Cannot unmarshal string into Go struct field Config.JobCleanupDuration of type time.Duration

[adriananeci]

Describe the bug

Kuberhealthy cannot parse the configmap when jobCleanupDuration option is defined because it cannot unmarshal string into Go struct field Config.JobCleanupDuration of type time.Duration

time="2021-02-25T05:12:29Z" level=info msg="WARNING: Failed to read configuration file from disk: error unmarshaling JSON: while decoding JSON: json: cannot unmarshal string into Go struct field Config.JobCleanupDuration of type time.Duration"

Steps To Reproduce

• Start kuberhealthy with the following associated configmap

---
apiVersion: v1
kind: ConfigMap
metadata:
  name: kuberhealthy
  namespace: test
data:
  kuberhealthy.yaml: |-
    listenAddress: ":8080" # The port for kuberhealthy to listen on for web requests
    enableForceMaster: false # Set to true to enable local testing, forced master mode
    logLevel: "debug" # Log level to be used
    listenNamespace: "test" # Kuberhealthy will only monitor khcheck resources from this namespace
    jobCleanupDuration: 3m # The maximum age of khjobs before being reaped
    maxCheckPods: 1 # The maximum number of check pods in Completed state before being reaped

• Check the logs and you'll see a warning like

time="2021-02-25T05:12:29Z" level=info msg="WARNING: Failed to read configuration file from disk: error unmarshaling JSON: while decoding JSON: json: cannot unmarshal string into Go struct field Config.JobCleanupDuration of type time.Duration"

• Try to adjust jobCleanupDuration value in configmap
• Check kuberhealthy logs and you'll see

time="2021-02-25T10:07:35Z" level=debug msg="configReloader: configuration file hash has changed to: 1807f3c6108d085b5b5b6b2a87294b16"
time="2021-02-25T10:07:41Z" level=error msg="configReloader: Error reloading config: error unmarshaling JSON: while decoding JSON: json: cannot unmarshal string into Go struct field Config.JobCleanupDuration of type time.Duration"

Expected behavior
No warnings/errors during starting or configmap changes

Versions

• Cluster OS: Flatcar
• Kubernetes Version: 1.18.14
• Kuberhealthy Release: v2.4.0

[integrii]

This is really strange considering that the helm chart defaults to 15m in the values.yaml... We will need to reproduce this one and figure out what the difference is

[joshulyne]

Fixed in #862! Use image: kuberhealthy/kuberhealthy:v2.4.1!

[integrii]

@joshulyne and I looked into this and it looks like the root cause of the failed time.Duration parsing is due to the package used here. I am going to open a new PR for this one here soon. Reopening.

[integrii]

Looks like I can piggyback this change into #884