Service Level for Yandex Managed Databases

This document defines Service Levels for Yandex Managed Service (for PostgreSQL), Yandex Managed Service (for ClickHouse), Yandex StoreDoc, Yandex Managed Service (for Valkey™), Yandex Managed Service (for MySQL®), Yandex Managed Service (for OpenSearch), Yandex Managed Service (for Sharded PostgreSQL) (collectively, “Service”) and is an integral part of Yandex.Cloud Customer Agreement (“Agreement”) and SLA available at: https://yandex.com/legal/cloud_sla/. Capitalized terms used herein but not defined herein shall have the meanings set forth in Agreement and Linked Documents.

Service Level in Region “Russia”:

DB Cluster, read

99.99%

DB Cluster, write

99.95%

Service Level in Region “Kazakhstan”:

DB Cluster, read

99.97%

DB Cluster, write

99.93%

Compensation Amount in Region “Russia”:

Availability of DB Cluster – read:

Service Uptime Percentage in Accrual Period

Compensation, % of Service Fee for Service in Accrual Period

< 99.99%-99.00%

10.0%

< 99.00%-90.00%

15.0%

less than 90.00%

30.0%

Availability of DB Cluster – write:

Service Uptime Percentage in Accrual Period

Compensation, % of Service Fee for Service in Accrual Period

< 99.95%-99.00%

10.0%

< 99.00%-90.00%

15.0%

less than 90.00%

30.0%

Compensation Amount in Region “Kazakhstan”:

Availability of DB Cluster – read:

Service Uptime Percentage in Accrual Period

Compensation, % of Service Fee for Service in Accrual Period

< 99.97%-99.00%

10.0%

< 99.00%-90.00%

15.0%

less than 90.00%

30.0%

Availability of DB Cluster – write:

Service Uptime Percentage in Accrual Period

Compensation, % of Service Fee for Service in Accrual Period

< 99.93%-99.00%

10.0%

< 99.00%-90.00%

15.0%

less than 90.00%

30.0%

Definition of Service Unavailability:

Yandex Managed Service for PostgreSQL that fall under the Service Level

A PostgreSQL cluster is subject to the Service Level if the cluster meets the following conditions:

  • a cluster contains two or more DB Hosts without configured replication sources;

  • DB Hosts have a host class with a CPU share of at least 100%;

  • DB Hosts without configured replication sources are located in different Availability Zones, provided that no more than half of all DB Hosts without configured replication sources are in each Availability Zone (applicable only for the "Russia" Region).

Determining Unavailability for a PostgreSQL Managed Service / Yandex Managed Service for PostgreSQL:

Unavailability of a DB Cluster meeting the above conditions for reading or writing via supported clients.

Yandex Managed Service for MySQL that fall under the Service Level

MySQL cluster is subject to the Service Level if the cluster meets the following conditions:

  • a cluster contains two or more DB Hosts without configured replication sources;

  • DB Hosts have a host class with a CPU share of at least 100%;

  • DB Hosts without configured replication sources are located in different Availability Zones, provided that no more than half of all DB Hosts without configured replication sources are in each Availability Zone (applicable only for the "Russia" Region);

  • the cluster setting "rpl_semi_sync_master_wait_for_slave_count" is equal to the maximum number of hosts within one Availability Zone (applicable only for the "Russia" Region);

Determining Unavailability for a MySQL-managed Service / Yandex Managed Service for MySQL:

Unavailability of a DB Cluster meeting the above conditions for reading or writing via supported clients.

Yandex Managed Service for ClickHouse that fall under the Service Level

An unsharded ClickHouse cluster

An unsharded ClickHouse cluster is subject to the Service Level if the cluster meets the following conditions:

  • a cluster contains two or more DB Hosts;

  • a cluster contains three or more Zookeeper or ClickHouse Keeper hosts;

  • DB Hosts have a host class with a CPU share of at least 100%;

  • DB Cluster hosts are located in different Availability Zones (applicable only for the "Russia" Region);

  • Zookeeper or ClickHouse Keeper hosts are located in different Availability Zones, provided that less than half of all Zookeeper or ClickHouse Keeper Hosts are located in each Availability Zone (applicable only for the "Russia" Region);

Sharded ClickHouse Cluster

A sharded ClickHouse cluster is subject to the Service Level if the cluster meets the following conditions:

  • a cluster contains three or more Zookeeper or ClickHouse Keeper hosts;

  • DB Hosts have a host class with a CPU share of at least 100%;

  • Zookeeper or ClickHouse Keeper hosts are located in different Availability Zones, provided that less than half of all Zookeeper or ClickHouse Keeper Hosts are located in each Availability Zone (applicable only for the "Russia" Region);

  • each Shard meets the following conditions:

  • Shard contains two or more DB Hosts;

  • Shard hosts are located in different Availability Zones (applicable only for the "Russia" Region);

Determining Unavailability for Yandex Managed Service for ClickHouse:

Unavailability of a DB Cluster meeting the above conditions for reading or writing via supported clients.

DATABASE Shard Unavailability — unavailability of a Shard meeting the conditions above for reading or writing via supported clients.

Yandex StoreDoc clusters that fall under the Service Level

An unsharded StoreDoc cluster

An unsharded StoreDoc cluster is subject to the Service Level if the cluster meets the following conditions:

  • the cluster contains two or more undisclosed DB Hosts (the hidden setting is false);

  • DB Hosts have a host class with a CPU share of at least 100%;

  • undisclosed Cluster Hosts are located in different Availability Zones, provided that no more than half of all undisclosed DB Hosts are in each Availability Zone (applicable only for the "Russia" Region);

Sharded StoreDoc Cluster

A sharded StoreDoc cluster is subject to the Service Level if each Shard in the cluster meets the following conditions:

  • a cluster contains two or more mongos or mongoinfra hosts;

  • a cluster contains two or more mongocfg or mongoinfra hosts;

  • Cluster Hosts have a host class with a CPU share of at least 100%;

  • mongos or mongoinfra cluster hosts are located in different Availability Zones (applicable only for the "Russia" Region);

  • mongocfg or mongoinfra Cluster Hosts are located in different Availability Zones, provided that no more than half of all mongocfg Hosts are in each Availability Zone (applicable only for the "Russia" Region);

  • Each Shard meets the following conditions:

  • Shard contains two or more undisclosed mongod Shard hosts (the hidden setting is false);

  • undisclosed mongod Shard hosts are located in different Availability Zones, provided that no more than half of all undisclosed mongod Shard hosts are in each Availability Zone (applicable only for the "Russia" Region);

Determining Unavailability for Yandex StoreDoc:

Unavailability of a DB Cluster meeting the above conditions for reading or writing via supported clients;

DATABASE Shard Unavailability — unavailability of a Shard meeting the conditions above for reading or writing via supported clients.

Yandex Managed Service for Valkey that fall under the Service Level

An unsharded Valkey cluster

An unsharded Valkey cluster is subject to the Service Level if the cluster meets the following conditions:

  • a cluster contains two or more DB Hosts;

  • DB Hosts have a host class with a CPU share of at least 100%;

  • DB Hosts are located in different Availability Zones, provided that no more than half of all DB Hosts are in each Availability Zone (applicable only for the "Russia" Region);

Sharded Valkey Cluster

A sharded Valkey cluster is subject to the Service Level if each Shard in the cluster meets the following conditions:

  • Shard contains two or more Shard hosts;

  • Shard hosts have a host class with a CPU share of at least 100%;

  • Shard hosts are located in different Availability Zones, provided that no more than half of all Shard Hosts are in each Availability Zone (applicable only for the "Russia" Region);

Determining Unavailability for Yandex Managed Service for Valkey:

Unavailability of a DB Cluster meeting the above conditions for reading or writing via supported clients; DATABASE Shard Unavailability — unavailability of a Shard meeting the conditions above for reading or writing via supported clients.

Yandex Managed Service for OpenSearch that fall under the Service Level

An OpenSearch cluster is subject to the Service Level if the cluster meets the following conditions:

  • a cluster contains three or more DB Hosts with the MANAGER role;

  • a cluster contains two or more DB Hosts with the DATA role;

  • DB Hosts with the MANAGER role are located in different Availability Zones, provided that no more than half of all DB Hosts with the MANAGER role are in each Availability Zone (applicable only for the Russia Region);

  • DB Hosts with the DATA role are located in different Availability Zones, provided that no more than half of all DB Hosts with the DATA role are in each Availability Zone (applicable only for the "Russia" Region);

  • cluster and index settings, as well as the configuration of DB Hosts with the DATA role, should allow data replication between different Availability Zones (applicable only for the "Russia" Region).

Determining Unavailability for Yandex Managed Service for OpenSearch:

Unavailability of a DB Cluster meeting the above conditions for reading or writing via supported clients.

A period of time starting from the moment of creation of a new DB Cluster or a DB Cluster from a backup, until the completion of the technical processes of creation, configuration, and, in the case of recovery from a backup, the end of the data recovery process is not counted as the Unavailability of the above Services.

Requirement for placement in several areas of availability is not applied during determination of the unavailability of Services mentioned above in the Region "Kazakhstan".

The list of supported clients for a specific type of DBMS can be found in technical documents of Service posted on Site.

Yandex Managed Service for Sharded PostgreSQL that fall under the Service Level

A Sharded PostgreSQL cluster is subject to the Service Level if the cluster meets the following conditions:

  • The cluster contains two or more Router or Infra hosts;

  • The cluster contains two or more Coordinator or Infra hosts;

  • The cluster’s Router or Infra hosts are located in different Availability Zones;

  • The cluster’s Coordinator or Infra hosts are located in different Availability Zones, provided that no more than half of all Coordinator hosts are located in any single Availability Zone;

  • Each cluster shard meets the following conditions:

    • The shard contains two or more DB hosts, excluding hosts with a fixed replication source (cascaded replicas);

    • All shard hosts have a host class with a guaranteed vCPU share of at least 100%;

    • All Shard Hosts have a host class with a vCPU count of at least 2;

    • Shard Hosts without a fixed replication source are located in different Availability Zones;

    • In each Availability Zone, no more than half of all Shard Hosts without a fixed replication source;

The Service Level Agreement does not apply to cases of Service Unavailability:

  • due to circumstances beyond Yandex’s control, including disruptions in external networks;

  • due to the Client’s failure to comply with the Service’s operating instructions and recommendations, such as running out of free disk space, placing an excessively high load on the system, etc. (for more details, see the Areas of Responsibility for MDB and Users);

  • during the process of updating the major version of the cluster or shards (for more details, see the update policy);

Determining Unavailability for a Sharded PostgreSQL Managed Service / Yandex Managed Service for Sharded PostgreSQL:

Unavailability of a DB Cluster meeting the above conditions for reading or writing via supported clients.

Definitions

Shard — the part of DB Cluster placed in one or several DB Hosts for load balancing.The process of forming a Shard is called sharding.

_______________________

Web address: https://yandex.com/legal/cloud_sla_mdb

Date of placement: June 16, 2026

Effective date: June 16, 2026

Previous version of the document: https://yandex.com/legal/cloud_sla_mdb/en/27122025

Previous version of the document: https://yandex.com/legal/cloud_sla_mdb/en/11112025

Previous version of the document: https://yandex.com/legal/adv_cloud_sla_mdb/15102025/

Previous version of the document: https://yandex.com/legal/cloud_sla_mdb/en/21122024

Previous version of the document: https://yandex.com/legal/cloud_sla_mdb/en/10122024

Previous version of the document: https://yandex.com/legal/adv_cloud_sla_mdb/27062024/

Previous version of the document: https://yandex.com/legal/adv_cloud_sla_mdb/01112023/

Previous version of the document: https://yandex.com/legal/adv_cloud_sla_mdb/13042021/

Previous version of the document: https://yandex.com/legal/adv_cloud_sla_mdb/18122020/

Previous version of the document: https://yandex.com/legal/adv_cloud_sla_mdb/23042019/