denis256 notes and projects

Apache Airflow Helm Chart 1.22.0 was released on June 13, 2026. This version bumps the default Airflow image to 3.2.2 and sets a new minimum requirement for the Helm binary itself.

The full release notes and downloads are on the GitHub release page.

Helm and Image Requirements

Operators using this chart must first verify their local environment before performing an upgrade. The minimum Helm version is now 3.19.0. This change is important because older versions of the Helm client might lack the necessary features to parse the updated templates or handle the complexity of the Airflow 3 architecture. If you run older versions of the Helm binary, you might see cryptic rendering errors during the install or upgrade process.

The default Airflow image is now updated to 3.2.2. This is a minor bump from the previous version 3.2.1. Most users will want this update to get the latest bug fixes and security patches that the Airflow team has shipped recently. While you can still override the image in your local values file, the chart is now tested and optimized against this specific 3.2.2 release. Using the default image ensures that all the internal paths and environment variables match the expectations of the chart maintainers.

Service Links and Networking

A significant change in this release is the support for configuring enableServiceLinks. This is a standard Kubernetes feature that automatically injects environment variables for every service in the namespace into every pod. While this was helpful in the early days of Kubernetes, it often creates a very noisy environment in namespaces with many services.

The maintainers have announced that enableServiceLinks will default to false in the upcoming Chart 2.0 release. In the current 1.22.0 version, you now have the option to control this behavior. If your custom DAGs or internal tools rely on these environment variables to find other services, you should explicitly set enableServiceLinks: true in your configuration now. However, the recommended path is to migrate your logic to use DNS based service lookups. DNS is more stable and does not depend on the order in which services and pods are created. It also keeps your pod environment cleaner and prevents issues where a large number of services might exceed the shell limit for environment variables.

New Deployment and Monitoring Options

Several new features were added to this release to help with observability and resource management in complex clusters.

• OpenTelemetry Support: The chart now includes an optional OTel service. This addition allows operators to integrate Airflow with modern observability stacks more easily. By using the OTel service, you can export metrics and traces to collectors like Prometheus or Jaeger without having to manually patch the deployment with sidecars or complex configuration overrides.
• Service Account Token Volumes: For the database cleanup cron job, you can now specify a serviceAccountTokenVolume. This is particularly useful for users running in hardened Kubernetes environments where the default service account token projection is disabled or restricted. It allows the cleanup job to authenticate against the Kubernetes API or other services as needed.
• Worker Log Persistence: A new requirePersistence option is available for the worker logGroomerSidecar. This setting ensures that logs are preserved correctly. If the worker pod is evicted or restarted, having persistence requirements prevents data loss before the logs can be shipped to long term storage or an external logging service.

Critical Bug Fixes and Refactors

The 1.22.0 release includes a long list of fixes that improve the stability of the Airflow components.

One major fix addresses the Helm chart executor label to support executor aliases. This is vital for users who use custom executor class paths or mixed executor setups. Previously, the label might not have correctly identified the executor type, leading to issues with how pods were scheduled or grouped.

The triggerer component received a fix for KEDA database connection rendering. This ensures that the triggerer can scale correctly based on external events when using KEDA as the autoscaler. Without this fix, the connection string might not render properly in certain configurations, causing the KEDA scaler to fail.

Other notable fixes include:

• Networking: Fixes for task log access when using NetworkPolicies. This ensures that logs can be reached even in restricted network environments for both Airflow 2 and Airflow 3 deployments.
• RBAC: Improved launcher RBAC rules for executor class paths. This allows the launcher to correctly manage the permissions needed for different executor types.
• Templates: Standardizing on Dag capitalization throughout the chart wording. While this is a small change, it helps with consistency across the documentation and labels.
• Lifecycle Hooks: Fixes for database cleanup and pod lifecycle hooks. These hooks are now applied correctly, ensuring that preStop and postStart actions are executed as expected.

Where to get it

You can download the Apache Airflow Helm Chart 1.22.0 and read the full technical notes at the following locations:

• The official GitHub release page
• The Airflow project repository
• Use the tag helm-chart/1.22.0 when installing via Helm

This update is a stable release. All operators are encouraged to upgrade, especially if they are moving toward Airflow 3 or require better observability through OpenTelemetry.