description "Custom reserved memory settings" \
   --cache-parameter-group-family redis7

    You need only one custom parameter group per engine family. The percent value works the same for all node types.

  2. Set reserved-memory-percent to the value you want. For more information about modifying parameter groups, see Modifying an ElastiCache parameter group.

    aws elasticache modify-cache-parameter-group \
   --cache-parameter-group-name my-parameter-group \
   --parameter-name-values "ParameterName=reserved-memory-percent, ParameterValue=30"

  3. Apply the custom parameter group to your cluster or replication group.

    To modify a cluster, run the following Amazon CLI command. For more information about modifying clusters, see Modifying an ElastiCache cluster.

    aws elasticache modify-cache-cluster \
   --cache-cluster-id my-cluster \
   --cache-parameter-group-name my-parameter-group \
   --apply-immediately

    To modify a replication group, run the following Amazon CLI command. For more information about modifying replication groups, see Modifying a replication group.

    aws elasticache modify-replication-group \
   --replication-group-id my-replication-group \
   --cache-parameter-group-name my-parameter-group \
   --apply-immediately