Configure Cloud SQL and the external server for replication

Stay organized with collections Save and categorize content based on your preferences.

This page describes how to configure your external server for replication to Cloud SQL, create a source representation instance on Cloud SQL, and replicate the data to Cloud SQL. You need to go through all the steps on this page before proceeding to the replication steps.

An alternative to the steps described on this page is the Database Migration Service, which offers continuous replication or one-time database migration from an external server to Cloud SQL.

Before you begin

Terminology

• External server. The PostgreSQL server external to Cloud SQL that you want to replicate data from. It's also referred to as the source database or the external database server. It can be another Cloud SQL instance or any other database server, such as on-premises, Amazon Relational Database Service (RDS), and so on.

• Source representation instance. A mock of a Cloud SQL instance that represents the external server to the Cloud SQL replica. It's visible in the Google Cloud console and appears like a regular Cloud SQL instance, but it doesn't contain data, require configuration or maintenance, or affect billing.

• Cloud SQL replica. The Cloud SQL instance that replicates from the external server. Also known as the external primary read replica.

• Replication user account. The PostgreSQL user account on the external server with sufficient permissions to allow replication between the external server and the Cloud SQL replica.

• Managed import. The process of importing data directly from the external server to the Cloud SQL replica. In this situation, Cloud SQL connects to the external server using the replication user account and runs the data dump directly on the external server to import data to the Cloud SQL replica.

Set up a Google Cloud project

1. Sign in to your Google Cloud account. If you're new to Google Cloud, create an account to evaluate how our products perform in real-world scenarios. New customers also get $300 in free credits to run, test, and deploy workloads.

2. In the Google Cloud console, on the project selector page, select or create a Google Cloud project.

   Go to project selector

3. Make sure that billing is enabled for your Google Cloud project.

4. Enable the Cloud SQL Admin API.

   Enable the API

5. In the Google Cloud console, on the project selector page, select or create a Google Cloud project.

   Go to project selector

6. Make sure that billing is enabled for your Google Cloud project.

7. Enable the Cloud SQL Admin API.

   Enable the API

8. Make sure you have the Cloud SQL Admin, Storage Admin, and Compute Viewer roles on your user account.

   Go to the IAM page

Install the Google Cloud SDK

To configure replication, install Google Cloud SDK for your external server. You might want to install the SDK on your external server unless it's already installed elsewhere.

Set up the external server for replication

Cloud SQL supports continuous migrations from source databases to Cloud SQL destination databases.

Supported source databases for PostgreSQL include:

• Self-managed PostgreSQL (on premises or on any cloud VM that you fully control) 9.4, 9.5, 9.6, 10, 11, 12, 13, 14, 15, 16, and 17
• Amazon RDS 9.6.10+, 10.5+, 11.1+, 12, 13, 14, 15, 16, and 17
• Amazon Aurora 10.11+, 11.6+, 12.4+, 13.3+, 14.6+, 15.2+, 16, and 17
• Microsoft Azure Database for PostgreSQL Flexible Server 11 and later
• Cloud SQL 9.6, 10, 11, 12, 13, 14, 15, 16, and 17

Configuring your source requires configuring both the source instance and underlying source databases.

External server checklist

If the Cloud SQL replica is enabled with a private IP address because the outgoing private IP address isn't static, configure the external server's firewall to allow the internal IP range allocated for the private services access of the VPC network that the Cloud SQL replica uses as its private network.

The source database server's firewall must be configured to allow the entire internal IP range allocated for the private service connection of the VPC network that the Cloud SQL destination instance uses as the privateNetwork field of its ipConfiguration settings.

To find the internal IP range:

1. In the Google Cloud console, go to the VPC networks page.

   Go to the VPC networks page

2. Select the VPC network that you want to use.

3. Click the Private service connection tab.

Configure your source instance

To configure your source instance, follow these steps:

1. If your source instance does not include the postgres database, create it.
2. Install the pglogical package on the source instance.

3. Set the following parameters, as needed.

   1. If the source PostgreSQL instance is Amazon RDS, then include these parameters in a new parameter group and attach the parameter group to the instance.

   • If the source is Cloud SQL, set the cloudsql.logical_decoding and cloudsql.enable_pglogical flags to on.

     To enable flags in Cloud SQL, see Configuring database flags.

   • Set shared_preload_libraries to include pglogical by using the following command:

     ALTERSYSTEMSETshared_preload_libraries='pglogical';

   • Set wal_level to logical by using the following command:

     ALTERSYSTEMSETwal_level='logical';

     If the source PostgreSQL instance is Amazon RDS, to enable WAL logs at the logical level, set the rds.logical_replication parameter to 1.

   • Set wal_sender_timeout to 0 by using the following command:

     ALTERSYSTEMSETwal_sender_timeout=0;

     The value 0 disables the timeout mechanism that's used to terminate inactive replication connections.

   • Set max_replication_slots to the maximum number of replication slots that the source instance can support. Use the following command, after replacing MAX_REPLICATION_SLOTS with the number:

     ALTERSYSTEMSETmax_replication_slots=MAX_REPLICATION_SLOTS;

     Cloud SQL requires one slot for each database that's migrated. Specify at least the number of subscriptions expected to connect, with some reserves for table synchronization.

     For example, if the source instance has 5 databases and 2 migration jobs are created for the source, then the number of replication slots must be at least 5 * 2 = 10, in addition to the number of replication slots that you already use.

   • Set max_wal_senders to at least the same as max_replication_slots, in addition to the number of senders already used on your instance. Use the following command, replacing MAX_WAL_SENDERS with the total number of WAL sender processes running simultaneously:

     ALTERSYSTEMSETmax_wal_senders=MAX_WAL_SENDERS;

     For example, if the max_replication_slots parameter is set to 10, and you're already using 2 senders, then the number of WAL sender processes running at the same time would be 10 + 2 = 12.

   • Set max_worker_processes to at least the number of databases in the source instance, in addition to the number of worker processes already used on your instance. Use the following command, after replacing MAX_WORKER_PROCESSES with the total number:

     ALTERSYSTEMSETmax_worker_processes=MAX_WORKER_PROCESSES;

   2. If the source PostgreSQL instance is Microsoft Azure Database for PostgreSQL Flexible Server, then perform the following actions to support migrating data from an external server to a Cloud SQL instance:

      • Set shared_preload_libraries to include pglogical by using the following command:

        ALTERSYSTEMSETshared_preload_libraries='pglogical';

      • Set wal_level to logical. For more information, see Logical replication and logical decoding in Azure Database for PostgreSQL - Flexible Server.

      • Set max_replication_slots to the maximum number of replication slots that the source instance can support. Use the following command, after replacing MAX_REPLICATION_SLOTS with the number:

        ALTERSYSTEMSETmax_replication_slots=MAX_REPLICATION_SLOTS;

        Cloud SQL requires one slot for each database that's migrated. Specify at least the number of subscriptions expected to connect, with some reserves for table synchronization.

        For example, if the source instance has 5 databases and 2 migration jobs are created for the source, then the number of replication slots must be at least 5 * 2 = 10, in addition to the number of replication slots that you already use.

      • Set max_wal_senders to at least the same as max_replication_slots, in addition to the number of senders already used on your instance. Use the following command, replacing MAX_WAL_SENDERS with the total number of WAL sender processes running simultaneously:

        ALTERSYSTEMSETmax_wal_senders=MAX_WAL_SENDERS;

        For example, if the max_replication_slots parameter is set to 10, and you're already using 2 senders, then the number of WAL sender processes running at the same time would be 10 + 2 = 12.

      • Set max_worker_processes to at least the number of databases in the source instance, in addition to the number of worker processes already used on your instance. Use the following command, after replacing MAX_WORKER_PROCESSES with the total number:

        ALTERSYSTEMSETmax_worker_processes=MAX_WORKER_PROCESSES;

      • Set azure.extensions to include pglogical. For more information, see Server parameters in Azure Database for PostgreSQL - Flexible Server.

The parameters that you're setting in this step apply to a PostgreSQL database server that's running. You can also make these changes persistent by including them in the postgresql.conf file.

4. If the Cloud SQL replica will use a private IP, configure the external server's firewall to allow the internal IP range allocated for the private services access of the VPC network of the replica.
5. To apply the configuration changes, restart the source instance.

Enable replication delay monitoring for PostgreSQL versions preceding 9.6

If you're migrating from a PostgreSQL version lower than 9.6, then the replication delay metric isn't available by default. You can use one of three alternatives to track this metric and ensure minimal downtime when you promote the database:

• Option 1: Enable the Cloud SQL external server to track the replication delay by granting access to a specific query. Using a user with the SUPERUSER privilege, perform the following:

  1. Define the following function to allow the external server to query for the replication delay.

     CREATEORREPLACEFUNCTIONpg_stat_replication_user() RETURNSTABLE( pidinteger, usesysidoid, usernamename, application_nametext, client_addrinet, client_hostnametext, client_portinteger, backend_starttimestampwithtimezone, backend_xminxid, statetext, sent_locationpg_lsn, write_locationpg_lsn, flush_locationpg_lsn, replay_locationpg_lsn, sync_priorityinteger, sync_statetext ) LANGUAGESQL SECURITYDEFINER AS$$SELECT* FROMpg_catalog.pg_stat_replication;$$;

  2. Grant the EXECUTE permission to the user by running the following commands:

     1. REVOKE EXECUTE ON FUNCTION pg_stat_replication_user() FROM public;
     2. GRANT EXECUTE ON FUNCTION pg_stat_replication_user() to {replication_user};

• Option 2: Grant the SUPERUSER privilege directly to the user used to connect to the source instance. This allows the external server to read the replication delay directly.

• Option 3: Track the replication delay independently by using the following query:

  SELECTcurrent_timestamp,application_name, pg_xlog_location_diff(pg_current_xlog_location(),pg_stat_replication.sent_location)ASsent_location_lag, pg_xlog_location_diff(pg_current_xlog_location(),pg_stat_replication.write_location)ASwrite_location_lag, pg_xlog_location_diff(pg_current_xlog_location(),pg_stat_replication.flush_location)ASflush_location_lag, pg_xlog_location_diff(pg_current_xlog_location(),pg_stat_replication.replay_location)ASreplay_location_lag FROMpg_stat_replication WHEREapplication_namelike'cloudsql%';

  In this option, Cloud SQL doesn't reflect the replication delay metric in the graphs or API responses.

Configure your source databases

The Cloud SQL external server migrates all databases under your source instance other than the following:

• For on-premise sources: template databases template0 and template1
• For Amazon RDS sources: template0, template1, and rdsadmin
• For Cloud SQL sources: template databases template0 and template1

Do the following on each database in your source instance that isn't included in the preceding list:

1. To install the pglogical extension, run the following command on every database on your source instance:

   CREATEEXTENSIONIFNOTEXISTSpglogical

   For tables that don't have primary keys, Cloud SQL supports the migration of the initial snapshot and INSERT statements during the change data capture (CDC) phase. Migrate UPDATE and DELETE statements manually.

2. Connect to the instance and run the following commands to set privileges for the user on each of the migrated databases, as well as the default postgres database.

   The user that you use to connect to the source instance is configured as the user in the Connection Profiles page. You can create a new user or reuse an existing one.

   • On all schemas on each database to migrate, except for the information schema and schemas starting with pg_, run the following command:

     GRANTUSAGEonSCHEMAschematousername; GRANTSELECTonALLSEQUENCESinSCHEMAschematousername; GRANTSELECTonALLTABLESinSCHEMAschematousername;

   • On each database to migrate, run the following command:

     GRANTUSAGEonSCHEMApglogicaltoPUBLIC;

   • To get replication information from source databases, run the following command on all databases:

     GRANTSELECTonALLTABLESinSCHEMApglogicaltousername;

   • If your source is Amazon RDS, then run the following command:

     GRANTrds_replicationtousername;

     If your source is Microsoft Azure Database for PostgreSQL version earlier than 14, then run the following command:

     GRANTSELECTonALLTABLESinSCHEMApglogicaltousername;

     You can ignore an error that appears on the pglogical.node interface.

     If your source is Microsoft Azure Database for PostgreSQL version 14 and later, then run the following command:

     GRANTpg_read_all_datatousername;

     Otherwise, run the following command:

     ALTERUSERusernamewithREPLICATION;

Set up a source representation instance

The source representation instance references the external server. It contains only the request data from the external server. Create the request data and use it in a curl command that creates the source representation instance in Cloud SQL.

Create the request data

The request data contains basic information about your external server in JSON format. The request data can be configured for a Cloud SQL replica on a public or private network and should contain this information.

{"name":"SOURCE_NAME", "region":"REGION", "databaseVersion":"DATABASE_VERSION", "onPremisesConfiguration":{"selectedObjects":"SELECTED_OBJECTS", "hostPort":"SOURCE_HOST", "username":"USERNAME", "password":"PASSWORD", "caCertificate":"SOURCE_CERT", "clientCertificate":"CLIENT_CERT", "clientKey":"CLIENT_KEY"}}

//exampleofsource.jsonforexternalserverthat//-initiatesreplicationfromaCloudSQLmanagedimport//-doesn'tuseSSL/TLS{"name":"cloudsql-source-instance","region":"us-central1","databaseVersion":"POSTGRES_9_6","onPremisesConfiguration":{"selectedObjects":[{"database":"db1"},{"database":"db2"}],"hostPort":"192.0.2.0:3306","username":"replicationUser","password":"486#@%*@"}}

Create a source representation instance

Before you start this step, create a JSON file that contains your source request data.

Then, to create the source representation instance in Cloud SQL, open a terminal and run the following commands:

gcloudauthlogin ACCESS_TOKEN="$(gcloudauthprint-access-token)"curl--header"Authorization: Bearer ${ACCESS_TOKEN}"\--header'Content-Type: application/json'\--data@JSON_PATH\-XPOST\https://sqladmin.googleapis.com/sql/v1beta4/projects/PROJECT-ID/instances

gcloudauthloginACCESS_TOKEN="$(gcloud auth print-access-token)"curl--header "Authorization: Bearer ${ACCESS_TOKEN}" \--header 'Content-Type: application/json' \--data @./source.json \-XPOST\https://sqladmin.googleapis.com/sql/v1beta4/projects/MyProject/instances

Update a source representation instance

If you update the request data from the external server, you can update the existing source representation instance to use the modified values.

Modify the request data

Update the request data to include any fields that have changed. This includes the hostPort, username, password, caCertificate, clientCertificate, and clientKey fields. After updating the request data, use it in a curl command to update the instance in Cloud SQL.

The following example shows updating the username and password fields with a different username and password:

{"name":"SOURCE_NAME", "region":"REGION", "databaseVersion":"DATABASE_VERSION", "onPremisesConfiguration":{"selectedObjects":"SELECTED_OBJECTS", "username":"NEW_USERNAME", "password":"NEW_PASSWORD"}}

//exampleofsource.jsonforexternalserverthat//-initiatesreplicationfromaCloudSQLmanagedimport//-doesn'tuseSSL/TLS{"name":"cloudsql-source-instance","region":"us-central1","databaseVersion":"POSTGRES_9_6","onPremisesConfiguration":{"selectedObjects":[{"database":"db1"},{"database":"db3"}],"username":"newReplicationUser","password":"525#@%*@"}}

Modify a source representation instance

Before you start this step, create a JSON file that contains your modified request data.

Then, to modify the source representation instance in Cloud SQL, open a terminal and run the following commands:

gcloudauthlogin ACCESS_TOKEN="$(gcloudauthprint-access-token)"curl--header"Authorization: Bearer ${ACCESS_TOKEN}"\--header'Content-Type: application/json'\--data@JSON_PATH\-XPATCH\https://sqladmin.googleapis.com/sql/v1beta4/projects/PROJECT-ID/instances/SOURCE_NAME

gcloudauthloginACCESS_TOKEN="$(gcloud auth print-access-token)"curl--header "Authorization: Bearer ${ACCESS_TOKEN}" \--header 'Content-Type: application/json' \--data @./source.json \-XPATCH\https://sqladmin.googleapis.com/sql/v1beta4/projects/MyProject/instances/cloudsql-source-instance

Set up a Cloud SQL replica

The Cloud SQL replica eventually contains the data from the external server. In this step, you create the request data and use it in a curl command that creates the Cloud SQL replica in Cloud SQL.

Create the request data

The request data contains basic information about your external server and Cloud SQL replica in JSON format. The request data can be configured for a Cloud SQL replica on a public or private network and should contain this information:

{"settings":{"tier":"TIER", "dataDiskSizeGb":"DISK_SIZE", "ipConfiguration":{"ipv4Enabled":"PUBLIC_IP_STATUS", "privateNetwork":"projects/PROJECT_ID/global/networks/NETWORK_NAME"}, "availabilityType":"AVAILABILITY_TYPE"}, "masterInstanceName":"SOURCE_REPRESENTATION_INSTANCE_NAME", "region":"SOURCE_REGION", "databaseVersion":"DATABASE_VERSION", "name":"REPLICA_NAME"}

{"settings":{"tier":"db-custom-4-15360", "dataDiskSizeGb":"100"}, "masterInstanceName":"source-instance", "region":"us-central1", "databaseVersion":"POSTGRES_16", "name":"replica-instance"}

Create the Cloud SQL replica

Before you start this step, create a JSON file that contains your replica request data. Then, to create a Cloud SQL replica, open a Cloud Shell terminal and run these commands:

gcloudauthlogin ACCESS_TOKEN="$(gcloudauthprint-access-token)"curl--header"Authorization: Bearer ${ACCESS_TOKEN}"\--header'Content-Type: application/json'\--data@JSON_PATH\-XPOST\https://sqladmin.googleapis.com/sql/v1beta4/projects/PROJECT-ID/instances

gcloudauthlogin ACCESS_TOKEN="$(gcloudauthprint-access-token)"curl--header"Authorization: Bearer ${ACCESS_TOKEN}"\--header'Content-Type: application/json'\--data@./replica.json\-XPOST\https://sqladmin.googleapis.com/sql/v1beta4/projects/MyProject/instances

Verify your setup

To ensure your instances were set up correctly, go to the Cloud SQL Instances page.

You should see your source representation instance and the Cloud SQL replica, in a listing similar to the following:

Also make sure that you have the cloudsql.instances.migrate permission on the Cloud SQL replica. This permission is included in the cloudsql.admin or cloudsql.editorIAM roles.

Add users to the Cloud SQL replica

You cannot import database user accounts from the external server, but you can create them on a Cloud SQL replica. Do this before you replicate from the external server.

Get the Cloud SQL replica's outgoing IP address

You can use the outgoing IP address of the Cloud SQL replica to create a secure connection between the external server and the Cloud SQL replica. You won't be charged for this IP address.

gcloudsqlinstancesdescribeREPLICA_NAME--format="default(ipAddresses)"

Allow incoming connections on the external server

The Cloud SQL replica needs to connect to the external server for replication to succeed. You must configure the network firewall for your external server to accept connections from the Cloud SQL replica's outgoing IP address if the following conditions apply:

• The external server is behind a firewall or some other network restriction.
• Your Cloud SQL replica is using a public IP.

To connect to the Cloud SQL replica, you use the replica's primary IP address. This IP address is displayed in the Google Cloud console.

Update the source representation instance to allow replication to the Cloud SQL replica

After you set up the source representation instance for the Cloud SQL replica, you might need to update the source representation instance. For example, these scenarios require an update to your configurations:

• The host, port, or IP of the external server changes.
• You want to use a different PostgreSQL replication user.
• The password of the PostgreSQL replication user changes.
• The SSL certificates used to securely connect to the external server change.

Seed the Cloud SQL replica

For the initial loading of data from the external server into the Cloud SQL replica, use a managed import. It uses a service that extracts data from the external server and imports it into the Cloud SQL instance directly. For more information, see Using a managed import to set up replication from external databases.

Monitor replication

When the Cloud SQL replica finishes the initial data load, it connects to the external server and applies all updates that were made after the export operation. Confirm your replication status.

It's important to check the replication status before promoting the replica to a standalone instance. If the replication process isn't successfully completed, a promoted replica doesn't have all the changes from your external server.

If replication delay is not trending toward 0, take steps to address it. You might want to check these metrics: /postgresql/external_sync/initial_sync_complete, postgresql/external_sync/max_replica_byte_lag, and database/replication/state. View the list of Cloud SQL metrics.

If you want to migrate a subset of databases from the source representation instance to the destination Cloud SQL instance, then check the following per-database metrics:

After the Cloud SQL replica has caught up with the external server and there's no replication delay on the Cloud SQL replica, connect to your database. Run the appropriate database commands to make sure that the contents are as expected when compared with the external server. Retain your external server until the necessary validations are done.

Set up a cascading replica

After migration, you can create cascading read replicas under your Cloud SQL replica before promoting the Cloud SQL replica.

To create a cascading replica, run the following commands:

Promote the replica

Promote your replica by following these steps:

1. Promote the replica to a primary instance.
2. Add read replicas to your instance.
3. Optional: Configure your instance for high availability (HA). To prevent additional downtime, you can enable HA while setting up a replica by setting AVAILABILITY_TYPE to REGIONAL.

Limitations

• If you install extensions on your external source databases that Cloud SQL doesn't support, then when you migrate the databases to a destination instance, Cloud SQL won't migrate these extensions. To ensure a smooth migration, verify that no objects or applications reference the extensions. Before proceeding with the migration, we recommend removing the extensions along with any references from the source databases.

  For more information about the extensions that Cloud SQL supports, see Configure PostgreSQL extensions.

• If you install the pg_cron extension on your external source databases, then when you migrate the databases to a destination instance, Cloud SQL doesn't migrate the extension or any cron settings associated with the extension. After you migrate the databases and promote the replica, Google recommends that you re-enable the pg_cron extension on each migrated database.

• You can't migrate data from version 11 of a source Microsoft Azure server to version 11 of a destination Cloud SQL for PostgreSQL instance. To resolve this, migrate the data to a Cloud SQL instance that has a version of 12 or later.

What's next

• Learn how to use a managed import to set up replication from external databases.