This page describes how to upgrade the database major version by upgrading your Cloud SQL instance in-place rather than by migrating data.
Introduction
Database software providers periodically release new major versions that contain new features, performance improvements, and security enhancements. Cloud SQL takes in new versions after they're released. After Cloud SQL offers support for a new major version, you can upgrade your instances to keep your database updated.
You can upgrade the database version of an instance in-place or by migrating data. In-place upgrades are a simpler way to upgrade your instance's major version. You don't need to migrate data or change application connection strings. With in-place upgrades, you can retain the name, IP address, and other settings of your current instance after the upgrade. In-place upgrades don't require you to move data files and can be completed faster. In some cases, the downtime is shorter than what migrating your data entails.
The Cloud SQL for the PostgreSQL in-place upgrade operation uses thepg_upgrade
utility.
Plan a major version upgrade
- Confirm that you have the required role to perform a major version upgrade: Cloud SQL Owner or Cloud SQL Admin.
Choose a target major version.
gcloud
For information about installing and getting started with the gcloud CLI, see Install the gcloud CLI. For information about starting Cloud Shell, see Use Cloud Shell.
To check the database versions that you can target for an in-place upgrade on your instance, do the following:
- Run the following command.
- In the output of the command,
locate the section that is labeled
upgradableDatabaseVersions
. - Each subsection returns a database version that is available for upgrade. In each subsection, review the following fields.
majorVersion
: the major version that you can target for the in-place upgrade.name
: the database version string that includes the major version.displayName
: the display name for the database version.
gcloud sql instances describe INSTANCE_NAME
Replace INSTANCE_NAME with the name of the instance.
REST v1
To check which target database versions are available for a major version in-place upgrade, use the
instances.get
method of the Cloud SQL Admin API.Before using any of the request data, make the following replacements:
- INSTANCE_NAME: The instance name.
HTTP method and URL:
GET https://sqladmin.googleapis.com/v1/projects/PROJECT_ID/instances/INSTANCE_NAME
To send your request, expand one of these options:
You should receive a JSON response similar to the following:
upgradableDatabaseVersions: { major_version: "POSTGRES_15_0" name: "POSTGRES_15_0" display_name: "PostgreSQL 15.0" }
REST v1beta4
To check which target database versions are available for major version in-place upgrade of an instance, use the
instances.get
method of the Cloud SQL Admin API.Before using any of the request data, make the following replacements:
- INSTANCE_NAME: The instance name.
HTTP method and URL:
GET https://sqladmin.googleapis.com/sql/v1beta4/projects/PROJECT_ID/instances/INSTANCE_NAME
To send your request, expand one of these options:
You should receive a JSON response similar to the following:
upgradableDatabaseVersions: { major_version: "POSTGRES_15_0" name: "POSTGRES_15_0" display_name: "PostgreSQL 15.0" }
For the complete list of the database versions that Cloud SQL supports, see Database versions and version policies.
Consider the features offered in each database major version and address incompatibilities.
New major versions introduce incompatible changes that might require you to modify the application code, the schema, or the database settings. Before you can upgrade your database instance, review the release notes of your target major version to determine the incompatibilities that you must address.
Perform the precheck for upgrades.
Test the upgrade with a dry run.
Perform a dry run of the end-to-end upgrade process in a test environment before you upgrade the production database. You can clone your instance to create an identical copy of the data on which to test the upgrade process.
In addition to validating that the upgrade completes successfully, run tests to ensure that the application behaves as expected on the upgraded database.
Decide on a time to upgrade.
Upgrading requires the instance to become unavailable for a period of time. Plan to upgrade during a time period when database activity is low.
Prepare for a major version upgrade
Before you upgrade, complete the following steps.
-
Check the
LC_COLLATE
value for thetemplate
andpostgres
databases. The character set for each database must been_US.UTF8
.If the
LC_COLLATE
value for thetemplate
andpostgres
databases isn'ten_US.UTF8
, then the major version upgrade fails. To fix this, if either database has a character set other thanen_US.UTF8
, then change theLC_COLLATE
value toen_US.UTF8
before you perform the upgrade.To change the encoding of a database:
- Dump your database.
- Drop your database.
- Create a new database with the different encoding (for this example,
en_US.UTF8
). - Reload your data.
Another option is to rename the database:
- Close all connections to the database.
- Rename the database.
- Update your application configurations to use the new database name.
- Create a new, empty database with the default encoding.
We recommend that you perform these steps on a cloned instance before applying them to a production instance.
Manage your remaining PostgreSQL extensions.
Most extensions work on the upgraded database major version. Drop any extensions that are no longer supported in your target version. For example, drop the
chkpass
extension if you're upgrading to PostgreSQL 11 or later versions.You can upgrade PostGIS and its related extensions to their latest supported versions manually.
Sometimes, upgrading from PostGIS versions 2.x can create a situation where there are leftover database objects that aren't associated with the PostGIS extension. This can block the upgrade operation. For information about resolving this issue, see Fixing a broken postgis raster install.
Sometimes, upgrading to PostGIS version 3.1.7 or later can't complete due to objects using deprecated functions. This can block the upgrade operation. To check the upgrade status, run
To learn more about upgrading your PostGIS extensions, see Upgrading PostGIS. For issues associated with upgrading PostGIS, see Check the version of your PostgreSQL instance.SELECT PostGIS_full_version();
. If there are warnings present, then drop any objects using the deprecated functions and update the PostGIS extension to any intermediate or higher version. After you complete these actions, run theSELECT PostGIS_full_version();
command again. Verify that no warnings appear. Then, proceed with the upgrade operation.- Manage your custom database flags. Check the names of any custom database flags that you configured for your PostgreSQL instance. For issues associated with these flags, see Check the custom flags for your PostgreSQL instance.
- When performing an upgrade from one major version to another,
attempt to connect to each database to see if there are any compatibility issues.
Ensure that your databases can connect to each other. Check the
datallowconn
field for each database to ensure that a connection is allowed. At
value means that it's allowed, and anf
value indicates that a connection can't be established. - If you use the Datadog installation to upgrade your Cloud SQL instance to PostgreSQL 10 or later versions, then before you perform the upgrade, drop the pg_stat_activity() function.
Known limitations
The following limitations affect in-place major version upgrades for Cloud SQL for PostgreSQL:
- You can't perform an in-place major version upgrade on an external replica.
- Upgrading instances that have more than 1,000 databases from one version to another might take a long time and time out.
- Use the
select * from pg_largeobject_metadata;
statement to query for the number of large objects in each PostgreSQL database of your Cloud SQL instance. If the result from all of your databases is more than 10 million large objects, then the upgrade fails. Cloud SQL rolls back to the previous version of your database. - Before you perform an in-place major version upgrade to PostgreSQL 16 and later, upgrade the PostGIS extension for all of your databases to version 3.4.0. For PostgreSQL 18, upgrade to PostGIS version 3.6.0.
- Before you perform an in-place major version upgrade to PostgreSQL 17, upgrade the
rdkit
extension for all of your databases to version 4.6.1. - Before you perform an in-place major version upgrade to PostgreSQL 16, 17, or 18, upgrade the
pg_squeeze
extension for all of your databases to version 1.6, 1.7, or 1.8 respectively. - If you're using PostgreSQL versions 9.6, 10, 11, or 12, then version 3.4.0 of the PostGIS extension isn't supported. Therefore, to perform an in-place major version upgrade to PostgreSQL 16 and later, you must first upgrade to an intermediate version of PostgreSQL (versions 13, 14, or 15).
If you install the
pg_ivm
extension for your instance, then you can't perform a major version upgrade. To fix this, uninstall this extension and then perform the upgrade. For more information about the extensions, see Configure PostgreSQL extensions.If you enable the vacuum_defer_cleanup_age and force_parallel_mode flags, then you can't perform a major version upgrade. To fix this, delete these flags and then perform the upgrade. For more information about the flags, including how to delete them, see Configure database flags.
Assess upgrade readiness for your instance
Cloud SQL lets you run a precheck on your instance before a major version upgrade. This precheck is a long-running operation (LRO) that checks if your instance is ready for an upgrade. It helps find potential problems like incompatibilities, configuration issues, or data problems prior to the upgrade operation.
The precheck either confirms your instance can be upgraded, or lists issues you need to fix first and their solutions. These issues might be due to incompatible extensions, unsupported dependencies, or data format problems.
The precheck mainly reads your instance's metadata and performs checks. These tasks don't affect your instance's performance or cause downtime. We highly recommend running the precheck as it helps prevent upgrade failures and unexpected downtime.
When you run the precheck, one of the following happens:
- No issues found: The precheck finished successfully, and no problems were found.
- Upgrade blocking issues found: The precheck finished successfully, but it found errors that stop the upgrade. The issues must be resolved prior to the upgrade.
- Non-blocking warnings found: The precheck finished successfully and found warnings, but none of them stop the upgrade.
Depending on the precheck's results, you can either proceed with the upgrade or fix the identified issues before upgrading.
Limitations
When using the major version upgrade precheck, consider these limitations:
- The instance state must be set to
RUNNING
. - The instance must be a primary instance. precheck doesn't support replica instances.
The instance must not have any blocking operations pending. If a blocking operation is pending, then the precheck results in an error with the following message:
Operation failed because another operation was already in progress. Try your request after the current operation is complete.
The precheck needs to connect to all databases on the instance. If a database is inaccessible, locked, or unresponsive, the precheck might fail or show errors. Although the precheck doesn't affect your instance's performance or cause downtime, we recommend running the precheck when database load is low.
Before you begin
- Make sure the Cloud SQL Admin API is enabled for your instance.
- Confirm you have the
cloudsql.instances.preCheckMajorVersionUpgrade
IAM permission.
Perform the precheck
To perform the major version upgrade precheck, do the following:
gcloud
Run the precheck:
gcloud sql instances pre-check-major-version-upgrade INSTANCE_NAME \ --target-database-version=TARGET_DATABASE_VERSION \ --project=PROJECT_ID \ [--async]
Replace the following:
- INSTANCE_NAME: the name of the instance.
- TARGET_DATABASE_VERSION: the major version you want to upgrade your instance to. To find the database version, see Plan an upgrade.
- PROJECT_ID: the ID of your Google Cloud project.
Get the precheck operation name:
Use the
gcloud sql operations list
command with the--instance
flag:gcloud sql operations list --instance=INSTANCE_NAME
Replace the following:
- INSTANCE_NAME: the name of the instance.
Monitor the status of the precheck.
Use the
gcloud sql operations describe
command:gcloud sql operations describe OPERATION_NAME
Replace the following:
- OPERATION_NAME: the precheck operation name retrieved in the previous step.
REST v1
Run the precheck.
Before using any of the request data, make the following replacements:
- PROJECT_ID: The project ID
- INSTANCE_ID: The instance ID
- TARGET_DATABASE_VERSION: The major version to upgrade to. To find a list of available database versions, see Plan an upgrade.
HTTP method and URL:
POST https://sqladmin.googleapis.com/sql/v1b/projects/PROJECT-ID/instances/INSTANCE_ID/preCheckMajorVersionUpgrade
Request JSON body:
{ "preCheckMajorVersionUpgradeContext": { "targetDatabaseVersion": "TARGET_DATABASE_VERSION" } }
To send your request, expand one of these options:
You should receive a JSON response similar to the following:
{ "message": "Precheck description of finding", "message_type": "ERROR", "actions_required": [ "Precheck action required to fix the finding" ] }
Get the precheck operation name.
Use the GET request with
operations.list
method after replacingPROJECT_ID
with the ID of the project.GET https://sqladmin.googleapis.com/v1/projects/PROJECT_ID/operations
Replace the following:
- PROJECT_ID: the ID of your Google Cloud project.
Monitor the status of the precheck.
Use the GET request with
operations.list
method:GET https://sqladmin.googleapis.com/v1/projects/PROJECT_ID/operation/OPERATION_NAME
Replace the following:
- PROJECT_ID: the ID of your Google Cloud project.
- OPERATION_NAME: the precheck operation name retrieved in the previous step.
REST v1beta4
Run the precheck.
Before using any of the request data, make the following replacements:
- PROJECT_ID: The project ID
- INSTANCE_ID: The instance ID
- TARGET_DATABASE_VERSION: The major version to upgrade to. To find a list of available database versions, see Plan an upgrade.
HTTP method and URL:
POST https://sqladmin.googleapis.com/sql/v1beta4/projects/PROJECT-ID/instances/INSTANCE_ID/preCheckMajorVersionUpgrade
Request JSON body:
{ "preCheckMajorVersionUpgradeContext": { "targetDatabaseVersion": "TARGET_DATABASE_VERSION" } }
To send your request, expand one of these options:
You should receive a JSON response similar to the following:
{ "message": "Precheck description of finding", "message_type": "ERROR", "actions_required": [ "Precheck action required to fix the finding" ] }
Get the precheck operation name.
Use the GET request with
operations.list
method after replacingPROJECT_ID
with the ID of the project.GET https://sqladmin.googleapis.com/v1/projects/PROJECT_ID/operations
Replace the following:
- PROJECT_ID: the ID of your Google Cloud project.
Monitor the status of the precheck.
Use the GET request with
operations.list
method:GET https://sqladmin.googleapis.com/v1/projects/PROJECT_ID/operation/OPERATION_NAME
Replace the following:
- PROJECT_ID: the ID of your Google Cloud project.
- operation_name: the precheck operation name retrieved in the previous step.
Review precheck findings
After the precheck finishes, your instance is either ready for upgrade, or it has issues that need your attention.
Ready for upgrade
If the precheck finishes successfully and the preCheckResponse
array is
empty, it means no issues or warnings were found. Your instance is ready for
the major version upgrade. To continue, see
Perform the major version upgrade.
Not ready for upgrade
If the precheck ran successfully and the preCheckResponse
array contains
issues, your instance isn't ready for the upgrade and needs attention. The
identified issues might or might not block the upgrade. These issues are
noted in the preCheckResponse
with the following message types:
Type | Description | Blocking upgrade? |
---|---|---|
INFO |
An informational message. | No |
WARNING |
A potential issue was found, but it doesn't block the upgrade. Cloud SQL recommends reviewing and addressing the warning before upgrading to ensure full compatibility. | No |
ERROR |
A critical issue that blocks the upgrade was found. These issues might cause the upgrade to fail. You must resolve them before upgrading your instance. | Yes |
If your instance only has INFO
or WARNING
messages, you can upgrade it,
but you might have issues after the upgrade. We recommend reviewing the
message details and resolving the issue before upgrading. If your instance
has ERROR
messages, you must resolve these issues before upgrading.
Each issue type includes a message
and an actions_required
field. Review
each issue to understand its type and how to resolve it. For more information
about common issues and their solutions, see
Common major version upgrade precheck errors.
After you resolve the issues, re-run the precheck to confirm your instance is ready for the upgrade. Then, proceed with upgrading your instance once the precheck is clear.
Perform the major version upgrade
You can upgrade the major version of a single Cloud SQL instance, or you can upgrade the major version of a primary instance and include all of its replicas in the upgrade, including cascading replicas and cross-region replicas.
Upgrade the major version of a single instance
When you initiate an upgrade operation for a single instance, Cloud SQL does the following:
- Checks the configuration of your instance to ensure that the instance is compatible for an upgrade.
- After Cloud SQL verifies the configuration, then Cloud SQL makes the instance unavailable.
- Makes a pre-upgrade backup.
- Performs the upgrade on the instance.
- Makes your instance available.
- Makes a post-upgrade backup.
Console
-
In the Google Cloud console, go to the Cloud SQL Instances page.
- To open the Overview page of an instance, click the instance name.
- Click Edit.
- In the Instance info section, click the Upgrade button and confirm that you want to go to the upgrade page.
- On the Choose a database version page, click the Database version for upgrade list and select one of the available database major versions.
- Click Continue.
- In the Instance ID box, enter the name of the instance and then click the Start upgrade button.
Verify that the upgraded database major version appears below the instance name on the instance Overview page.
gcloud
Start the upgrade.
Use the
gcloud sql instances patch
command with the--database-version
flag.Before running the command, replace the following:
- INSTANCE_NAME: The name of the instance.
- DATABASE_VERSION: The enum for the database major version, which must be later than the current version. Specify a database version for a major version that is available as an upgrade target for the instance. You can obtain this enum as the first step of Plan for upgrade. If you need a complete list of database version enums, then see SqlDatabaseEnums.
gcloud sql instances patch INSTANCE_NAME \ --database-version=DATABASE_VERSION
Major version upgrades take several minutes to complete. You might see a message indicating that the operation is taking longer than expected. You can either ignore this message or run the
gcloud sql operations wait
command to dismiss the message.Get the upgrade operation name.
Use the
gcloud sql operations list
command with the--instance
flag.Before running the command, replace the INSTANCE_NAME variable with the name of the instance.
gcloud sql operations list --instance=INSTANCE_NAME
Monitor the status of the upgrade.
Use the
gcloud sql operations describe
command.Before running the command, replace the OPERATION variable with the upgrade operation name retrieved in the previous step.
gcloud sql operations describe OPERATION
REST v1
Start the in-place upgrade.
Use a PATCH request with the
instances:patch
method.Before using any of the request data, replace these variables:
- PROJECT_ID: The ID of the project.
- INSTANCE_NAME: The name of the instance.
HTTP method and URL:
PATCH https://sqladmin.googleapis.com/v1/projects/PROJECT_ID/instances/INSTANCE_NAME
Request JSON body:
{ "databaseVersion": DATABASE_VERSION }
Replace DATABASE_VERSION with the enum for the database major version, which must be later than the current version. Specify a database version for a major version that is available as an upgrade target for the instance. You can obtain this enum as the first step of Plan for upgrade. If you need a full list of database version enums, then see SqlDatabaseVersion.
Get the upgrade operation name.
Use a GET request with the
operations.list
method after replacing PROJECT_ID with the ID of the project.HTTP method and URL:
GET https://sqladmin.googleapis.com/v1/projects/PROJECT_ID/operations
Monitor the status of the upgrade.
Use a GET request with the
operations.get
method after replacing the following variables:- PROJECT_ID: The ID of the project.
- OPERATION_NAME: The upgrade operation name retrieved in the previous step.
HTTP method and URL:
GET https://sqladmin.googleapis.com/v1/projects/PROJECT_ID/operation/OPERATION_NAME
Terraform
To update the version of the database, use a Terraform resource and the Terraform provider for Google Cloud, version 4.34.0 or later.
Apply the changes
To apply your Terraform configuration in a Google Cloud project, complete the steps in the following sections.
Prepare Cloud Shell
- Launch Cloud Shell.
-
Set the default Google Cloud project where you want to apply your Terraform configurations.
You only need to run this command once per project, and you can run it in any directory.
export GOOGLE_CLOUD_PROJECT=PROJECT_ID
Environment variables are overridden if you set explicit values in the Terraform configuration file.
Prepare the directory
Each Terraform configuration file must have its own directory (also called a root module).
-
In Cloud Shell, create a directory and a new
file within that directory. The filename must have the
.tf
extension—for examplemain.tf
. In this tutorial, the file is referred to asmain.tf
.mkdir DIRECTORY && cd DIRECTORY && touch main.tf
-
If you are following a tutorial, you can copy the sample code in each section or step.
Copy the sample code into the newly created
main.tf
.Optionally, copy the code from GitHub. This is recommended when the Terraform snippet is part of an end-to-end solution.
- Review and modify the sample parameters to apply to your environment.
- Save your changes.
-
Initialize Terraform. You only need to do this once per directory.
terraform init
Optionally, to use the latest Google provider version, include the
-upgrade
option:terraform init -upgrade
Apply the changes
-
Review the configuration and verify that the resources that Terraform is going to create or
update match your expectations:
terraform plan
Make corrections to the configuration as necessary.
-
Apply the Terraform configuration by running the following command and entering
yes
at the prompt:terraform apply
Wait until Terraform displays the "Apply complete!" message.
- Open your Google Cloud project to view the results. In the Google Cloud console, navigate to your resources in the UI to make sure that Terraform has created or updated them.
Delete the changes
To delete your changes, do the following:
- To disable deletion protection, in your Terraform configuration file set the
deletion_protection
argument tofalse
.deletion_protection = "false"
- Apply the updated Terraform configuration by running the following command and
entering
yes
at the prompt:terraform apply
-
Remove resources previously applied with your Terraform configuration by running the following command and entering
yes
at the prompt:terraform destroy
When you place an in-place upgrade request, Cloud SQL first performs a pre-upgrade check. If Cloud SQL determines that your instance isn't ready for an upgrade, then your upgrade request fails with a message suggesting how you can address the issue. See also Troubleshoot a major version upgrade.
Include replicas in the major version upgrade
If your primary instance has replicas, then you can include all replicas in the upgrade. Cloud SQL can upgrade all replicas of the primary instance, including cross-region replicas and cascading replicas.
When you include replicas in a major version upgrade, Cloud SQL does the following:
- Checks the configuration of your primary instance and replicas to ensure that the instance and replicas are compatible for an upgrade.
- Makes your primary instance unavailable.
- Makes a pre-upgrade backup of the primary instance.
- Stops replication for all replicas.
- Performs the upgrade on the primary instance.
- If the upgrade on the primary instance is successful, then the primary instance becomes available again and restarts replication.
- Cloud SQL takes a post-upgrade backup of the primary instance.
- Cloud SQL proceeds to upgrade all replicas.
Even if the major version upgrade of a replica fails, the primary instance continues to be available.
To include replicas in a major version upgrade, you can't use the Google Cloud console or Terraform. You can only use gcloud CLI or the Cloud SQL Admin API.
gcloud
Start the upgrade.
Use the
gcloud sql instances patch
command with the--database-version
and the flags.--include-replicas-for-major-version-upgrade
Before running the command, replace the following:
- INSTANCE_NAME: The name of the primary instance.
- DATABASE_VERSION: The enum for the database major version, which must be later than the current version. Specify a database version for a major version that is available as an upgrade target for the instance. You can obtain this enum as the first step of Plan for upgrade. If you need a complete list of database version enums, then see SqlDatabaseEnums.
gcloud sql instances patch INSTANCE_NAME \ --database-version=DATABASE_VERSION \ --include-replicas-for-major-version-upgrade
Major version upgrades take several minutes to complete. You might see a message indicating that the operation is taking longer than expected. You can either ignore this message or run the
gcloud sql operations wait
command to dismiss the message. Upgrading replicas can take several minutes to complete. To check the status of the upgrade, do the following:Get the upgrade operation name.
Use the
gcloud sql operations list
command with the--instance
flag.Before running the command, replace the INSTANCE_NAME variable with the name of the instance.
gcloud sql operations list --instance=INSTANCE_NAME
Monitor the status of the upgrade.
Use the
gcloud sql operations describe
command.Before running the command, replace the OPERATION variable with the upgrade operation name retrieved in the previous step.
gcloud sql operations describe OPERATION
REST
Start the in-place upgrade.
Use a PATCH request with the
instances:patch
method.Before using any of the request data, replace these variables:
- PROJECT_ID: The ID of the project.
- INSTANCE_NAME: The name of the instance.
HTTP method and URL:
PATCH https://sqladmin.googleapis.com/v1/projects/PROJECT_ID/instances/INSTANCE_NAME
Request JSON body:
{ "databaseVersion": DATABASE_VERSION "includeReplicasForMajorVersionUpgrade": true }
- Replace DATABASE_VERSION with the enum for the database major version, which must be later than the current version. Specify a database version for a major version that is available as an upgrade target for the instance. You can obtain this enum as the first step of Plan for upgrade. If you need a full list of database version enums, then see SqlDatabaseVersion.
- In the
includeReplicasForMajorVersionUpgrade
field, specifytrue
.
Get the upgrade operation name.
Use a GET request with the
operations.list
method after replacing PROJECT_ID with the ID of the project.HTTP method and URL:
GET https://sqladmin.googleapis.com/v1/projects/PROJECT_ID/operations
Monitor the status of the upgrade.
Use a GET request with the
operations.get
method after replacing the following variables:- PROJECT_ID: The ID of the project.
- OPERATION_NAME: The upgrade operation name retrieved in the previous step.
HTTP method and URL:
GET https://sqladmin.googleapis.com/v1/projects/PROJECT_ID/operation/OPERATION_NAME
Automatic upgrade backups
When you perform a major version upgrade, Cloud SQL automatically makes two on-demand backups, called upgrade backups:
- The first upgrade backup is the pre-upgrade backup, which is made immediately before starting the upgrade. You can use this backup to restore your database instance to its state on the previous version.
- The second upgrade backup is the post-upgrade backup, which is made immediately after new writes are allowed to the upgraded database instance.
When you view your list of
backups, the
upgrade backups are listed with type On-demand
. Upgrade backups are labeled so
that you can identify them quickly.
For example, if you're upgrading from PostgreSQL 14 to PostgreSQL 15, your
pre-upgrade backup is labeled Pre-upgrade backup, POSTGRES_14 to POSTGRES_15.
and your post-upgrade backup is labeled Post-upgrade backup, POSTGRES_14 to
POSTGRES_15.
As with other on-demand backups, upgrade backups persist until you delete them or delete the instance. If you have PITR enabled, you can't delete your upgrade backups while they're in your retention window. If you need to delete your upgrade backups, you must disable PITR or wait until your upgrade backups are no longer in your retention window.
Complete the major version upgrade
After you finish upgrading your primary instance, perform the following steps to complete your upgrade:
Refresh the database statistics.
Run
ANALYZE
on your primary instance to update the system statistics after the upgrade. Accurate statistics make sure that the PostgreSQL query planner processes queries optimally. Missing statistics can lead to bad query plans, which in turn might degrade performance and take up excessive memory.Perform acceptance tests.
Run tests to make sure that the upgraded system performs as expected.
Troubleshoot a major version upgrade
Cloud SQL returns an error message if you attempt an invalid upgrade command, for example, if your instance contains invalid database flags for the new version.
If your upgrade request fails, check the syntax of your upgrade request. If the request has a valid structure, try looking into the following suggestions.
View error logs
If any issues occur with a valid upgrade request, then Cloud SQL
publishes error logs to projects/PROJECT_ID/logs/cloudsql.googleapis.com%2Fpostgres-upgrade.log
. Each log entry contains a label with the
instance identifier to help you identify the instance with the upgrade error.
Look for such upgrade errors and resolve them.
To view error logs, use the Google Cloud console::
-
In the Google Cloud console, go to the Cloud SQL Instances page.
- To open the Overview page of an instance, click the instance name.
In the Operations and logs pane of the instance Overview page, click the View PostgreSQL error logs link.
The Logs Explorer page opens.
View logs as follows:
- To list all error logs in a project, select the log name in the Log name log filter.
For more information on query filters, see Advanced queries.
- To filter the upgrade error logs for a single instance, enter the
following query in the Search all fields box, after replacing
DATABASE_ID
with the project ID followed by the instance name in this format:
project_id:instance_name
.resource.type="cloudsql_database" resource.labels.database_id="DATABASE_ID" logName : "projects/PROJECT_ID/logs/cloudsql.googleapis.com%2Fpostgres-upgrade.log"
For example, to filter the upgrade error logs by an instance named
shopping-db
running in the projectbuylots
, use the following query filter:resource.type="cloudsql_database" resource.labels.database_id="buylots:shopping-db" logName : "projects/buylots/logs/cloudsql.googleapis.com%2Fpostgres-upgrade.log"
You can either review all logs reported within a given timeframe, or you can filter logs by severity. A common option for troubleshooting might include selecting the following filters:
- Emergency
- Alert
- Critical
- Error
Log entries with the pg_upgrade_dump
prefix indicate that an upgrade error had
occurred. For example:
pg_upgrade_dump: error: query failed: ERROR: out of shared memory
HINT: You might need to increase max_locks_per_transaction.
Additionally, log entries labeled with a .txt
secondary filename might list
other errors that you might want to resolve before attempting the upgrade again.
All filenames are found in the postgres-upgrade.log
file. To locate a filename,
look at the labels.FILE_NAME
field.
Filenames that might contain errors to resolve include:
tables_with_oids.txt:
This file contains tables that are listed with object identifiers (OIDs). Either delete the tables or modify them so that they don't use OIDs.tables_using_composite.txt:
This file contains tables that are listed using system-defined composite types. Either delete the tables or modify them so that they don't use these composite types.tables_using_unknown.txt:
This file contains tables that are listed using theUNKNOWN
data type. Either delete the tables or modify them so that they don't use this data type.tables_using_sql_identifier.txt:
This file contains tables that are listed using theSQL_IDENTIFIER
data type. Either delete the tables or modify them so that they don't use this data type.tables_using_reg.txt:
This file contains tables that are listed using theREG*
data type (for example,REGCOLLATION
orREGNAMESPACE
). Either delete the tables or modify them so that they don't use this data type.postfix_ops.txt:
This file contains tables that are listed using postfix (right-unary) operators. Either delete the tables or modify them so that they don't use these operators.
Check the memory
If the instance has insufficient shared memory, you might see this error
message: ERROR: out of shared memory.
This error is more likely to occur if
you have in excess of 10,000 tables.
Before you attempt an upgrade, set the value of the
max_locks_per_transaction
flag to approximately twice the number of tables in the instance. The instance
is restarted when you change the value of this flag.
Check the connections capacity
If your instance has insufficient connection capacity, you might see this
error message: ERROR: Insufficient connections.
Cloud SQL recommends that you increase the max_connections
flag value by the number of databases in your instance. The instance is
restarted when you change the value of this flag.
Check for an ambiguous column reference
Cloud SQL automatically performs a pre-upgrade check to identify user-defined
views that depend on system catalog views, such as pg_stat_activity
or
pg_stat_replication
. The column structure of these system catalog views can
change between major PostgreSQL versions. If you have views that select *
or
rely on the column order of these system views, then they might become
incompatible after an upgrade, resulting in an error, such as
ERROR: column reference "column_name" is ambiguous
.
The pre-upgrade check detects such views by checking for dependencies. If incompatible views are found, the upgrade process is stopped and an error message is displayed. This message lists the incompatible views in each database that need to be addressed.
Example Error Message
For
pg_stat_activity
related issues:Please remove the following usages of views that depend on functions returning data types of pg_stat_activity before attempting an upgrade: (database: my_db, schema name: public, view name: my_stat_activity_view)
For
pg_stat_replication
related issues:Please remove the following usages of views that depend on functions returning data types of pg_stat_replication before attempting an upgrade: (database: my_db, schema name: public, view name: my_replication_stats_view)
To resolve such issues and proceed with the upgrade: 1. Identify the views listed in the pre-upgrade check error message.
Drop these views using
DROP VIEW view_name;
.Retry the major version upgrade.
Once the upgrade is complete, recreate the views. Ensure the new view definitions are compatible with the schema of the system catalog views in the current PostgreSQL version. You might need to explicitly list columns instead of using
select *
to avoid future issues.
For a more-detailed example of the problem and further insights, see this stack overflow discussion
Check for SRFs in CASE statements
If you are upgrading your instance from version 9.6 and using set returning
functions in your CASE statements, then you might see this error message
ERROR: set-returning functions are not allowed in CASE
. This issue occurs as
from version 10 onwards using set-returning functions in CASE statements is disallowed.
To resolve this issue and upgrade your instance successfully, ensure that any CASE statements utilizing set-returning functions are modified to avoid their use before retrying the upgrade. Some commonly used SRFs include the following:
- unnest()
- generate_series()
- array_agg()
- regexp_split_to_table()
- jsonb_array_elements()
- json_array_elements()
- sonb_each()
- json_each()
Check views created on custom casts
If you have a view created on a custom cast, then an error message similar to the following appears: ERROR: cannot cast type <type_1> to <type_2>
.
This issue occurs because of permission issues on custom created casts.
To resolve this issue, update your instance to [PostgreSQL version].R20240910.01_02
For more information, see Self-service maintenance.
Check event trigger ownership
In Cloud SQL, all event triggers must be owned by a user with the
cloudsqlsuperuser
role. Cloud SQL performs a pre-upgrade check to
validate ownership of all event triggers. If an event trigger is owned by a
user who lacks the cloudsqlsuperuser
role, then the upgrade process is halted
and you might get an error message, such as:
Please ensure that the owners of all event triggers have the cloudsqlsuperuser role assigned to them before attempting an upgrade: (database: your_db, triggerName your_trigger, owner: non_super_user)
To resolve this issue, either change the owner of the event trigger to a user
that has the cloudsqlsuperuser
role, such as postgres
, or grant the
cloudsqlsuperuser
role to the current owner.
To identify event triggers with owners lacking the required role, run the following command:
SELECT t.evtname AS trigger_name, r.rolname AS current_owner FROM pg_event_trigger t JOIN pg_roles r ON t.evtowner = r.oid WHERE NOT pg_has_role(r.rolname, 'cloudsqlsuperuser', 'member');
The results show any event trigger with an owner who doesn't have the
cloudsqlsuperuser
role.
Check generated columns from unlogged tables
If you have an unlogged table which has generated columns you might see the error
message ERROR: unexpected request for new relfilenumber in binary upgrade mode
.
This issue occurs due to discrepancies in the persistence characteristics between
tables and their sequences for generated columns.
To address this issue, do the following:
- Drop unlogged tables: if possible, drop any unlogged tables that are linked to generated columns. Make sure that data loss can be safely mitigated before proceeding.
-
Convert to permanent tables: temporarily, convert unlogged tables to permanent
tables using the following steps:
- Convert the table to a logged table
ALTER TABLE
SET LOGGED; - Perform major version upgrade
- Convert the table back to an unlogged table
ALTER TABLE
SET UNLOGGED
- Convert the table to a logged table
You can identify all such tables by using the following query :
SELECT relnamespace::regnamespace, c.relname AS table_name, a.attname AS column_name, a.attidentity AS identity_type FROM pg_catalog.pg_class c JOIN pg_catalog.pg_attribute a ON a.attrelid = c.oid WHERE a.attidentity IN ('a', 'd') AND c.relkind = 'r' AND c.relpersistence = 'u' ORDER BY c.relname, a.attname;
Check the custom flags for your PostgreSQL instance
If you're upgrading to a PostgreSQL instance, version 14 or higher, then check the names of any custom database flags that you configured for the instance. This is because PostgreSQL placed additional restrictions on allowed names for custom parameters.
The first character of a custom database flag must be alphabetic (A-Z or a-z). All subsequent characters can be alphanumeric, the underscore (_) special character, or the dollar sign ($) special character.
Remove extensions
If you're upgrading your Cloud SQL instance,
then you might see this error message: pg_restore: error: could not execute
query: ERROR: role "16447" does not exist
.
To resolve this issue, follow these steps:
- Remove the
pg_stat_statements
andpgstattuple
extensions. - Perform the upgrade.
- Reinstall the extensions.
Common major version upgrade precheck errors
Issues found by the major version upgrade precheck fall into these categories:
Incompatible extensions: These are Cloud SQL for PostgreSQL extensions on your instance that don't work with the new major version.
Unsupported dependencies: These are dependencies that either aren't supported by the new major version or need updates to work with it.
Database incompatibilities: These are problems with your database or data that might happen after a major version upgrade. This includes differences in database structures, data types, encoding, collation, or system catalog changes specific to the new version.
Incompatible extensions
The following table lists common errors related to incompatible extensions that the major version upgrade precheck might find:
Type | Error example | Resolution |
---|---|---|
Unsupported or deprecated extension | Your installation contains unsupported extensions for the
new version. These extensions must be removed before attempting an
upgrade: (database: %s, Extension name: %s) |
Remove the extension from all databases that use it with
DROP EXTENSION $extension_name; . |
Incompatible extension version | Your installation contains incompatible version extensions.
These extensions must be upgraded to a compatible version before
attempting an upgrade: (database: %s, Extension name: %s) |
Update the extension to a version that works with your target Cloud SQL for PostgreSQL version. For compatible versions, see Configure Cloud SQL for PostgreSQL extensions. |
PostGIS unpackaged files |
PostGIS version upgrade has not been completed, unpackaged
raster files present. Follow the steps at
https://postgis.net/documentation/tips/tip-removing-raster-from-2-3/ to
fix before major version upgrade. |
Clean up the unpackaged raster files. |
PostGIS deprecated functions |
PostGIS version upgrade has not been completed, deprecated
functions present. Please drop all objects using deprecated functions
and upgrade to a different version of PostGIS before major version
upgrade. |
Find and remove or change any database objects that use deprecated
PostGIS functions before upgrading the PostGIS
extension. |
Extension ownership | Please ensure that the owner of the postgres_fdw extension
has the cloudsqlsuperuser role assigned to them before attempting an
upgrade: (database: my_db, extension name: postgres_fdw, owner: some_user) |
Change the extension owner using
ALTER EXTENSION postgres_fdw OWNER TO postgres; . |
Unsupported dependencies
The following table lists common errors related to unsupported dependencies that the major version upgrade precheck might find:
Type | Error example | Resolution |
---|---|---|
Event trigger ownership | Please ensure that the owners of all event triggers have the
cloudsqlsuperuser role assigned to them before attempting an upgrade:
(database: your_db, triggerName your_trigger, owner: non_super_user) |
Connect to the identified database using psql or
Cloud SQL Studio and change the trigger's owner to a
postgres user. |
Uncommitted prepared statements | Please commit/rollback the following usages of 'Uncommitted
Prepared Statements'... (database: my_db, gid: my_prepared_xact) |
Either commit or roll back the prepared statement. |
Deprecated flags | flag "force_parallel_mode" is deprecated in new postgres
version, Please delete this flag before retrying again |
Remove the database flag from the instance configuration. |
Database incompatibilities
The following table lists common errors related to data format incompatibilities that the major version upgrade precheck might find:
Type | Error example | Resolution |
---|---|---|
Unknown data type | Please remove the following usages of 'Unknown' data types
before attempting an upgrade: (database: my_db, relation: my_table,
attribute: my_column) |
Remove the column or table, or change the table's data type using
ALTER TABLE my_table ALTER COLUMN my_column TYPE TEXT; . |
reg* data type |
Please remove the following usages of 'reg*' data types before
attempting an upgrade: (database: my_db, relation: my_table, attribute:
my_column) |
Remove the column or change its data type. |
Removed data type | Please remove the following usages of 'sql_identifier' data
types before attempting an upgrade: ... |
Convert to TEXT , timestamptz , or another
suitable data type. |
aclitem Internal Format |
Please remove the following usages of 'aclitem' data types before attempting an upgrade: ... |
Stop using aclitem in your database table definitions. |
System-defined composite data types | Please remove the following usages of 'composite' data types
before attempting an upgrade: (database: my_db, relation: my_table,
attribute: my_column) |
Change the identified columns to use a user-defined composite type or a standard data type. System composite types may not be consistent across major versions. |
Tables with OIDS |
Please remove the following usages of tables with OIDs before attempting an upgrade: (database: my_db, relation: my_table) |
Update the table using ALTER TABLE my_table SET WITHOUT OIDS; . |
User-defined postfix operators |
Please remove the following usages of 'postfix operators'
before attempting an upgrade: (database: my_db, operation id: 12345,
operation namespace: public, operation name: !!, type namespace:
public, type name: mytype) |
Remove the custom postfix operators. You might need to
rewrite your code to use prefix operators or function calls instead. |
Incompatible polymorphic functions | Please remove the following usages of 'incompatible polymorphic'
functions before attempting an upgrade: (database: my_db, object kind:
function, object name: public.my_poly_func) |
Remove or change the function to remove incompatible polymorphic functions. This might mean adjusting function signatures or logic to work with Cloud SQL for PostgreSQL 14 and later. |
User-defined encoding conversions | Please remove the following usages of user-defined encoding
conversions before attempting an upgrade: (database: my_db, namespace
name: public, encoding conversions name: my_encoding_conv) |
Remove the user-defined encoding conversion. You might need to recreate it after the upgrade with a signature that works with the new version. |
Check for an ambiguous column reference | Cloud SQL automatically checks for user-defined views that
rely on system catalog views. The column structure of these system
catalog views might change between major versions.Please remove the following usages of views that depend on functions returning data types of pg_stat_activity before attempting an upgrade: (database: my_db, schema name: public, view name: my_stat_activity_view)
|
Find the views listed in the error message and remove them using the
DROP VIEW command. After the upgrade, recreate the views. |
Unlogged tables with generated columns or logged sequences | Please drop the following usages of 'Unlogged Tables with Logged
Sequence' before attempting an upgrade: (database: your_db, table name: problematic_table) |
You can either convert the table to LOGGED , or remove
it using the DROP TABLE command. Recreate the table after
the upgrade. |
Fix the empty search path issue | Please update the search path of the 'll_to_earth' function (database: your_db, search path: ) |
The earthdistance extension uses earth
and cube types without specifying the function's search path.
Update the search path using
ALTER FUNCTION ll_to_earth SET search_path = public; . |
Restore the primary instance to the previous major version
If your upgraded database system doesn't perform as expected, then you might need to restore your primary instance to the previous version. You do so by restoring your pre-upgrade backup to a Cloud SQL recovery instance, which is a new instance running the pre-upgrade version.
To restore a primary instance to the previous version, perform the following steps:
Identify your pre-upgrade backup.
Create a recovery instance.
Create a new Cloud SQL instance using the major version that Cloud SQL was running when the pre-upgrade backup was made. Set the same flags and instance settings that the original instance uses.
Restore your pre-upgrade backup.
Restore your pre-upgrade backup to the recovery instance. This might take several minutes to complete.
Add your read replicas.
If you're using read replicas, then add the read replicas individually.
Connect your application.
Having recovered your database system, update your application with details about the recovery instance and its read replicas. You can resume serving traffic on the pre-upgrade version of your database.
FAQs
The following questions might come up when upgrading the database major version.
- Yes. Your instance remains unavailable for a period of time while Cloud SQL performs the upgrade.
- How long does an upgrade take?
Upgrading a single instance typically takes less than 10 minutes. If your instance configuration has a small number of vCPUs or memory, then your upgrade might take more time.
If your instance hosts too many databases or tables, or your databases are very large, then the upgrade might take hours or even time out because the total upgrade time corresponds to the number of objects in your databases. If you have multiple instances that need to be upgraded, then your upgrade time increases proportionately. If you include replicas in your upgrade, then the upgrade operation can take up to an hour to complete, depending on the number of replicas that your primary instance has.
- Can I monitor each step in my upgrade process?
- While Cloud SQL lets you monitor whether an upgrade operation is still in progress, you can't track the individual steps in each upgrade.
- Can I cancel my upgrade after I've started it?
- No, you can't cancel an upgrade once it has started. If your upgrade fails, Cloud SQL automatically recovers your instance on the previous version.
- What happens to my settings during an upgrade?
When you perform an in-place major version upgrade, Cloud SQL retains your database settings, including your instance name, IP address, explicitly configured flag values, and user data. However, the default value of the system variables might change. For example, the default value of the
password_encryption
flag in PostgreSQL 13 and earlier ismd5
. When you upgrade to PostgreSQL 14, the default value of this flag changes toscram-sha-256
.To learn more, see Configure database flags. If a certain flag or value is no longer supported in your target version, then Cloud SQL automatically removes the flag during the upgrade.
What's next
- Learn about options for connecting to an instance.
- Learn about importing and exporting data.
- Learn more about [setting database flags][flags].