disruptionBudget maxUnavailable value is ignored if 0

[topfunky]

What I did:

I set my pod disruption budget's maxUnavailable to 0 in my Helm values override file. I then installed the helm chart to a minikube cluster.

What I saw:

When running helm install, I saw an error that told me the maxUnavailable had been set to -1, not the 0 that I had used.

$ helm install -f helm-consul-values.yaml ./consul-helm

Error: release pouring-monkey failed: PodDisruptionBudget.policy "pouring-monkey-consul
-server" is invalid: spec.maxUnavailable: Invalid value: -1: must be greater than or eq
ual to 0

What I expected:

I expected that setting maxUnavailable to 0 would be interpreted by the chart as a valid value.

Other context:

If I set the value to 1, everything works as expected. This suggests that a piece of logic is interpreting 0 as false and fails to use the value. Values 1 and greater are interpreted correctly.

My Helm values file has (this is the config that causes the above error):

server:
  replicas: 1
  bootstrapExpect: 1
  disruptionBudget:
    enabled: true
    maxUnavailable: 0

The issue may be related to code in _helpers.tpl which checks if .Values.server.disruptionBudget.maxUnavailable but should look for if .Values.server.disruptionBudget.enabled.

[adilyse]

I've addressed the invalid -1 value in this PR. There is unfortunately still a larger issue of not being able to explicitly set this value to 0 that is a limitation of Helm templating (see PR for details).

[adilyse]

I've addressed this with the following changes:

• I've changed the calculation of the default maxUnavailableto provide a valid value when replicas=1. This should fix the issue in the vast majoring of cases.

• I've added documentation about how someone can directly set this value to 0, since it's more involved than it should be.

Hopefully that helps!