Migrate to CRDS and Kube builder (#33)

* initial machine actuator update implementation

* update kubelet and control plane version

* init kubebuilder project

* add API types for machine and cluster provider config

* prepare for cluster api vendor upgrade

* vendor: update cluster-api and vendor controller-runtime dependencies

* move cloud directory under pkg

* update cluster actuator to use controller manager

* update machine actuator and machine status

Comment out test cases that depend on the cluster v1 alpha1 client

* remove unused providerconfig

* update manager to reigster acutators properly

* fix package import for azureconfigv1

* re-run codegen

* remove unused controller

* make it compile

* move arm templates under resourcemanagement package and update Dockerfile

* update kustomize config files

* use generated cluster name in the template file

* fix one machine update test case that doesnt utilize cluster api client

* update e2e test client

* vendor: update vendored dependencies

* manually add upstream crd configs

dep prune does not keep this folder for some reason even with non-go set to true

* set logtostderr flag for manager

* remove exampleconfigs and duplicate boilerplate.txt

* move configtemplate to clusterctl/examples

* add vendors target, dev image push and build to makefile

* add targets for machine and cluster actuator unit tests

* update build and test pipeline

* dont prune cluster-api package

To preserve config directory

* set NODE_NAME env var to avoid self deletion

* fix machine status update

* disable e2e pipeline

* update .gitignore

* cleanup generate-yaml script and remove unused makefiles

* fix package import and remove unused test files

* remove unused provider components template

* update kustomize secret generator to read from env variables

* replace minikube with kubeadm-dind

CRD subresources requires Kubernetes >= 1.11 (or else we'd have to use feature
flag to enable). The none vm-driver for minikube does not support versions >
1.10. A fork is used for the shell script which fixes an issue with teeing to
stderr on the Azure pipelines Ubuntu agent.

* setup local docker registry and configure the nodes to pull the images

* fix the rest of the E2E steps and re-enable the pipeline

* change docker image placeholder

* export service principal values to a file

* update README with the new setup instructions

* vendor: update packages

* add skipping creating service principal if flag is set

* update e2e tests

* fix machine annotation update for first creation

* only update the status once for node recreation

* install specific control plane version through kubeadm

* upgrade kubectl client version with the control plane

* switch to alpine 3.8

* use explicit argument indices for formatting strings

.gitignore

@@ -1,8 +1,10 @@
 *.test
 .idea/
+bin/
 .env
 exampleconfigs/provider-components.yaml
 minikube.kubeconfig
 generatedconfigs/
 sshkey*
 kubeconfig
+*.out

Dockerfile

@@ -0,0 +1,18 @@
+# Build the manager binary
+FROM golang:1.10.3 as builder
+
+# Copy in the go src
+WORKDIR /go/src/github.com/platform9/azure-provider
+COPY pkg/    pkg/
+COPY cmd/    cmd/
+COPY vendor/ vendor/
+
+# Build
+RUN CGO_ENABLED=0 GOOS=linux GOARCH=amd64 go build -a -o manager github.com/platform9/azure-provider/cmd/manager
+
+# Copy the controller-manager into a thin image
+FROM alpine:3.8
+WORKDIR /root/
+COPY --from=builder /go/src/github.com/platform9/azure-provider/manager .
+COPY --from=builder /go/src/github.com/platform9/azure-provider/pkg/cloud/azure/services/resourcemanagement/template/deployment-template.json .
+ENTRYPOINT ["./manager"]