SAP Converged Charts

This repository contains Helm charts required by SAP Converged Cloud.

Structure

Charts are grouped logically into:

• common: Reusable charts
• global: Singletons that only exist once in a global context
• openstack: Openstack and dependent or related services
• system: Infrastructure required by the control plane

This structure is just a logical grouping, it does not represent deployable units or imply other semantics.

Charts

On the second level we expect a chart. This can be a single chart or a meta-chart that describe a dependent set of compononents. Meta-charts contain sub-charts or reference charts from other repositories using Helm dependencies.

.
└── system
    ├── dns
    │   └── charts
    │       ├── bind
    │       └── unbound
    ├── kube-system
    │   └── charts
    │       ├── ingress
    │       └── dashboard
    └── prometheus
        └── charts
            ├── kube-state-metrics
            ├── prometheus-collector
            └── prometheus-frontend

We imply that the highest chart will be deployed as a Helm release. In this example, releasing dns will install/update bind and unbound.

In order to be able to relate charts to running Kubernetes pods, we also imply that a chart will be deployed in a namespace with the same name.

$ kubectl get pods --all-namespaces                                                                                                                 0 ↵
NAMESPACE         NAME                                               READY     STATUS    RESTARTS   AGE
dns               bind1-2290429089-joidj                             2/2       Running   0          5d
dns               bind2-3590597799-1vcv0                             2/2       Running   0          5d
dns               unbound1-3007389427-shh2y                          1/1       Running   0          9d
dns               unbound1-3577488147-ld1rd                          1/1       Running   0          5d
kube-system       ingress-controller-d3snv                           1/1       Running   4          13d
kube-system       ingress-controller-j9bpf                           1/1       Running   2          18d

This has the benefits that:

• Values required for releasing a chart can be found at the same place in cc/regions
• Cleanup of a failed release, is as easy as deleting the namespace.
• For testing a chart can deployed in a seperate testing namespace.
• Pods and other Kubernetes primitives are reflected at a known place in Kubernetes

Install/Update of a Chart/Release

Per convention we use the name of the meta-chart as namespace and name of the release. Values are pulled in from a secret repository.

helm upgrade dns ./system/dns --namespace dns --values ../secrets/staging/system/dns.yaml --install