Upgrade Advisory

This documentation is for Flux (v1) and Helm Operator (v1). Both projects are in maintenance mode and will soon reach end-of-life. We strongly recommend you familiarise yourself with the newest Flux and start looking at your migration path.

For documentation regarding the latest Flux, please refer to this section.

Rollbacks

From time to time a release made by the Helm Operator may fail, this section of the guide will explain how you can recover from a failed release by enabling rollbacks.

Enabling rollbacks

When rollbacks for a HelmRelease are enabled, the Helm Operator will detect a faulty upgrade, including post-upgrade helm test if enabled failures, and instruct Helm to perform a rollback, it will not attempt a new upgrade unless it detects a change in values and/or the chart, or retries have been enabled. Changes are detected by comparing the failed release to a fresh dry-run release.

Rollbacks can be enabled by setting .rollback.enable:

spec:
  rollback:
    enable: true

Wait interaction

When rollbacks are enabled, resource waiting defaults to true since this is necessary to validate whether the release should be rolled back or not.

Tweaking the rollback configuration

To get more fine-grained control over how the rollback is performed by Helm, the .rollback of the HelmRelease resources offers a couple of additional settings.

spec:
  rollback:
    enable: true
    disableHooks: false
    force: false
    recreate: false
    timeout: 300

The definition of the listed keys is as follows:

  • enable: Enables the performance of a rollback when a release fails.
  • disableHooks (Optional): When set to true, prevent hooks from running during rollback. Defaults to false when omitted.
  • force (Optional): When set to true, force resource update through delete/recreate if needed. Defaults to false when omitted.
  • recreate (Optional): When set to true, performs pods restart for the resource if applicable. Defaults to false when omitted.
  • timeout (Optional): Time to wait for any individual Kubernetes operation during rollback in seconds. Defaults to 300 when omitted.

Enabling retries of rolled back releases

Sometimes the cause of an upgrade failure may be transient. To guard yourself against this it is possible to instruct the Helm Operator to retry the upgrade of a rolled back release by setting .rollback.retry to true. This will cause the Helm Operator to retry the upgrade until the .rollback.maxRetries is reached:

spec:
  rollback:
    enable: true
    retry: true
    maxRetries: 5

The definition of the listed keys is as follows:

  • enable: Enables the performance of a rollback when a release fails.
  • retry (Optional): When set to true, retries the upgrade of a failed release until maxRetries is reached. Defaults to false when omitted.
  • maxRetries (Optional): The maximum amount of retries that should be attempted for a rolled back release. Defaults to 5 when omitted, use 0 for an unlimited amount of retries.
Last modified 2021-06-03: Integrate helm-op docs properly (d5756e4)