Use point-in-time recovery (PITR)

This page describes how to use point-in-time recovery (PITR) to restore your primary Cloud SQL instance.

To learn more about PITR, see Point-in-time recovery (PITR).

Log storage for PITR

On May 31, 2024, we launched storing transaction logs for PITR in Cloud Storage. Since this launch, the following conditions apply:

  • All Cloud SQL instances that you create with PITR enabled before this date store the transaction logs that are used for PITR on disk.

  • All Cloud SQL instances created with PITR enabled on or after this date store these logs in Cloud Storage.

Transaction logs update regularly and use storage space. Cloud SQL deletes transaction logs with their associated automatic backups automatically. This happens after the value set for thetransactionLogRetentionDays parameter is met. For more information about this parameter, see Logs and disk usage.

For instances that store transaction logs only on disk, you can configure Cloud SQL to store the logs on Cloud Storage by first deactivating and then re-enabling PITR. You can't move logs from Cloud Storage back to disk.

To ensure that logs for your instance are stored in Cloud Storage instead of on disk, complete the following actions:

Logs and disk usage

For instances having transaction logs stored in Cloud Storage, the logs are stored in the same region as the primary instance. This log storage (up to 35 days for Cloud SQL Enterprise Plus edition and seven days for Cloud SQL Enterprise edition, the maximum length for PITR) generates no additional cost per instance.

Cloud SQL generates transaction logs regularly and these logs use storage space. Cloud SQL deletes the transaction logs automatically with their associated automatic backups. This happens after the value set for the transactionLogRetentionDays parameter is met. This parameter specifies the number of days that Cloud SQL retains transaction logs. For Cloud SQL Enterprise Plus edition, you can set the number of days of retained transaction logs from 1 to 35, and for Cloud SQL Enterprise edition, you can set this value from 1 to 7.

To find out how much disk the transaction logs use, check the bytes_used_by_data_type metric for the instance. The value for the data type returns the size of the transaction logs on the disk. For instances that store transaction logs used for PITR on disk, Cloud SQL purges data from the disk daily to meet the transactionLogRetentionDays PITR setting, as described in Automated backup retention.

If your instance has PITR enabled, and if the size of your transaction logs on disk is causing an issue for your instance:

  • You can deactivate PITR and re-enable it to ensure that logs are stored in Cloud Storage in the same region as the instance. However, any existing logs are deleted, so you can't perform a point-in-time restore earlier than the time that you re-enabled PITR.
  • You can increase the instance storage size, but the transaction log size increase in disk usage might be temporary.
  • We recommend enabling automatic storage increases to avoid unexpected storage issues. This recommendation applies only if your instance has PITR enabled and your logs are stored on disk.
  • If you want to delete logs and recover storage, then you can deactivate PITR without re-enabling it. However, decreasing the storage used doesn't shrink the size of the disk provisioned for the instance.

  • Logs are purged once daily, not continuously. Setting the log retention to two days means that at least two days of logs, and at most three days of logs, are retained. We recommend setting the number of backups to one more than the days of log retention.

    For example, if you specify 7 for the value of the transactionLogRetentionDays parameter, then for the backupRetentionSettings parameter, set the number of retainedBackups to 8.

For more information about PITR, see Point-in-time recovery (PITR).

Database recovery model for PITR

When you enable PITR on an instance, Cloud SQL automatically sets the recovery model of the existing and subsequent databases to the full recovery model.

For more information about SQL Server recovery models, see the Microsoft documentation.

Enable PITR

When you create a new instance in the Google Cloud console, the Automated backups setting is automatically enabled.

The following procedure enables PITR on an existing primary instance.

Console

  1. In the Google Cloud console, go to the Cloud SQL Instances page.

    Go to Cloud SQL Instances

  2. Open the more actions menu More actions icon. for the instance you want to enable PITR on and click Edit.
  3. Under Customize your instance, expand the Data Protection section.
  4. Select the Enable point-in-time recovery checkbox.
  5. In the Days of logs field, enter the number of days to retain logs, from 1-35 for Cloud SQL Enterprise Plus edition, or 1-7 for Cloud SQL Enterprise edition.
  6. Click Save.

gcloud

  1. Display the instance overview: 
    gcloud sql instances describe INSTANCE_NAME
  2. If you see enabled: false in the backupConfiguration section, enable scheduled backups: 
    gcloud sql instances patch INSTANCE_NAME \
--backup-start-time=HH:MM

    Specify the backup-start-time parameter using 24-hour time in UTC±00 time zone.

  3. Enable PITR: 
    gcloud sql instances patch INSTANCE_NAME \
--enable-point-in-time-recovery

    If you're enabling PITR on a primary instance, you can also configure the number of days for which you want to retain transaction logs by adding the following parameter:

    --retained-transaction-log-days=RETAINED_TRANSACTION_LOG_DAYS
  4. Confirm your change: 
    gcloud sql instances describe INSTANCE_NAME

    In the backupConfiguration section, you see pointInTimeRecoveryEnabled: true if the change was successful.

REST v1

Before using any of the request data, make the following replacements:

  • PROJECT_ID: the ID or project number of the Google Cloud project that contains the instance
  • INSTANCE_NAME: the name of the primary or read replica instance that you're configuring for high availability
  • START_TIME: the time (in hours and minutes)

HTTP method and URL: 

PATCH https://sqladmin.googleapis.com/v1/projects/PROJECT_ID/instances/INSTANCE_NAME

Request JSON body: 

{
  "settings":
  {
    "backupConfiguration":
    {
      "startTime": "START_TIME",
      "enabled": true,
      "pointInTimeRecoveryEnabled": true
    }
  }
}

To send your request, expand one of these options:

You should receive a JSON response similar to the following:

REST v1beta4

Before using any of the request data, make the following replacements:

  • PROJECT_ID: the ID or project number of the Google Cloud project that contains the instance
  • INSTANCE_NAME: the name of the primary or read replica instance that you're configuring for high availability
  • START_TIME: the time (in hours and minutes)

HTTP method and URL: 

PATCH https://sqladmin.googleapis.com/sql/v1beta4/projects/PROJECT_ID/instances/INSTANCE_NAME

Request JSON body: 

{
  "settings":
  {
    "backupConfiguration":
    {
      "startTime": "START_TIME",
      "enabled": true,
      "pointInTimeRecoveryEnabled": true
    }
  }
}

To send your request, expand one of these options:

You should receive a JSON response similar to the following:

Perform PITR

Console

  1. In the Google Cloud console, go to the Cloud SQL Instances page.

    Go to Cloud SQL Instances

  2. Open the more actions menu More actions icon. for the instance you want to recover and click Create clone.
  3. Optionally, on the Create a clone page, update the ID of the new clone.
  4. Select Clone from an earlier point in time.
  5. Enter a PITR time.
  6. Select All databases, or specify a database name.
    If you specify a database name, only one may be selected. By default, PITR is for all databases.
  7. Click Create clone.

gcloud

Create a clone using PITR.

Replace the following:

  • SOURCE_INSTANCE_NAME - Name of the instance you're restoring from.
  • NEW_INSTANCE_NAME - Name for the clone.
  • TIMESTAMP - UTC timezone for the source instance in RFC 3339 format. For example, 2012-11-15T16:19:00.094Z.
gcloud sql instances clone SOURCE_INSTANCE_NAME \
NEW_INSTANCE_NAME \
--point-in-time 'TIMESTAMP'

REST v1

Before using any of the request data, make the following replacements:

  • project-id: The project ID
  • target-instance-id: The target instance ID
  • source-instance-id: The source instance ID
  • restore-timestamp The point-in-time to restore up to

In the JSON request, you optionally can specify up to one specific database name as the following: "databaseNames": "my-database"

HTTP method and URL: 

POST https://sqladmin.googleapis.com/v1/projects/project-id/instances/source-instance-id/clone

Request JSON body: 

{
  "cloneContext":
  {
    "kind": "sql#cloneContext",
    "destinationInstanceName": "target-instance-id",
    "pointInTime": "restore-timestamp"
  }
}

To send your request, expand one of these options:

You should receive a JSON response similar to the following:

REST v1beta4

Before using any of the request data, make the following replacements:

  • project-id: The project ID
  • target-instance-id: The target instance ID
  • source-instance-id: The source instance ID
  • restore-timestamp The point-in-time to restore up to

In the JSON request, you optionally can specify up to one specific database name as the following: "databaseNames": "my-database"

HTTP method and URL: 

POST https://sqladmin.googleapis.com/sql/v1beta4/projects/project-id/instances/source-instance-id/clone

Request JSON body: 

{
  "cloneContext":
  {
    "kind": "sql#cloneContext",
    "destinationInstanceName": "target-instance-id",
    "pointInTime": "restore-timestamp"
  }
}

To send your request, expand one of these options:

You should receive a JSON response similar to the following:

Deactivate PITR

Console

  1. In the Google Cloud console, go to the Cloud SQL Instances page.

    Go to Cloud SQL Instances

  2. Open the more actions menu More actions icon. for the instance you want to deactivate and select Edit.
  3. Under Customize your instance, expand the Data Protection section.
  4. Clear Enable point-in-time recovery.
  5. Click Save.

gcloud

  1. Deactivate point-in-time recovery: 
    gcloud sql instances patch INSTANCE_NAME \
--no-enable-point-in-time-recovery
  2. Confirm your change: 
    gcloud sql instances describe INSTANCE_NAME

    In the backupConfiguration section, you see pointInTimeRecoveryEnabled: false if the change was successful.

REST v1

Before using any of the request data, make the following replacements:

  • project-id: The project ID
  • instance-id: The instance ID

HTTP method and URL: 

PATCH https://sqladmin.googleapis.com/v1/projects/project-id/instances/instance-id

Request JSON body: 

{
  "settings":
  {
    "backupConfiguration":
    {
      "enabled": false,
      "pointInTimeRecoveryEnabled": false
    }
  }
}

To send your request, expand one of these options:

You should receive a JSON response similar to the following:

REST v1beta4

Before using any of the request data, make the following replacements:

  • project-id: The project ID
  • instance-id: The instance ID

HTTP method and URL: 

PATCH https://sqladmin.googleapis.com/sql/v1beta4/projects/project-id/instances/instance-id

Request JSON body: 

{
  "settings":
  {
    "backupConfiguration":
    {
      "enabled": false,
      "pointInTimeRecoveryEnabled": false
    }
  }
}

To send your request, expand one of these options:

You should receive a JSON response similar to the following:

Check the storage location of transaction logs used for PITR

You can check where your Cloud SQL instance is storing the transaction logs used for PITR.

gcloud

To determine whether your instance stores logs for PITR on disk or Cloud Storage, use the following command:

   gcloud sql instances describe INSTANCE_NAME

Replace INSTANCE_NAME with the name of the instance.

You can also check the storage location of the transaction logs for multiple instances in the same project. To determine the location for multiple instances, use the following command:

   gcloud sql instances list --show-transactional-log-storage-state

Example response:

NAME  DATABASE_VERSION         LOCATION       TRANSACTIONAL_LOG_STORAGE_STATE
my_01 SQLSERVER_2019_STANDARD  us-central-1   DISK
my_02 SQLSERVER_2019_STANDARD  us-central-1   CLOUD_STORAGE
...

In the output of the command, the transactionalLogStorageState field or the TRANSACTIONAL_LOG_STORAGE_STATE column provides information about where the transaction logs for PITR are stored for the instance. The possible transaction log storage states are the following:

  • DISK: the instance stores the transaction logs used for PITR on disk.
  • CLOUD_STORAGE: the instance stores the transaction logs used for PITR in Cloud Storage.

Set transaction log retention

To set the number of days to retain transaction logs:

Console

  1. In the Google Cloud console, go to the Cloud SQL Instances page.

    Go to Cloud SQL Instances

  2. Open the more actions menu More actions icon. for the instance you want to set the transaction log on and select Edit.
  3. Under Customize your instance, expand the Data Protection section.
  4. In the Enable point-in-time recovery section, expand Advanced options.
  5. Enter the number of days to retain logs, from 1-35 for Cloud SQL Enterprise Plus edition or 1-7 for Cloud SQL Enterprise edition.
  6. Click Save.

gcloud

Edit the instance to set the number of days to retain transaction logs.

Replace the following:

  • INSTANCE_NAME: The name of the instance you want to set the transaction log on.

  • DAYS_TO_RETAIN: The number of days of transaction logs to keep. For Cloud SQL Enterprise Plus edition, the valid range is between 1 and 35 days, with a default of 14 days. For Cloud SQL Enterprise edition, the valid range is between 1 and 7 days, with a default of 7 days.

    If no value is specified, then the default value is used. This is valid only when PITR is enabled. Keeping more days of transaction logs requires a bigger storage size.

  gcloud sql instances patch INSTANCE_NAME \
    --retained-transaction-log-days=DAYS_TO_RETAIN

REST v1

Before using any of the request data, make the following replacements:

  • PROJECT_ID: the project ID.
  • INSTANCE_ID: the instance ID.

  • DAYS_TO_RETAIN: the number of days to retain transaction logs. For Cloud SQL Enterprise Plus edition, the valid range is between 1 and 35 days, with a default of 14 days. For Cloud SQL Enterprise edition, the valid range is between 1 and 7 days, with a default of 7 days.

    If no value is specified, then the default value is used. This is valid only when PITR is enabled. Keeping more days of transaction logs requires a bigger storage size.

HTTP method and URL: 

PATCH https://sqladmin.googleapis.com/v1/projects/PROJECT_ID/instances/INSTANCE_ID

Request JSON body: 

{
  "settings":
  {
    "backupConfiguration":
    {
      "transactionLogRetentionDays": "DAYS_TO_RETAIN"
    }
  }
}

To send your request, expand one of these options:

You should receive a JSON response similar to the following:

REST v1beta4

Before using any of the request data, make the following replacements:

  • PROJECT_ID: the project ID.
  • INSTANCE_ID: the instance ID.

  • DAYS_TO_RETAIN: the number of days to retain transaction logs. For Cloud SQL Enterprise Plus edition, the valid range is between 1 and 35 days, with a default of 14 days. For Cloud SQL Enterprise edition, the valid range is between 1 and 7 days, with a default of 7 days.

    If no value is specified, then the default value is used. This is valid only when PITR is enabled. Keeping more days of transaction logs requires a bigger storage size.

HTTP method and URL: 

PATCH https://sqladmin.googleapis.com/sql/v1beta4/projects/PROJECT_ID/instances/INSTANCE_ID

Request JSON body: 

{
  "settings":
  {
    "backupConfiguration":
    {
      "transactionLogRetentionDays": "DAYS_TO_RETAIN"
    }
  }
}

To send your request, expand one of these options:

You should receive a JSON response similar to the following:

Troubleshoot

Issue Troubleshooting

argument --point-in-time: Failed to parse date/time:
Unknown string format: 2021-0928T30:54:03.094;
received: 2021-0928T30:54:03.094Z

OR

Invalid value at 'body.clone_context.point_in_time'
(type.googleapis.com/google.protobuf.Timestamp), Field 'pointInTime',
Invalid time format: Failed to parse input,

 The timestamp you provided is invalid.

HTTP Error 400: Successful backup required for carrying out the operation was not found.

OR

Successful backup required for carrying out the operation was not found. or Time where no backups can be found.

 The timestamp that you provided is for a time where backups or when binlog coordinates could not be found.

What's next