Yandex Cloud
Documentation
PresentationContact UsTry it for free
© 2026 Direct Cursus Technology L.L.C.
Yandex Managed Service for Valkey™
  • Getting started
    • Resource relationships
    • High availability clusters
    • Networking in Yandex Managed Service for Valkey™
    • Sharding
    • Autoscaling
    • Backups
    • Replication and persistence
    • Supported clients
    • Memory management
    • Available commands
    • Valkey™ modules
    • Quotas and limits
    • Storage
    • Maintenance
    • Valkey™ settings
    • Valkey™ versioning policy
  • Access management
  • Inspections and recommendations
  • Pricing policy
  • Terraform reference
  • Monitoring metrics
  • Audit Trails events
  • Public materials
  • Release notes
  • FAQ
  1. Concepts
  2. Autoscaling

Autoscaling in Yandex Managed Service for Valkey™

Written by
Yandex Cloud
Updated at October 27, 2026
View in Markdown

Autoscaling is changing the number of Valkey™ cluster shards depending on current load.

With autoscaling enabled, Yandex Managed Service for Valkey™ automatically adjusts the number of shards to match current load. This enables stable cluster operation without manually increasing or decreasing the number of shards.

For scaling, Yandex Managed Service for Valkey™ regularly collects cluster shard utilization metrics. It can then take the following decisions based on these metrics:

  • Increase the number of shards if the cluster is nearing overload.
  • Decrease the number of shards if the cluster is consistently underloaded.
  • Make no changes if utilization is within the specified thresholds.

To prevent reactions to brief load spikes, Master Autoscaler takes decisions based on aggregated metrics over a few minutes. Scaling only gets initiated when the critical value persists for ten minutes.

When the number of shards is automatically changed, cluster rebalancing starts to redistribute the data. As a result, some existing slots will be moved to the new shards.

Autoscaling advantages:

  • Cluster performance does not degrade as the load grows.
  • When idling, excessive shards get deleted, so you pay less for you cluster.
  • Query execution time does not increase as the cluster load grows and does not depend on the number of shards.
  • Computing resource allocation becomes more efficient: shards adjust to the changing app needs.

Autoscaling managementAutoscaling management

You can enable or disable shard autoscaling when creating or updating a sharded Valkey™ cluster.

You can set the following:

  • Autoscaling range:

    • Minimum number: at least 1.
    • Maximum number: maximum 10 and not less than the minimum number.
  • Thresholds for monitored shard utilization metrics for upward and downward scaling.

    For autoscaling management, the following per-shard Yandex Managed Service for Valkey™ metrics are used:

    • CPU usage
    • RAM usage

    You can set upper and lower thresholds for each metric. Based on the actual metric values, Yandex Managed Service for Valkey™ changes the number of shards as follows:

    • Scales the cluster up by one shard if any of the specified metrics continuously exceeds its threshold on at least one shard for ten minutes straight. None of the metrics has priority over others.
    • Scales the cluster down by one shard if all specified metrics on all shards remain below their lower thresholds for ten minutes straight.

    If you leave some thresholds unspecified, Yandex Managed Service for Valkey™ will use the default values for them. You can then edit the threshold values based on the results of cluster monitoring in autoscaling mode.

Was the article helpful?

Previous
Sharding
Next
Backups
© 2026 Direct Cursus Technology L.L.C.