ThirdPartyResource API is removed in k8s version 1.8

[rince86]

As of kubernetes 1.8 installation method is not valid as:

"ThirdPartyResource is deprecated as of Kubernetes 1.7 and has been removed in version 1.8 in accordance with the deprecation policy for beta features."

Monasca deployment will fail with:

helm install monasca/monasca --name monasca --namespace monitoring
Error: error validating "": error validating data: couldn't find type: v1beta1.ThirdPartyResource

due to API does not longer exists.

Ref:
https://kubernetes.io/docs/tasks/access-kubernetes-api/extend-api-third-party-resource/

[mhoppal]

Hi @rince86 yes that is a known bug we are working on fixing here in the short term.

Could you get around it by disabling the thirdpartyresource for now? -> https://github.com/monasca/monasca-helm/blob/master/monasca/values.yaml#L1596 https://github.com/monasca/monasca-helm/blob/master/monasca/values.yaml#L1597

FYI @timothyb89

[timothyb89]

Yep, this should be enough to install on k8s 1.8:

helm install monasca/monasca \
    --name monasca \
    --namespace monitoring \
    --set alarm_definition_controller.resource_enabled=false \
    --set alarm_definition_controller.controller_enabled=false

We're hoping to switch over to CustomResourceDefinitions "soon" but don't currently have an ETA.

@mhoppal maybe it would be best to turn off the ADC by default for now?

[rince86]

Thanks @mhoppal , @timothyb89. Deployment with alarms set worked, however comparing to deployment in 1.7 several pods are absent:

influx-init-job-n6rcg
monasca-adc-370665528-qchdj
monasca-alarms-init-job-8tz3s
monasca-grafana-init-job-jljjf
monasca-kafka-init-job-vjpr8
mysql-init-job-hdp2f

I assume that is expected and should not break anything?

[timothyb89]

monasca-adc should disappear, yes. The various init-job pods are supposed to run to completion and exit. By default kubectl hides completed pods, so you may need to use kubectl get pods --show-all to see them.

If the init jobs never run at all, that would be a bigger issue. Assuming other pods aren't in a crash loop (particularly monasca-persister and anything else that uses kafka) I'd guess that they just completed normally.

Note that I am seeing some other issues in minikube 1.8.0 that I've filed issues for: #348 and #349