Manage PostgreSQL extensions¶

One of the specific PostgreSQL features is the ability to provide it with additional functionality via Extensions . Percona Distribution for PostgreSQL comes with a number of extensions . These extensions are available for the database cluster managed by the Operator as well.

Built-in extensions¶

You can enable or disable built-in extensions in the extensions.builtin section of your deploy/cr.yaml file. Set an option to true to enable an extension, or to false to disable it. To see which extensions are enabled by default, check the deploy/cr.yaml Custom Resource manifest.

extensions:
  ...
  builtin:
    pg_stat_monitor: true
    pg_audit: true
    pgvector: false
    pg_repack: false

Apply changes after editing with kubectl apply -f deploy/cr.yaml command. This causes the Operator to restart the Pods of your cluster.

Add custom extensions¶

The needed extension may not be in the list of extensions supplied with Percona Distribution for PostgreSQL, or it’s a custom extension developed by the end-user. To add such a custom extension is not straightforward in a containerized database in a Kubernetes environment. It requires building a custom PostgreSQL image.

Starting with version 2.3, the Operator provides an alternative way to extend Percona Distribution for PostgreSQL by downloading pre-packaged extensions from and external storage on the fly.

Advanced configuration

Custom extensions configuration is an advanced feature that requires careful consideration. Adding custom extensions may violate the immutability of Pod images, which can lead to unexpected behavior and maintenance challenges. Use this feature only if you are certain what you are doing and understand the implications. Or reach out to our experts for assistance with adding custom extensions into your infrastructure.

Here’s how it works:

You build and package a custom extension. The package must have a strict structure. See Packaging requirements for details.
You upload the extension to a cloud storage.
In the extensions section of the Custom Resource, specify the storage configuration and the extension information.
The Operator downloads the extension and installs it.
In PostgreSQL, you create the extension for every database where you want to use it.

Understanding which files are required for a given extension may not be easy. To figure this out, you can spin up a Docker container or a virtual machine, install Percona Distribution for PostgreSQL and developer tools there, then build and install the extension from source. Then copy all the installed files to the archive.

Check the Example configuration for the steps that can help you in building and adding your own custom extension.

Packaging requirements¶

Custom extensions require specific packaging for the Operator to use them. The package must be a .tar.gz archive that follows this naming format:

${EXTENSION}-pg${PG_MAJOR}-${EXTENSION_VERSION}

The archive must be created with usr at the root and must include all the required files in the correct directory structure:

The control file and any shared library must be in the LIBDIR directory
All required SQL script files must be in the SHAREDIR/extension directory. At least one SQL script is required.

The SHAREDIR corresponds to /usr/pgsql-${PG_MAJOR}/share and LIBDIR to /usr/pgsql-${PG_MAJOR}/lib.

For example, the directory for pg_cron extension should look as follows:

tree ~/pg_cron-1.6.7/
/home/user/pg_cron-1.6.7/
└── usr
    └── pgsql-17
        ├── lib
        │   └── pg_cron.so
        └── share
            └── extension
                ├── pg_cron--1.0--1.1.sql
                ├── pg_cron--1.0.sql
                ├── pg_cron--1.1--1.2.sql
                ├── pg_cron--1.2--1.3.sql
                ├── pg_cron--1.3--1.4.sql
                ├── pg_cron--1.4--1.4-1.sql
                ├── pg_cron--1.4-1--1.5.sql
                ├── pg_cron--1.5--1.6.sql
                └── pg_cron.control

The resulting .tar archive has the name pg_cron-pg17-1.6.7.tar.gz.

Example configuration¶

The following is an example workflow showing how to build and package the pg_cron extension. This example is intended to illustrate the general process and give you an idea of the required steps. However, the exact workflow and specifics may differ for your custom extension. Always review your extension’s build and packaging requirements and adapt accordingly.

Considerations¶

You must build your extension on a host with the same operating system and architecture as the one used for Percona Distribution for PostgreSQL images to prevent library incompatibility. Otherwise, your extension may not load or may not function correctly.

To check the operating system, do the following:

Connect to one of the database Pods:

kubectl exec -it cluster1-instance1-xrcf-0 -n <namespace> -c database -- bash

List the installed packages:

rpm -qa|grep percona

Sample output

percona-release-1.0-32.noarch
percona-postgresql17-libs-17.6-1.el9.x86_64
percona-postgresql17-17.6-1.el9.x86_64
percona-postgresql-client-common-280-1.el9.noarch
percona-telemetry-agent-1.0.5-1.el9.x86_64
percona-pg-telemetry17-1.1.3-1.el9.x86_64
percona-postgresql17-server-17.6-1.el9.x86_64
percona-pgbackrest-2.56.0-1.el9.x86_64
percona-pg_stat_monitor17-2.2.0-1.el9.x86_64
percona-pgaudit17-17.1-1.el9.x86_64
percona-pgvector_17-0.8.0-3.el9.x86_64
percona-wal2json17-2.6-1.el9.x86_64
percona-postgresql17-contrib-17.6-1.el9.x86_64
percona-postgresql-common-280-1.el9.noarch
percona-pg_repack17-1.5.2-2.el9.x86_64
percona-pgaudit17_set_user-4.1.0-3.el9.x86_64
percona-patroni-4.0.6-1.el9.x86_64

Check the operating system version:

cat /etc/redhat-release

Sample output

Red Hat Enterprise Linux release 9.6 (Plow)

Your extension must be compatible with PostgreSQL version you are running. To check the version, run the following command:

kubectl -n <namespace> get pg cluster1 -o go-template='{{.spec.image}}'

Sample output

docker.io/perconalab/percona-postgresql-operator:main-ppg17-postgres

In this example configuration, we use a Docker container to build the pg_cron extension. However, you can use any environment that matches the distribution’s operating system, such as a virtual machine or a Kubernetes Pod, not just Docker.
We assume you have deployed a Percona Distribution for PostgreSQL cluster in Kubernetes. If not, use the Quickstart guide to deploy it.

Prepare your build environment¶

Run the following commands as the root user or with sudo privileges.

Start a Docker container and establish a shell session inside. In this example we use a Red Hat Universal Base Image 9 on x86_64 architecture.
```
docker run -it --name pg redhat/ubi9:latest /bin/bash
```

Install basic tools:

dnf install git make 'dnf-command(config-manager)'

Install additional PostgreSQL packages:
- Add the Extra Packages for Enterprise Linux by installing the epel-release package:
```
dnf install -y https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm
```
- Add the codeready builder repository that contains additional packages for use by developers:
```
dnf config-manager --add-repo https://dl.rockylinux.org/pub/rocky/9/CRB/x86_64/os/
```
- Import GPG keys
```
rpm --import https://dl.rockylinux.org/pub/rocky/RPM-GPG-KEY-Rocky-9
```
- Install perl-IPC-Run to run and interact with child processes:
```
dnf install perl-IPC-Run -y
```

Install build tools:

dnf groupinstall "Development tools"

Troubleshooting tip: If development tools fail to install, add BaseOS and AppStream repos:

dnf config-manager --add-repo https://dl.rockylinux.org/pub/rocky/9/BaseOS/aarch64/os/
dnf config-manager --add-repo https://dl.rockylinux.org/pub/rocky/9/AppStream/aarch64/os/
dnf clean all && dnf makecache

Then retry the installation.

Install PostgreSQL developer packages from Percona repositories:
- Install percona-release repository management tool:
```
dnf install https://repo.percona.com/yum/percona-release-latest.noarch.rpm
```
- Enable PostgreSQL repository:
```
percona-release setup ppg17
```
- Disable the potsgresql module supplied with the operating system:
```
dnf -qy module disable postgresql
```
- Install PostgreSQL developer packages:
```
dnf install percona-postgresql17-devel percona-postgresql17-libs percona-postgresql17
```

Build the extension¶

Download the extension source:

git clone https://github.com/citusdata/pg_cron.git

Navigate to the cloned extension and switch to the desired version. In this example we use version 1.6.7:
```
cd pg_cron
git checkout v1.6.7
```
Ensure pg_config is in your path:
```
export PATH=/usr/pgsql-17/bin:$PATH
```
Build and install the extension
```
make && sudo PATH=$PATH make install
```
As the result you should see the binaries in the following paths: /usr/pgsql-17/share/extension/pg_cron and /usr/pgsql-17/lib/.

Package the extension¶

Create a .tar archive of the extension:

tar -czvf pg_cron-pg17-1.6.7.tar.gz \
  /usr/pgsql-17/lib/pg_cron.so \
  /usr/pgsql-17/share/extension/pg_cron*

Check that the package structure follows the requirements.
Copy the archive to the local machine. Run this command on the local machine:
```
docker cp pg:/pg_cron-pg17-1.6.7.tar.gz ./
```

Upload a custom extension to the cloud storage¶

After packaging the extension, upload it to a cloud storage. In our example we use AWS S3 storage. You can upload the extension via the Amazon web interface or using the aws command line tool as shown below:

Export the AWS S3 access credentials as the environment variables:

export AWS_ACCESS_KEY_ID=<your-access-key-id-here> 
export AWS_SECRET_ACCESS_KEY=<your-secret-key-here>

Upload the extension to your storage. Use your value for the bucket and specify your path to the archive:

aws s3 cp path/to/pg_cron-pg17-1.6.7.tar.gz s3://my-bucket

Create a Secret with the storage credentials¶

After the upload is complete, place the access credentials for the cloud storage in a Secret.

Create a Secrets file with the credentials that the Operator needs to access extensions stored on Amazon S3:
- The metadata.name key is the name you will use to refer to your Kubernetes Secret.
- The data.AWS_ACCESS_KEY_ID and data.AWS_SECRET_ACCESS_KEY keys contain base64-encoded credentials used to access the storage.
  
  To encode credentials, use this command:
  in Linuxin macOS
  For GNU/Linux:
  
  echo -n 'plain-text-string' | base64 --wrap=0
  For Apple macOS:
  
  echo -n 'plain-text-string' | base64
Here’s the example Secrets file extensions-secret.yaml:
extensions-secret.yaml
```
apiVersion: v1
kind: Secret
metadata:
  name: cluster1-extensions-secret
type: Opaque
data:
  AWS_ACCESS_KEY_ID: <base64 encoded secret>
  AWS_SECRET_ACCESS_KEY: <base64 encoded secret>
```

Create the Secrets object from this file:

kubectl apply -f extensions-secret.yaml -n <namespace>

Configure the Operator to load and install the custom extension¶

Specify both the storage and extension details in the Custom Resource so the Operator can download and install it.

In the extensions.storage subsection of the Custom Resource, specify the following information:
- storage details such as the bucket where your extension resides, region and endpoint to access the storage
- the Secret name with the storage credentials that you created before.
```
extensions:
  ...
  storage:
    type: s3
    bucket: pg-extensions
    region: eu-central-1
    endpoint: s3.eu-central-1.amazonaws.com
    secret:
      name: cluster1-extensions-secret
```

In the extensions.custom subsection, specify the extension name and version:

extensions:
  ...
  custom:
  - name: pg_cron
    version: 1.6.1

Some extensions (such as pg_cron in our example) may require additional shared memory. If this is the case, you need to configure PostgreSQL to preload it at startup:

```yaml … patroni: dynamicConfiguration: postgresql: parameters: shared_preload_libraries: pg_cron …
Apply the configuration:
```
kubectl apply -f deploy/cr.yaml -n <namespace>
```
This causes the Operator to restart the Pods of your cluster.

Enable custom extension in PostgreSQL¶

The installed extension is not enabled by default. You need to explicitly enable it in PostgreSQL for all databases where you want to use it.

Here’s how to do it:

Connect to the primary Pod:

kubectl exec -it cluster1-instance1-69r8-0 -c database -n <namespace> -- bash

Connect to the required database in PostgreSQL and create the extension for this database using the CREATE EXTENSION statement:
```
CREATE EXTENSION pg_cron;
```

Update custom extensions¶

To update your custom extension inside the Operator, do the following:

Prepare the *.tar archive of the extension’s new version. See the Packaging requirements section for the archive’s structure and naming format
Reference the new version of the extension in the Custom Resource. For example, you update pg_cron extension to version 1.6.8. Then your configuration looks like this:
```
extensions:
  ...
  custom:
  - name: pg_cron
    version: 1.6.8
```
Apply the configuration for the changes to come into place:
```
kubectl apply -f deploy/cr.yaml -n <namespace>
```

Last update: 2025-11-18