Working with your Managed Service for PostgreSQL® cluster

In Managed Service for PostgreSQL, databases are organized into collections called clusters. You can create, modify or delete clusters.

If you want to move data from an external PostgreSQL cluster to a Managed PostgreSQL cluster, use migration or replication.

How to create clusters

Web console
CLI
Terraform

In the sidebar, go to Storage → PostgreSQL.
Click Create cluster.
On the page that opens, enter the cluster name and an optional description.
(Optional) Select one of the pooling modes:
- Transaction: A connection is allocated from a connection pool for every transaction. Once the transaction is completed, the connection returns to the pool. The transaction mode is suitable for a large number of idle connections. Additionally, this mode allows you to allocate a smaller pooling size in comparison with the session mode. However, the transaction mode offers a limited support of PostgreSQL features; not all PostgreSQL clients are compatible with this mode.
- Session (set by default): A connection is allocated from a connection pool for every client session. Once the session is terminated, the connection returns to the pool. More PostgreSQL clients support the session mode, over the transaction mode. Moreover, the session mode does not limit the number of PostgreSQL features that you can use. However, a database has to consume more resources to maintain long-lasting connections. You may need to increase the pooling size.
(Optional) In the Pooling size field, specify the number of connections allowed in a pool for a single database.
In the Access field, select whether the cluster should have a public IP address and access through the internet.
In the Resources section, configure the cluster resources:
1. Select the number of vCPUs and RAM to allocate to the cluster.
2. Specify the storage capacity.
Set the number of the cluster hosts.
In the Database section, specify the parameters for the connection to the database:
1. Enter the database name.
2. Enter the username. The following usernames are reserved; do not specify them: admin, repl, monitor, postgres, public and none.
3. Enter the password. It must contain at least:
  - 8 characters
  - One lowercase character
  - One uppercase character
  - One special character
(Optional) In the Backup settings section, configure the cluster backup:
1. Schedule the backup. The backup is created every day at the selected time.
2. Specify how long, in days, every backup and write-ahead logs (WALs) should be kept.
(Optional) Configure PostgreSQL additional parameters.
Supported PostgreSQL parameters
Click Create cluster.

The cluster will have the Provisioning status while it is being created. Before you perform any operations, wait for the Running status.

Install and configure the Nebius AI Cloud CLI.
Create a cluster:
```
nebius msp postgresql v1alpha1 cluster create \
    --name <cluster_name> \
    --description "<description>" \
    --network-id <network_ID> \
    --config-version 16 \
    --config-pooler-config-pooling-mode <transaction|session> \
    --config-pooler-config-max-pool-size <pooling_size> \
    --config-public-access=<true|false> \
    --config-template-resources-platform cpu-e2 \
    --config-template-resources-preset <preset> \
    --config-template-disk-type network-ssd \
    --config-template-disk-size-gibibytes <storage_size> \
    --config-template-hosts-count <number_of_hosts> \
    --bootstrap-db-name <database_name> \
    --bootstrap-user-name <username> \
    --bootstrap-user-password <password> \
    --backup-backup-window-start "<HH:MM:SS>" \
    --backup-retention-policy <number_of_days>d \
    <PostgreSQL_parameters>
```
This command contains the following parameters:
- --name: The cluster name.
- --description: The cluster description.
- --network-id: The network ID.
- --config-version: The PostgreSQL version. The service currently supports only version 16.
- --config-pooler-config-pooling-mode: The mode of how to allocate connections from the database connection pool. Managed PostgreSQL supports the following options:
  - Transaction: A connection is allocated from a connection pool for every transaction. Once the transaction is completed, the connection returns to the pool. The transaction mode is suitable for a large number of idle connections. Additionally, this mode allows you to allocate a smaller pooling size in comparison with the session mode. However, the transaction mode offers a limited support of PostgreSQL features; not all PostgreSQL clients are compatible with this mode.
  - Session (set by default): A connection is allocated from a connection pool for every client session. Once the session is terminated, the connection returns to the pool. More PostgreSQL clients support the session mode, over the transaction mode. Moreover, the session mode does not limit the number of PostgreSQL features that you can use. However, a database has to consume more resources to maintain long-lasting connections. You may need to increase the pooling size.
- --config-pooler-config-max-pool-size: The number of connections allowed in a pool for a single database.
- --config-public-access: If true, the cluster has both a public and a private IP address, and the cluster is accessible through the internet.
- --config-template-resources-platform: The resource platform for the database hosts. Managed PostgreSQL currently supports only the cpu-e2 platform.
- --config-template-resources-preset: The resource preset. Specify one of the presets available for cpu-e2.
- --config-template-disk-type: The disk type for the database hosts. Managed PostgreSQL currently supports only network-ssd disks.
- --config-template-disk-size-gibibytes: The number of gibibytes in the cluster storage.
- --config-template-hosts-count: The number of the cluster hosts.
- --bootstrap-db-name: The database name.
- --bootstrap-user-name: The username of the database user. The following usernames are reserved; do not specify them: admin, repl, monitor, postgres, public and none.
- --bootstrap-user-password: The password for this user. The password must contain at least:
  - 8 characters
  - One lowercase character
  - One uppercase character
  - One special character
- --backup-backup-window-start: The backup schedule, in the HH:MM:SS format. The backup is created every day at the selected time.
- --backup-retention-policy: How long, in days, every backup and write-ahead logs (WALs) should be kept. For example, 7d or 30d.
To configure the database, you can also use PostgreSQL parameters.
Supported PostgreSQL parameters
- —config-postgresql-config-16-autovacuum-work-mem: From -1 to 2147483647 (in kilobytes).
- —config-postgresql-config-16-autovacuum-vacuum-cost-delay: From -1 to 100 (in milliseconds).
- —config-postgresql-config-16-autovacuum-vacuum-cost-limit: From -1 to 10000.
- —config-postgresql-config-16-autovacuum-naptime: From 1 to 2147483 (in seconds).
- —config-postgresql-config-16-autovacuum-vacuum-scale-factor: From 0 to 100.
- —config-postgresql-config-16-autovacuum-analyze-scale-factor: From 0 to 100.
- —config-postgresql-config-16-statement-timeout: From 0 to 2147483647 (in milliseconds).
- —config-postgresql-config-16-idle-in-transaction-session-timeout: From 0 to 2147483647 (in milliseconds).
- —config-postgresql-config-16-search-path
- —config-postgresql-config-16-default-transaction-read-only
- —config-postgresql-config-16-max-connections: From 100 to 12800.
- —config-postgresql-config-16-shared-buffers: From 128 to 8589934584 (in kilobytes).

Install and configure the Nebius AI Cloud provider for Terraform.
Create the following configuration file:
```
resource "nebius_msp_postgresql_v1alpha1_cluster" "<cluster_name>" {
  name        = "<cluster_name>"
  description = "<cluster_description>"
  parent_id   = "<project_ID>"
  network_id  = "<network_ID>"

  config = {
    version = "16"
    pooler_config = {
      pooling_mode  = "<TRANSACTION|SESSION>"
      max_pool_size = <pooling_size>
    }
    public_access = <true|false>
    template = {
      resources = {
        platform = "cpu-e2"
        preset   = "<preset>"
      }
      disk = {
        type           = "network-ssd"
        size_gibibytes = <storage_volume>
      }
      hosts = {
        count = <number_of_hosts>
      }
    }
    postgresql_config_16 = {
      <PostgreSQL_parameters>
    }
  }

  bootstrap = {
    db_name       = "<database_name>"
    user_name     = "<username>"
    user_password = "<password>"
  }

  backup = {
    backup_window_start = "HH:MM:SS"
    retention_policy    = "<number_of_days>d"
  }
}
```
The file contains the following parameters:
- name: The cluster name.
- description: The cluster description.
- parent_id: Project ID.
- network_id: Network ID.
- config: The cluster configuration parameters:
  - version: The PostgreSQL version. The service currently supports only version 16.
  - pooler_config.pooling_mode: The mode of how to allocate connections from the database connection pool. Managed PostgreSQL supports the following options:
    - Transaction: A connection is allocated from a connection pool for every transaction. Once the transaction is completed, the connection returns to the pool. The transaction mode is suitable for a large number of idle connections. Additionally, this mode allows you to allocate a smaller pooling size in comparison with the session mode. However, the transaction mode offers a limited support of PostgreSQL features; not all PostgreSQL clients are compatible with this mode.
    - Session (set by default): A connection is allocated from a connection pool for every client session. Once the session is terminated, the connection returns to the pool. More PostgreSQL clients support the session mode, over the transaction mode. Moreover, the session mode does not limit the number of PostgreSQL features that you can use. However, a database has to consume more resources to maintain long-lasting connections. You may need to increase the pooling size.
  - pooler_config.max_pool_size: The number of connections allowed in a pool for a single database.
  - public_access: If true, the cluster has both a public and a private IP address, and the cluster is accessible through the internet.
  - template.resources.platform: The resource platform for the database hosts. Managed PostgreSQL currently supports only the cpu-e2 platform.
  - template.resources.preset: The resource preset. Specify one of the presets available for cpu-e2.
  - template.disk.type: The disk type for the database hosts. Managed PostgreSQL currently supports only network-ssd disks.
  - template.disk.size_gibibytes: The number of gibibytes in the cluster storage.
  - template.hosts.count: The number of the cluster hosts.
  - postgresql_config_16: Additional PostgreSQL parameters for the database configuration.
    Supported PostgreSQL parameters
    postgresql_config_16 = { autovacuum_work_mem = <in_kilobytes> autovacuum_vacuum_cost_delay = <in_milliseconds> autovacuum_vacuum_cost_limit = <> autovacuum_naptime = <in_seconds> autovacuum_vacuum_scale_factor = <> autovacuum_analyze_scale_factor = <> statement_timeout = <in_milliseconds> idle_in_transaction_session_timeout = <in_milliseconds> search_path = "<>" default_transaction_read_only = <true|false> max_connections = <> shared_buffers = <in_kilobytes> }
    Supported PostgreSQL parameters:
    
    autovacuum_work_mem: From -1 to 2147483647.
    
    autovacuum_vacuum_cost_delay: From -1 to 100.
    
    autovacuum_vacuum_cost_limit: From -1 to 10000.
    
    autovacuum_naptime: From 1 to 2147483.
    
    autovacuum_vacuum_scale_factor: From 0 to 100.
    
    autovacuum_analyze_scale_factor: From 0 to 100.
    
    statement_timeout: From 0 to 2147483647.
    
    idle_in_transaction_session_timeout: From 0 to 2147483647.
    
    search_path
    
    default_transaction_read_only
    
    max_connections: From 100 to 12800.
    
    shared_buffers: From 128 to 8589934584.
- bootstrap.db_name: The database name.
- bootstrap.user_name: The username of the database user. The following usernames are reserved; do not specify them: admin, repl, monitor, postgres, public and none.
- bootstrap.user_password: The password for this user. The password must contain at least:
  - 8 characters
  - One lowercase character
  - One uppercase character
  - One special character
- backup.backup_window_start: The backup schedule, in the HH:MM:SS format. The backup is created every day at the specified time.
- backup.retention_policy: How long, in days, every backup and write-ahead logs (WALs) should be kept. For example, 7d or 30d.
Check that the configuration is correct:
```
terraform validate
```
Apply the changes:
```
terraform apply
```

How to restore clusters from backups

Managed PostgreSQL creates daily backups for each cluster, and you can create a new cluster from a backup. For more details and instructions, see Backups in Managed Service for PostgreSQL®.

How to modify clusters

Web console
CLI
Terraform

In the sidebar, go to Storage → PostgreSQL.
Open the page of the required cluster and then go to the Settings tab.
Change the database pooling mode. Managed PostgreSQL supports the following options:
- Transaction: A connection is allocated from a connection pool for every transaction. Once the transaction is completed, the connection returns to the pool. The transaction mode is suitable for a large number of idle connections. Additionally, this mode allows you to allocate a smaller pooling size in comparison with the session mode. However, the transaction mode offers a limited support of PostgreSQL features; not all PostgreSQL clients are compatible with this mode.
- Session (set by default): A connection is allocated from a connection pool for every client session. Once the session is terminated, the connection returns to the pool. More PostgreSQL clients support the session mode, over the transaction mode. Moreover, the session mode does not limit the number of PostgreSQL features that you can use. However, a database has to consume more resources to maintain long-lasting connections. You may need to increase the pooling size.
In the Access field, select whether the cluster should have a public IP address and access through the internet.
In the Resource preset field, change the number of vCPUs and RAM that should be allocated to the cluster.
Change the number of hosts in the cluster.
In the Database section, change the password of the database user. The password must contain at least:
- 8 characters
- One lowercase character
- One uppercase character
- One special character
Configure PostgreSQL additional parameters.
Supported PostgreSQL parameters

Get the ID of the cluster via its name:

export NB_PG_CLUSTER_ID=$(nebius msp postgresql v1alpha1 cluster \
  get-by-name --name <cluster_name> --format json | jq -r '.metadata.id')

Alternatively, you can use the listing command: nebius msp postgresql v1alpha1 cluster list.

Update the cluster settings:
```
nebius msp postgresql v1alpha1 cluster update \
  --id $NB_PG_CLUSTER_ID \
  --config-pooler-config-pooling-mode <transaction|session> \
  --config-pooler-config-max-pool-size <pooling_size> \
  --config-public-access=<true|false> \
  --config-template-resources-preset <preset> \
  --config-template-hosts-count <number_of_hosts> \
  --bootstrap-user-password <password> \
  <PostgreSQL_parameters>
```
The command contains the following parameters:
- --config-pooler-config-pooling-mode: The mode of how to allocate connections from the database connection pool. Managed PostgreSQL supports the following options:
  - Transaction: A connection is allocated from a connection pool for every transaction. Once the transaction is completed, the connection returns to the pool. The transaction mode is suitable for a large number of idle connections. Additionally, this mode allows you to allocate a smaller pooling size in comparison with the session mode. However, the transaction mode offers a limited support of PostgreSQL features; not all PostgreSQL clients are compatible with this mode.
  - Session (set by default): A connection is allocated from a connection pool for every client session. Once the session is terminated, the connection returns to the pool. More PostgreSQL clients support the session mode, over the transaction mode. Moreover, the session mode does not limit the number of PostgreSQL features that you can use. However, a database has to consume more resources to maintain long-lasting connections. You may need to increase the pooling size.
- --config-pooler-config-max-pool-size: The number of connections allowed in a pool for a single database.
- --config-public-access: If true, the cluster has both a public and a private IP address, and the cluster is accessible through the internet.
- --config-template-resources-preset: The resource preset. Specify one of the presets available for cpu-e2.
- --config-template-hosts-count: The number of the cluster hosts.
- --bootstrap-user-password: The password for this user. The password must contain at least:
  - 8 characters
  - One lowercase character
  - One uppercase character
  - One special character
To configure the database, you can additionally use PostgreSQL parameters.
Supported PostgreSQL parameters
- —config-postgresql-config-16-autovacuum-work-mem: From -1 to 2147483647 (in kilobytes).
- —config-postgresql-config-16-autovacuum-vacuum-cost-delay: From -1 to 100 (in milliseconds).
- —config-postgresql-config-16-autovacuum-vacuum-cost-limit: From -1 to 10000.
- —config-postgresql-config-16-autovacuum-naptime: From 1 to 2147483 (in seconds).
- —config-postgresql-config-16-autovacuum-vacuum-scale-factor: From 0 to 100.
- —config-postgresql-config-16-autovacuum-analyze-scale-factor: From 0 to 100.
- —config-postgresql-config-16-statement-timeout: From 0 to 2147483647 (in milliseconds).
- —config-postgresql-config-16-idle-in-transaction-session-timeout: From 0 to 2147483647 (in milliseconds).
- —config-postgresql-config-16-search-path
- —config-postgresql-config-16-default-transaction-read-only
- —config-postgresql-config-16-max-connections: From 100 to 12800.
- —config-postgresql-config-16-shared-buffers: From 128 to 8589934584 (in kilobytes).
Only the parameters from the command above can be changed. To perform a full cluster update, add --full to the command. This will update all parameters with the default values or the values specified in the --full command.

Modify the manifest with the deployed infrastructure:
```
resource "nebius_msp_postgresql_v1alpha1_cluster" "<cluster_name>" {
  name        = "<cluster_name>"

  config = {
    pooler_config = {
      pooling_mode  = "<TRANSACTION|SESSION>"
      max_pool_size = <pooling_size>
    }
    public_access = <true|false>
    template = {
      resources = {
        preset   = "<preset>"
        ...
      }
      hosts = {
        count = <number_of_hosts>
      }
      postgresql_config_16 = {
        <PostgreSQL_parameters>
      }
      ...
    }
    postgresql_config_16 = {
      <PostgreSQL_parameters>
    }
    ...
  }

  bootstrap = {
    user_password = "<password>"
    ...
  }
  ...
}
```
You can change the following parameters:
- name: The cluster name.
- config: The cluster configuration parameters:
  - pooler_config.pooling_mode: The mode of how to allocate connections from the database connection pool. Managed PostgreSQL supports the following options:
    - Transaction: A connection is allocated from a connection pool for every transaction. Once the transaction is completed, the connection returns to the pool. The transaction mode is suitable for a large number of idle connections. Additionally, this mode allows you to allocate a smaller pooling size in comparison with the session mode. However, the transaction mode offers a limited support of PostgreSQL features; not all PostgreSQL clients are compatible with this mode.
    - Session (set by default): A connection is allocated from a connection pool for every client session. Once the session is terminated, the connection returns to the pool. More PostgreSQL clients support the session mode, over the transaction mode. Moreover, the session mode does not limit the number of PostgreSQL features that you can use. However, a database has to consume more resources to maintain long-lasting connections. You may need to increase the pooling size.
  - pooler_config.max_pool_size: The number of connections allowed in a pool for a single database.
  - public_access: If true, the cluster has both a public and a private IP address, and the cluster is accessible through the internet.
  - template.resources.preset: The resource preset. Specify one of the presets available for cpu-e2.
  - template.hosts.count: The number of the cluster hosts.
  - postgresql_config_16: Additional PostgreSQL parameters for the database configuration.
    Supported PostgreSQL parameters
    postgresql_config_16 = { autovacuum_work_mem = <in_kilobytes> autovacuum_vacuum_cost_delay = <in_milliseconds> autovacuum_vacuum_cost_limit = <> autovacuum_naptime = <in_seconds> autovacuum_vacuum_scale_factor = <> autovacuum_analyze_scale_factor = <> statement_timeout = <in_milliseconds> idle_in_transaction_session_timeout = <in_milliseconds> search_path = "<>" default_transaction_read_only = <true|false> max_connections = <> shared_buffers = <in_kilobytes> }
    Supported PostgreSQL parameters:
    
    autovacuum_work_mem: From -1 to 2147483647.
    
    autovacuum_vacuum_cost_delay: From -1 to 100.
    
    autovacuum_vacuum_cost_limit: From -1 to 10000.
    
    autovacuum_naptime: From 1 to 2147483.
    
    autovacuum_vacuum_scale_factor: From 0 to 100.
    
    autovacuum_analyze_scale_factor: From 0 to 100.
    
    statement_timeout: From 0 to 2147483647.
    
    idle_in_transaction_session_timeout: From 0 to 2147483647.
    
    search_path
    
    default_transaction_read_only
    
    max_connections: From 100 to 12800.
    
    shared_buffers: From 128 to 8589934584.
- bootstrap.user_password: The password for this user. The password must contain at least:
  - 8 characters
  - One lowercase character
  - One uppercase character
  - One special character
If you change other settings, the cluster will be firstly destroyed and then recreated with the new configuration.
Check that the configuration is correct:
```
terraform validate
```
Apply the changes:
```
terraform apply
```

How to delete clusters

Web console
CLI

In the sidebar, go to Storage → PostgreSQL.
Click the cluster you want to delete.
On the Settings tab, click Delete cluster.
In the dialog window, enter the cluster name and click Delete cluster.

Copy the ID of the cluster you want to delete. Run the following command:
```
nebius msp postgresql v1alpha1 cluster list --page-size 10
```
On the list of clusters, you’ll get the cluster IDs under metadata:id.

Export the ID of the cluster to an environment variable:

export CLUSTER_ID_DELETION=<cluster_ID_from_clipboard>

Run the following command to delete your cluster:

nebius msp postgresql v1alpha1 cluster delete --id $CLUSTER_ID_DELETION

You can later restore a deleted cluster from its backup. Managed PostgreSQL preserves backups of deleted clusters for the retention period configured in the backup settings.

Postgres, PostgreSQL and the Slonik Logo are trademarks or registered trademarks of the PostgreSQL Community Association of Canada, and used with their permission.

Object Storage

Managed Service for PostgreSQL®

Container Registry

Working with your Managed Service for PostgreSQL® cluster

How to create clusters

How to restore clusters from backups

How to modify clusters

How to delete clusters

Object Storage

Managed Service for PostgreSQL®

Container Registry

​How to create clusters

​How to restore clusters from backups

​How to modify clusters

​How to delete clusters

How to create clusters

How to restore clusters from backups

How to modify clusters

How to delete clusters