Setting Up PostgreSQL with Bitnami Helm Chart

First, let’s install a PostgreSQL database using the Bitnami Helm chart:

# Create a namespace for our PostgreSQL instance
kubectl create namespace postgres-demo

# Install PostgreSQL using Helm
helm install postgres16 oci://registry-1.docker.io/bitnamicharts/postgresql \
  --namespace postgres-demo \
  --set auth.postgresPassword=your-secure-password

Verify that the PostgreSQL pod is running:

kubectl get pods -n postgres-demo

Creating a Database with Sample Data

Connect to the PostgreSQL instance and create a sample database with tables:

kubectl exec -n postgres-demo -it postgres16-postgresql-0 -- bash
export PGHOST='localhost'
export PGUSER='postgres'
export PGPASSWORD='your-secure-password'
psql 
CREATE DATABASE testdb;
\c testdb
CREATE TABLE customers (id SERIAL PRIMARY KEY, name VARCHAR(100), email VARCHAR(100));
INSERT INTO customers (name, email) VALUES ('John Doe', 'john@example.com'), ('Jane Smith', 'jane@example.com');
SELECT * FROM customers;

Custom Blueprint for Granular Backup

Kasten K10 uses Kanister blueprints to define how applications should be backed up. For PostgreSQL, we’ll use a custom blueprint that creates individual binary dumps for each database, enabling granular restore capabilities:

apiVersion: cr.kanister.io/v1alpha1
kind: Blueprint
metadata:
  name: postgres-pg-dump
  namespace: kasten-io
actions:
  beforeBackup:
    kind: StatefulSet
    phases:
    - name: pgDumpall
      func: KubeExec
      objects:
        pgSecret:
          kind: Secret
          name: 'postgres16-postgresql'
          namespace: '{{ .StatefulSet.Namespace }}'
      args:
        pod: "{{ index .StatefulSet.Pods 0 }}"
        namespace: '{{ .StatefulSet.Namespace }}'
        command:
        - bash
        - -o
        - errexit
        - -o
        - pipefail
        - -c
        - |
          export PGHOST='localhost'
          export PGUSER='postgres'
          export PGPASSWORD='{{ index .Phases.pgDumpall.Secrets.pgSecret.Data "postgres-password" | toString }}'
          mkdir -p /bitnami/postgresql/backups
          
          # Back up global objects (roles, tablespaces) in plain text
          pg_dumpall --globals-only > /bitnami/postgresql/backups/globals.sql
          
          # Get a list of user databases (excluding system databases)
          databases=$(psql -U postgres -t -c "SELECT datname FROM pg_database WHERE datname NOT IN ('template0', 'template1', 'postgres')")

          # Back up each database in binary format
          for db in $databases; do
            pg_dump -U postgres -Fc $db > /bitnami/postgresql/backups/${db}.dump
          done

Apply this blueprint to your cluster:

Understanding the Blueprint

The blueprint performs several key actions:

• Connects to PostgreSQL using credentials from the PostgreSQL secret
• Creates separate dumps for global objects (users, roles) in plain text format
• Identifies all user databases by querying pg_database
• Creates individual binary dumps (using the -Fc custom format) for each database

The binary format (-Fc) is critical because it allows us to perform selective table restores later using the pg_restore tool, unlike plain SQL dumps that would require parsing.

Creating a Backup Policy in Kasten K10

With our blueprint in place, create a backup policy in Kasten K10 that uses it:

1. In the Kasten K10 dashboard, navigate to Policies and click Create New Policy
2. Name your policy (e.g., “PostgreSQL-Granular-Backup”)
3. Select your PostgreSQL application as the target
4. Choose an export target this is madatory for volume clone restore
5. Under Advanced Settings > Action Hooks, select your custom blueprint and the action beforeBackup
6. Choose the workload stefulset/postgres16-postgresql

Create and run the policy.

Performing Granular Table Restore

When you need to restore a single table, Kasten K10 performs a volume clone restore that creates a new PVC with the backup date as a suffix. Here’s how to leverage this for granular restores:

1. Run a restore operation from Kasten K10.
   • Navigate to the restpoint you want to restore
   • Select “Restore” and choose volume clone restore
   • Select the pvc
   • Click restore

You’ll see a new PVC with a name like data-postgres16-postgresql-0-2025-05-28-15-13-41

1. Patch the StatefulSet to mount the new PVC temporarily:

kubectl patch statefulset postgres16-postgresql -n postgres-demo --patch '
{
  "spec": {
    "template": {
      "spec": {
        "containers": [{
          "name": "postgresql",
          "volumeMounts": [{
            "name": "restore-volume",
            "mountPath": "/2025-05-28-15-13-41"
          }]
        }],
        "volumes": [{
          "name": "restore-volume",
          "persistentVolumeClaim": {
            "claimName": "data-postgres16-postgresql-0-2025-05-28-15-13-41"
          }
        }]
      }
    }
  }
}'

1. Restore a specific table from the binary dump:

kubectl exec -it postgres16-postgresql-0 -n postgres-demo -- bash
# Inside the pod, restore only the customers table
export PGHOST='localhost'
export PGUSER='postgres'
export PGPASSWORD='your-secure-password'
pg_restore -U postgres -d testdb --table=customers --clean /2025-05-28-15-13-41/backups/testdb.dump

1. Verify the restore:

psql -d testdb -c "SELECT * FROM customers;"

1. Clean up by removing the temporary volume mount:

   kubectl rollout undo statefulset/postgres16-postgresql -n postgres-demo
   

Benefits of This Approach

This granular restore approach offers several advantages:

1. Minimal downtime - only restore what you need without affecting other databases/tables
2. Resource efficiency - no need to restore the entire database for a single table
3. Flexibility - can restore tables across environments (development/testing)
4. Data integrity - binary dumps maintain data types and constraints accurately

Conclusion

By combining Kasten K10’s volume cloning capabilities with PostgreSQL’s binary dump format, we’ve enabled granular table-level restores that minimize downtime and maximize flexibility during recovery scenarios. This approach gives database administrators the precision tools they need for efficient recovery operations.