Replicating databases and account objects across multiple accounts¶

Step 1. Disable replication for a replication enabled database¶

Execute the SYSTEM$DISABLE_DATABASE_REPLICATION function to disable replication for a primary database, along with any secondary databases linked to it, in order to add it to a replication or failover group.

Execute the following SQL statement from the source account with the primary database:

SELECT SYSTEM$DISABLE_DATABASE_REPLICATION('mydb');

Step 2. Add the database to a primary failover group and create a secondary failover group¶

Once you have successfully disabled replication for a database, you can add the primary database to a failover group in the source account.

Then create a secondary failover group in the target account. When the secondary failover group is refreshed in the target account, the previously secondary database will automatically be added as a member of the secondary failover group and refreshed with the changes from the primary database.

For more details on creating primary and secondary failover groups, see Workflow.

Examples¶

Execute the following SQL statements in your preferred Snowflake client to enable account and database object replication and failover, and refresh objects.

Prerequisite: Enable replication for accounts in the organization¶

The organization administrator must enable replication for the source and target accounts.

To enable replication for accounts, an organization administrator uses the SYSTEM$GLOBAL_ACCOUNT_SET_PARAMETER function to set the ENABLE_ACCOUNT_DATABASE_REPLICATION parameter to true.

As an organization administrator, enable replication for each source and target account in your organization.

-- View the list of the accounts in your organization
-- Note the organization name and account name for each account for which you are enabling replication
SHOW ACCOUNTS;

-- Enable replication by executing this statement for each source and target account in your organization
SELECT SYSTEM$GLOBAL_ACCOUNT_SET_PARAMETER('<organization_name>.<account_name>', 'ENABLE_ACCOUNT_DATABASE_REPLICATION', 'true');

Though the SYSTEM$GLOBAL_ACCOUNT_SET_PARAMETER function supports the legacy account locator identifier, it causes unexpected results when an organization has multiple accounts that share the same locator (in different regions).

Step 1: Create a role with the CREATE FAILOVER GROUP privilege in the source account — Optional¶

Create a role and grant it the CREATE FAILOVER GROUP privilege. This step is optional. If you have already created this role, skip to Step 3: Create a primary failover group in a source account.

USE ROLE ACCOUNTADMIN;

CREATE ROLE myrole;

GRANT CREATE FAILOVER GROUP ON ACCOUNT
    TO ROLE myrole;

Step 2: Identify accounts enabled for replication and group membership¶

Before creating a primary failover group, identify the accounts enabled for replication and the existing failover and replication groups.

SHOW REPLICATION ACCOUNTS;

+------------------+-------------------------------+--------------+-----------------+-----------------+-------------------+--------------+
| snowflake_region | created_on                    | account_name | account_locator | comment         | organization_name | is_org_admin |
+------------------+-------------------------------+--------------+-----------------+-----------------+-------------------+--------------+
| AWS_US_WEST_2    | 2020-07-15 21:59:25.455 -0800 | myaccount1   | myacctlocator1  |                 | myorg             | true         |
+------------------+-------------------------------+--------------+-----------------+-----------------+-------------------+--------------+
| AWS_US_EAST_1    | 2020-07-23 14:12:23.573 -0800 | myaccount2   | myacctlocator2  |                 | myorg             | false        |
+------------------+-------------------------------+--------------+-----------------+-----------------+-------------------+--------------+
| AWS_US_EAST_2    | 2020-07-25 19:25:04.412 -0800 | myaccount3   | myacctlocator3  |                 | myorg             | false        |
+------------------+-------------------------------+--------------+-----------------+-----------------+-------------------+--------------+

SHOW FAILOVER GROUPS;

SHOW DATABASES IN FAILOVER GROUP myfg;

SHOW SHARES IN FAILOVER GROUP myfg;

Step 3: Create a primary failover group in a source account¶

Create a primary failover group and enable the replication and failover of specific objects from the current (source) account to one or more target accounts in the same organization.

You can create a replication or failover group using Snowsight or SQL.

• Create a replication or failover group using Snowsight

• Create a failover group using SQL

USE ROLE myrole;

CREATE FAILOVER GROUP myfg
    OBJECT_TYPES = USERS, ROLES, WAREHOUSES, RESOURCE MONITORS, DATABASES, INTEGRATIONS, NETWORK POLICIES
    ALLOWED_DATABASES = db1, db2
    ALLOWED_INTEGRATION_TYPES = API INTEGRATIONS
    ALLOWED_ACCOUNTS = myorg.myaccount2
    REPLICATION_SCHEDULE = '10 MINUTE';

Step 4: Create a role with the CREATE FAILOVER GROUP privilege in the target account — Optional¶

Create a role in the target account and grant it the CREATE FAILOVER GROUP privilege. This step is optional. If you have already created this role, skip to Step 5: Create a secondary failover group in the target account.

USE ROLE ACCOUNTADMIN;

CREATE ROLE myrole;

GRANT CREATE FAILOVER GROUP ON ACCOUNT
    TO ROLE myrole;

Step 5: Create a secondary failover group in the target account¶

Create a secondary failover group in the target account as a replica of the primary failover group in the source account.

Execute a CREATE FAILOVER GROUP … AS REPLICA OF statement in each target account for which you enabled replication in Step 3: Create a primary failover group in a source account (in this topic).

Executed from each target account:

USE ROLE myrole;

CREATE FAILOVER GROUP myfg
  AS REPLICA OF myorg.myaccount1.myfg;

Step 6. Refresh a secondary failover group in the target account manually — Optional¶

To manually refresh the objects in a target account, execute the ALTER FAILOVER GROUP … REFRESH command.

As a best practice, we recommend scheduling your secondary refreshes by setting the REPLICATION_SCHEDULE parameter using CREATE FAILOVER GROUP or ALTER FAILOVER GROUP.

Step 7. Grant the FAILOVER privilege on failover group to role — Optional¶

To execute the command to fail over a secondary failover group in a target account, you must use a role with the FAILOVER privilege on the failover group. The FAILOVER privilege is currently not replicated and must be granted in each source and target account.

For more information, see Replication of roles and grants.

For example, to grant the FAILOVER privilege to role my_failover_role on failover group my_fg, execute the following statement in the target account using a role with the OWNERSHIP privilege on the group:

GRANT FAILOVER ON FAILOVER GROUP myfg TO ROLE my_failover_role;

For instructions on creating a custom role with a specified set of privileges, see Creating custom roles.

For general information about roles and privilege grants for performing SQL actions on securable objects, see Overview of Access Control.

Specify schemas to replicate or skip¶

You can explicitly specify schemas to replicate or skip in a database in a failover group using the optional REPLICABLE_WITH_FAILOVER_GROUPS parameter.

ALTER DATABASE <name> SET REPLICABLE_WITH_FAILOVER_GROUPS = { 'YES' | 'NO' }
ALTER DATABASE <name> UNSET REPLICABLE_WITH_FAILOVER_GROUPS

ALTER SCHEMA <name> SET REPLICABLE_WITH_FAILOVER_GROUPS = { 'YES' | 'NO' }
ALTER SCHEMA <name> UNSET REPLICABLE_WITH_FAILOVER_GROUPS

USE ROLE ACCOUNTADMIN;

GRANT REPLICATE ON ACCOUNT TO ROLE replicationadmin;
GRANT USAGE ON DATABASE db1 TO ROLE replicationadmin;
GRANT USAGE ON SCHEMA db1.sch1 TO ROLE replicationadmin;

USE ROLE replicationadmin;

ALTER DATABASE db1 SET REPLICABLE_WITH_FAILOVER_GROUPS = 'NO';
ALTER SCHEMA sch1 SET REPLICABLE_WITH_FAILOVER_GROUPS = 'YES';

USE ROLE replicationadmin;

ALTER DATABASE db2 SET REPLICABLE_WITH_FAILOVER_GROUPS = 'YES';
ALTER SCHEMA sch2 SET REPLICABLE_WITH_FAILOVER_GROUPS = 'NO';

Refresh of schemas with REPLICABLE_WITH_FAILOVER_GROUPS set in target accounts¶

During a database refresh:

• Schemas with REPLICABLE_WITH_FAILOVER_GROUPS set to 'YES' are replicated from the source account to the target account.

• Schemas with REPLICABLE_WITH_FAILOVER_GROUPS set to 'NO' are not replicated, except in the following two scenarios:

  • The target schema is a replica of the source account schema. In this case, the target schema is always synchronized with its source schema.

  • The target schema has a name conflict with the source account schema. In this situation, the replication job fails due to the name conflict.

List databases and schemas with REPLICABLE_WITH_FAILOVER_GROUPS set in your account¶

You can list the values set for the REPLICABLE_WITH_FAILOVER_GROUPS parameter in the current account by querying the ACCOUNT_USAGE and INFORMATION_SCHEMA views.

USE ROLE replicationadmin;

SELECT database_name, replicable_with_failover_groups
  FROM db1.INFORMATION_SCHEMA.DATABASES;

+---------------+---------------------------------+
| DATABASE_NAME | REPLICABLE_WITH_FAILOVER_GROUPS |
+---------------+---------------------------------+
| DB1           | NO                              |
| DB2           | YES                             |
| DB3           | UNSET                           |
+---------------+---------------------------------+

SELECT schema_name, catalog_name, replicable_with_failover_groups
  FROM db1.INFORMATION_SCHEMA.SCHEMATA ORDER BY catalog_name;

+--------------------+--------------+---------------------------------+
| SCHEMA_NAME        | CATALOG_NAME | REPLICABLE_WITH_FAILOVER_GROUPS |
+--------------------+--------------+---------------------------------+
| PUBLIC             | DB1          | NO                              |
| SCH1               | DB1          | YES                             |
| SCH2               | DB1          | NO                              |
| SCH3               | DB1          | NO                              |
| INFORMATION_SCHEMA | DB1          | UNSET                           |
+--------------------+--------------+---------------------------------+

USE ROLE replicationadmin;

SELECT schema_name, catalog_name, replicable_with_failover_groups
  FROM db2.INFORMATION_SCHEMA.SCHEMATA
  ORDER BY catalog_name;

+--------------------+--------------+---------------------------------+
| SCHEMA_NAME        | CATALOG_NAME | REPLICABLE_WITH_FAILOVER_GROUPS |
+--------------------+--------------+---------------------------------+
| PUBLIC             | DB2          | YES                             |
| SCH1               | DB2          | YES                             |
| SCH2               | DB2          | NO                              |
| SCH3               | DB2          | YES                             |
| INFORMATION_SCHEMA | DB2          | UNSET                           |
+--------------------+--------------+---------------------------------+

Use SYSTEM$LINK_ACCOUNT_OBJECTS_BY_NAME() to apply global IDs¶

You can prevent the loss of some object types by linking matching objects with the same name in the source and target accounts. The SYSTEM$LINK_ACCOUNT_OBJECTS_BY_NAME function adds a global identifier to account objects in the target account.

Apply global identifiers to account objects in the target account of the types included in the object_types list for failover group myfg:

Execute the following SQL statement using the ACCOUNTADMIN role:

SELECT SYSTEM$LINK_ACCOUNT_OBJECTS_BY_NAME('myfg');

Initial replication of users and roles¶

The behavior of the initial refresh operation for USERS and ROLES object types can vary depending on whether or not there are matching objects with the same name in the target account.

• If there are existing users in the target account with the same name as users in the source account, the initial refresh operation fails and describes the two options you have to continue:

  • Force the refresh operation and allow any existing users in the target account to be dropped. The users in the source account will be replicated to the target account.

    To force a refresh for a group, use the FORCE parameter for the refresh command. For example, to force the refresh of a failover group, execute the following command:

    ALTER FAILOVER GROUP <fg_name> REFRESH FORCE;
    

    Copy

  • Link the account objects by name. The SYSTEM$LINK_ACCOUNT_OBJECTS_BY_NAME function links users with the same name in both the target account and the source account. Users in the target account that are linked are not deleted.

    To link account objects by name, execute the following command:

    SELECT SYSTEM$LINK_ACCOUNT_OBJECTS_BY_NAME('<rg_name>');
    

    Copy

    Note

    Any user in the target account that does not have a matching user in the source account with the same name is dropped.

• If there are no users in the target account with names matching users in the source account, the initial refresh operation in the target account drops all users. This can result in the following data and metadata loss:

  • If USERS are included in the OBJECT_TYPES list for a replication or failover group:

    • Worksheets are lost.

    • Query history is lost.

  • If USERS are included in the OBJECT_TYPES list, but ROLES is not:

    • Privilege grants to users are lost.

  • If ROLES are included in the OBJECT_TYPES list:

    • Privilege grants to share objects are lost.

To avoid dropping users or roles in the target account:

1. In the source account, manually recreate any users or roles that exist only in the target account before the initial replication.

2. In the target account, link matching objects with the same name in both accounts using the SYSTEM$LINK_ACCOUNT_OBJECTS_BY_NAME function.

Amazon S3¶

The configuration process depends on how you set up event notifications. For example, suppose you have an auto-ingest pipe that relies on an Amazon Simple Notification Service (SNS) topic to publish messages about the Snowflake stage location.

When you replicate the pipe to a target account, Snowflake automatically creates a new Amazon Simple Queue Service (SQS) queue. You must subscribe this SQS queue for your target account to the SNS topic to get notifications about the stage location.

• If you use Amazon S3 Event Notifications with Amazon Simple Queue Service (SQS), follow the instructions in Step 4: Configure event notifications.

  Important

  To ensure that the pipe has not missed any notifications, you should refresh the pipe after switching to the new SQS queue.

  You can also migrate from SQS to SNS. For more information, see Migrate to Amazon Simple Notification Service (SNS).

• If you use Amazon Simple Notification Service (SNS), see Subscribing the Snowflake SQS Queue to your SNS topic.

• If you use Amazon EventBridge, see Option 3: Setting up Amazon EventBridge to automate Snowpipe.

Microsoft Azure Blob Storage¶

A pipe that automatically loads data from files located on a stage in Microsoft Azure blob storage requires an Event Grid subscription, storage queue, and a notification integration bound to the storage queue. A secondary pipe in a target account needs a separate Event Grid, storage queue, and notification integration bound to the storage queue. The Event Grid in both source and target accounts must be configured as endpoints for the same Azure Storage source.

See the diagram below for configuration details:

Create a new Event Grid subscription and storage queue. Then, create a new notification integration in the target account and grant Snowflake access to your storage queue. For instructions, see Configuring Automation With Azure Event Grid.

External stage for Google Cloud Storage¶

A pipe that automatically loads data from files located in Google Cloud Storage requires a Google Pub/Sub subscription and a notification integration that references that subscription. Each replicated pipe in a target account also requires a Google Pub/Sub subscription and a notification integration that references that subscription. The Pub/Sub subscription in each source and target account must be subscribed to the same Pub/Sub Topic that receives notifications from the Google Cloud Storage source.

See the diagram below for configuration details:

Create a new subscription to your Pub/Sub topic and a new notification integration in your target account.

Then, grant Snowflake access to the Pub/Sub subscription. For instructions, see Configuring Automation Using GCS Pub/Sub.

Example¶

In the examples below, the database mydb is included in the failover group myfg. The database mydb contains the table mytable.

Modify a replication or failover group in a source account using Snowsight¶

To perform these actions, you must be signed in to the source account. If you are not signed in, the Status column displays a sign in message instead of the refresh status.

1. Sign in to Snowsight and navigate to Admin » Accounts.

2. Select Replication, then select Groups.

3. Locate the replication or failover group you want to edit, and select the More menu (…) in the last column of the row.

4. Select Edit.

5. To change the group name, enter a new name in the Group name box that meets the following requirements:

   • Must start with an alphabetic character and cannot contain spaces or special characters unless the identifier string is enclosed in double quotes (for example, “My object”). Identifiers enclosed in double quotes are also case-sensitive.

     For more information, see Identifier requirements.

   • Names for failover groups and replication groups in an account must be unique.

6. Choose Edit objects to add or remove share and account objects.

   Note

   Account objects can only be added to one replication or failover group. If a replication or failover group with any account objects already exists in your account, you can’t select those objects.

7. Choose Select databases to add or remove database objects.

8. Select the Replication frequency to change the replication schedule for a group.

9. Select Save to update the group.

   If saving the changes to the group is unsuccessful, refer to Troubleshoot issues with creating and editing replication groups using Snowsight for common errors and how to resolve them.

Modify a replication or failover group in a source account using SQL¶

You can modify a replication or failover group properties using the ALTER REPLICATION GROUP or ALTER FAILOVER GROUP command.

Pause or resume a replication schedule in a target account using Snowsight¶

To pause or resume a replication schedule, you must be signed in to the target account.

1. Sign in to Snowsight and navigate to Admin » Accounts.

2. Select Replication, then select Groups.

3. Locate the replication or failover group you want to edit, and select the More menu (…) in the last column of the row.

4. Select Pause or Resume.

Pause or resume a replication schedule in a target account using SQL¶

You can pause or resume a replication schedule in a target account using the ALTER REPLICATION GROUP or ALTER FAILOVER GROUP command. To pause, specify the SUSPEND parameter. To resume, specify the RESUME parameter.

Drop a primary replication or failover group using SQL¶

A primary replication or failover group can only be dropped after all the replicas of the group (that is, secondary replication or failover groups) have been dropped. Alternatively, you can promote a secondary failover group to serve as the primary failover group, then drop the former primary failover group.

Note that only the group owner can drop the group.

Drop a replication or failover group using Snowsight¶

You can delete a primary replication or failover group and any linked secondary groups.

1. Sign in to Snowsight and navigate to Admin » Accounts.

2. Select Replication, select Groups.

3. Locate the replication or failover group you want to delete. Select the More menu (…) in the last column of the row.

4. Select Drop, then select Drop group.

You cannot add a database to a group¶

Database '<database_name>' is already configured to replicate to
account '<account_name>' by replication group '<group_name>'.

Cannot directly add previously replicated object '<database_name>' to a
replication group. Please use the provided system functions to convert
this object first.

You cannot add a share to a group¶

Share '<share_name>' is already configured to replicate to
account '<account_name>' by replication group '<group_name>'.