This is a simple Helm chart and Ansible playbook to demonstrate Helm --timeout being used without --wait to accommodate long-running pre-chart hooks.
The chart simply consists of a pre-chart batch job that sleeps for six minutes and then terminates.
This is long enough to trigger Helm's default five minute timeout, causing it to fail.
If --timeout 6m is provided, the chart successfully installs.
https://helm.sh/docs/intro/using_helm/#helpful-options-for-installupgraderollback
You will need:
- Ansible
- Helm
- Kubernetes cluster
If the kubernetes.core collection is not already installed, run ansible-galaxy collection install -r requirements.yml.
This demo assumes the Kubernetes cluster has been configured in ~/.kube/config and will install the chart into the default namespace.
This will fail because the job takes six minutes to terminate and Helm's default duration is 5m.
sh helm-defaults-fails.shThis will succeed because the timeout has been set to 10m, but the job finishes in six.
sh helm-w-timeout-succeeds.shThis will fail because kubernetes.core.helm only adds --timeout, based on the wait_timeout module option, if wait: yes is specified.
sh ansible-helm-timeout-ignored.sh This will succeed wait has been enabled.
However, there are situations where wait is undesirable, such as an application that has a very long startup process.
Helm's --wait will block waiting for all resources to become ready.
sh ansible-helm-timeout-honored.sh