Separate agent and analyzer charts

[Sparika]

Main change

The chart is separated into two subcharts. One is the skydive-analyzer chart and the other is the skydive-agent chart. The root chart control the deployment of both which allows to deploy through helm either only agent daemonset, only analyzer deployment, or both.

The parameters to control the deployment are deployment.agent and deployment.analyzer. Both parameters default to true so the default behavior does not change.

Parameters for each charts are located on their respective values.yaml files but can be modified directly from the root values.yaml file. However, parameters must be prefixed by the sub-chart name as documented in the README.

Additional changes

• Fix API to be compatible with Kubernetes >= 1.16 (see issue Incompatible templates for Kubernetes >= 1.16 #2)
• Add deployment of the new Skydive UI from within the analyzer chart. However, the UI expect the analyzer to be available on port 8080. This requires other configurations apart from Helm/K8S.
• Updated Skydive version to 0.26
• Updated parameters documentation in README.md

[hunchback]

Thanks @Sparika for this I generally like your approach (and I can see how this is useful for the multi cluster k8s use case).

Please rebase
Please paste here proof that it works running helm install and helm remove for agent, analyzer and agent+analyzer scenario.

[Sparika]

hi @hunchback, thanks for the review. I have my hands full for the moment but I will rebase and show the result of the two commands once I have time to do it.

[hunchback]

hi @hunchback, thanks for the review. I have my hands full for the moment but I will rebase and show the result of the two commands once I have time to do it.

great!