The Klio Plugin v0.0.12

The Klio plugin for CloudNativePG allows you to leverage the backup and WAL streaming capabilities of Klio for your PostgreSQL clusters managed by CloudNativePG. It will add two containers to each PostgreSQL instance pod:

• A klio-plugin container that handles backup creation and management
• A klio-wal container that streams WAL files to the Klio server in real-time

Configuration

The Klio plugin integrates with CloudNativePG through the CNPG-I (CloudNativePG Interface) specification, enabling Klio to manage backups and WAL streaming for your PostgreSQL clusters. To use Klio with a CloudNativePG cluster, you need to:

1. Create a PluginConfiguration resource that defines how to connect to the Klio server
2. Reference the plugin in your Cluster resource specification

Prerequisites

Before configuring a cluster to use the Klio plugin, ensure you have:

• A running Klio Server resource deployed in your namespace
• Client credentials (username and password) stored in a Kubernetes Secret
• The server's TLS certificate available in a Secret

Creating a PluginConfiguration resource

The PluginConfiguration custom resource defines how the Klio plugin connects to and communicates with the Klio server. This resource contains connection details, authentication credentials, and optional configuration for metrics, profiling, and backup retention policies.

Basic example

Here's a minimal PluginConfiguration example:

apiVersion: klio.enterprisedb.io/v1alpha1
kind: PluginConfiguration
metadata:
  name: klio-plugin-config
  namespace: default
spec:
  serverAddress: klio-server.default
  clientSecretName: client-sample-tls
  serverSecretName: klio-server-tls
  # mode: standard  # Optional: standard (default) or read-only

Client credentials secret

The client credentials must be stored in a Kubernetes Secret of type kubernetes.io/tls, containing a secret to be presented to the Klio server.

This secret can be generated with cert-manager by following the documentation in the Klio server page, or be provided by the user.

Server Address

The serverAddress field specifies where the Klio server can be reached. This can be:

• A Kubernetes service name: klio-server.default (within the same namespace)
• A fully qualified domain name: klio-server.default.svc.cluster.local
• An external address: klio.example.com

Connections will be done using the default ports of the Klio base and WAL servers, respectively 51515 and 52000.

TLS configuration

The serverSecretName field references a Secret containing the TLS certificate used to secure communication with the Klio server. This is the same certificate configured on the Server resource.

Configuring a Cluster to use the Klio plugin

Once you have created a PluginConfiguration, reference it in your CloudNativePG Cluster resource:

apiVersion: postgresql.cnpg.io/v1
kind: Cluster
metadata:
  name: my-postgres-cluster
  namespace: default
spec:
  instances: 3

  postgresql:
    pg_hba:
      - local replication all peer # Allow replication connections locally

  plugins:
    - name: klio.enterprisedb.io
      enabled: true # Activate the Klio plugin (default)
      parameters:
        pluginConfigurationRef: klio-plugin-config

  storage:
    size: 10Gi

To be able to stream WAL files, ensure that your PostgreSQL configuration allows local replication connections. You can do this by adding an entry to the pg_hba section, as shown in the example above.

Plugin parameters

The plugins section in the Cluster specification requires:

• name: Must be set to klio.enterprisedb.io to identify the Klio plugin
• enabled: Set to true to activate the plugin. This is the default value.
• parameters.pluginConfigurationRef: The name of your PluginConfiguration resource

Advanced configuration options

The PluginConfiguration resource supports several advanced options to customize the plugin's behavior.

Retention policies

Define how long backups should be retained by configuring retention policies for Tier 1 and Tier 2 storage. Retention policies can be configured independently for each tier:

apiVersion: klio.enterprisedb.io/v1alpha1
kind: PluginConfiguration
metadata:
  name: klio-plugin-config
spec:
  serverAddress: klio-server.default
  clientSecretName: klio-client-credentials
  serverSecretName: klio-server-tls
  tier1:
    retention:
      keepLatest: 5
      keepHourly: 12
      keepDaily: 7
      keepWeekly: 4
      keepMonthly: 6
      keepAnnual: 2
  tier2:
    enableBackup: true
    enableRecovery: true
    retention:
      keepLatest: 10
      keepDaily: 30
      keepMonthly: 12
      keepAnnual: 5

Except for keepLatest, each option defines how many backups to retain for the specified time period. For example, keepDaily: 7 means that we should retain at most one backup for each of the past 7 days.

If multiple backups exist within the same time bucket, the most recent one is kept, unless preserved by a different keep rule. Backups that are not retained by any rule are deleted. Rule evaluation is done when a new backup is taken.

The Klio server will automatically delete WAL files that are no longer needed for recovery by any retained backup.

All retention settings are optional. For each unspecified retention level, the default Kopia value is applied:

keepLatest: 10
keepHourly: 48
keepDaily: 7
keepWeekly: 4
keepMonthly: 24
keepAnnual: 1

Set a rule to 0 to disable that retention level.

Operation Mode

The mode field controls whether the plugin can perform both backup and restore operations (standard) or only restore operations (read-only).

spec:
  mode: standard  # or read-only

Standard Mode (default)

In standard mode, the plugin can:

• Create backups to Tier 1 (and optionally Tier 2)
• Stream WAL files to the server
• Restore from both Tier 1 and Tier 2

This is the default mode and suitable for most deployments.

Read-Only Mode

Use read-only mode when connecting to a read-only Klio server for recovery operations only. This is useful for disaster recovery scenarios where you want to restore from a read-only server in a secondary region or datacenter.

apiVersion: klio.enterprisedb.io/v1alpha1
kind: PluginConfiguration
metadata:
  name: dr-restore-config
spec:
  serverAddress: dr-server.default
  clientSecretName: dr-client-credentials
  serverSecretName: dr-server-tls
  mode: read-only

  # Read-only mode requires:
  # - tier2 configuration with enableRecovery: true
  # - NO tier1 configuration
  tier2:
    enableRecovery: true
    # enableBackup must be false or omitted

Restrictions in read-only mode:

• Tier 2 must be configured with enableRecovery: true
• Tier 2 enableBackup must be false or omitted
• Tier 1 configuration is not allowed
• No backup or WAL streaming operations are performed

See the Read-Only Server Mode documentation for details on setting up a read-only Klio server.

Cluster name override

By default, the plugin uses the name of the CloudNativePG Cluster resource. You can override this if needed:

spec:
  clusterName: my-custom-cluster-name

This can be useful working with backups from different clusters, for example when restoring clusters or configuring replica clusters.

Tier 2 configuration

Tier 2 provides secondary storage (typically object storage like S3) for long-term backup retention and disaster recovery. Configure Tier 2 using the tier2 section:

spec:
  tier2:
    enableBackup: true
    enableRecovery: true
    retention:
      keepDaily: 30
      keepMonthly: 12

Options

• enableBackup: When set to true, backups and WAL files are automatically synchronized to Tier 2 storage after being stored in Tier 1. This ensures your backups are available in long-term storage.

• enableRecovery: When set to true, Klio will look for backups and WAL files in both Tier 1 and Tier 2 during restore operations. If a backup is available in both tiers, Tier 1 takes precedence as restore from it will be faster.

• retention: Configure a separate retention policy for Tier 2. Typically, you would configure longer retention periods for Tier 2 since object storage is more cost-effective for long-term storage.

See the Architecture documentation for more details on Tier 2 storage.

Observability

See the OpenTelemetry observability section for more details on how to monitor the Klio plugin using OpenTelemetry.

Performance profiling

Enable the pprof HTTP endpoint for performance profiling and troubleshooting:

spec:
  pprof: true

When enabled, the pprof endpoint is exposed and can be used with Go's profiling tools to analyze CPU usage, memory allocation, goroutines, and other runtime metrics.

Container customization

The PluginConfiguration resource allows you to customize the Klio sidecar containers by providing base container specifications that are used as the foundation for the sidecars. This feature enables you to add custom environment variables, volume mounts, resource limits, and other container settings without modifying the PostgreSQL container environment.

Basic example

apiVersion: klio.enterprisedb.io/v1alpha1
kind: PluginConfiguration
metadata:
  name: klio-plugin-config
spec:
  serverAddress: klio-server.default
  clientSecretName: klio-client-credentials
  serverSecretName: klio-server-tls
  containers:
    - name: klio-plugin
      env:
        - name: CUSTOM_ENV_VAR
          value: "my-value"
        - name: DEBUG_LEVEL
          value: "info"
    - name: klio-wal
      env:
        - name: WAL_BUFFER_SIZE
          value: "8192"

How container merging works

The containers you define serve as the base for the Klio sidecars, with the following merge behavior:

1. Your container is the base: When you define a container (e.g., klio-plugin), your specification serves as the starting point
2. Klio enforces required values: Klio sets its essential configuration:
   • Container name (klio-plugin, klio-wal, or klio-restore)
   • Container args (the command arguments needed for operation)
   • CONTAINER_NAME environment variable
3. Your customizations are preserved: All other fields you define remain intact
4. Template defaults fill gaps: For fields you don't specify, Klio applies sensible defaults (image, security context, standard volume mounts, etc.)

Available sidecar containers

The following containers can be customized:

• klio-plugin: Handles backup creation and management in PostgreSQL pods
• klio-wal: Streams WAL files to the Klio server in PostgreSQL pods
• klio-restore: Restores backups during recovery jobs

Example: Resource limits and environment variables

apiVersion: klio.enterprisedb.io/v1alpha1
kind: PluginConfiguration
metadata:
  name: klio-plugin-config
spec:
  serverAddress: klio-server.default
  clientSecretName: klio-client-credentials
  serverSecretName: klio-server-tls
  containers:
    - name: klio-plugin
      env:
        - name: LOG_LEVEL
          value: "debug"
        - name: OTEL_EXPORTER_OTLP_ENDPOINT
          value: "http://otel-collector:4317"
      resources:
        limits:
          memory: "512Mi"
          cpu: "1"
        requests:
          memory: "256Mi"
          cpu: "500m"
    - name: klio-wal
      env:
        - name: WAL_STREAM_TIMEOUT
          value: "30s"
      resources:
        limits:
          memory: "256Mi"
          cpu: "500m"
        requests:
          memory: "128Mi"
          cpu: "250m"