constellation/operators/constellation-node-operator
Malte Poll 4283601433
operators: infrastructure autodiscovery (#1958)
* helm: configure GCP cloud controller manager to search in all zones of a region

See also: d716fdd452/providers/gce/gce.go (L376-L380)

* operators: add nodeGroupName to ScalingGroup CRD

NodeGroupName is the human friendly name of the node group that will be exposed to customers via the Constellation config in the future.

* operators: support simple executor / scheduler to reconcile on non-k8s resources

* operators: add new return type for ListScalingGroups to support arbitrary node groups

* operators: ListScalingGroups should return additionally created node groups on AWS

* operators: ListScalingGroups should return additionally created node groups on Azure

* operators: ListScalingGroups should return additionally created node groups on GCP

* operators: ListScalingGroups should return additionally created node groups on unsupported CSPs

* operators: implement external scaling group reconciler

This controller scans the cloud provider infrastructure and changes k8s resources accordingly.
It creates ScaleSet resources when new node groups are created and deletes them if the node groups are removed.

* operators: no longer create scale sets when the operator starts

In the future, scale sets are created dynamically.

* operators: watch for node join/leave events using a controller

* operators: deploy new controllers

* docs: update auto scaling documentation with support for node groups
2023-07-05 07:27:34 +02:00
..
api operators: infrastructure autodiscovery (#1958) 2023-07-05 07:27:34 +02:00
config operators: infrastructure autodiscovery (#1958) 2023-07-05 07:27:34 +02:00
controllers deps: bump ginkgo version (#1885) 2023-06-06 18:15:48 +02:00
external/github.com/medik8s/node-maintenance-operator/config/crd/bases [node operator] Add nodemaintenance CRD 2022-08-09 10:29:04 +02:00
hack [node operator] Initial commit 2022-08-09 10:29:04 +02:00
internal operators: infrastructure autodiscovery (#1958) 2023-07-05 07:27:34 +02:00
sgreconciler operators: infrastructure autodiscovery (#1958) 2023-07-05 07:27:34 +02:00
.dockerignore [node operator] Initial commit 2022-08-09 10:29:04 +02:00
.gitignore join: synchronize control plane joining (#776) 2022-12-09 18:30:20 +01:00
BUILD.bazel operators: infrastructure autodiscovery (#1958) 2023-07-05 07:27:34 +02:00
bundle.Dockerfile join: synchronize control plane joining (#776) 2022-12-09 18:30:20 +01:00
go.mod operators: infrastructure autodiscovery (#1958) 2023-07-05 07:27:34 +02:00
go.sum operators: infrastructure autodiscovery (#1958) 2023-07-05 07:27:34 +02:00
main.go operators: infrastructure autodiscovery (#1958) 2023-07-05 07:27:34 +02:00
Makefile deps: add Kubernetes v1.27, remove Kubernetes v1.24 (#1669) 2023-05-05 13:22:53 +02:00
PROJECT upgrade: support Kubernetes components (#839) 2023-01-03 12:09:53 +01:00
README.md upgrade: fix broken reference from constellation-os to constellation-version (#939) 2023-01-11 16:07:07 +01:00

constellation-node-operator

The constellation node operator manages the lifecycle of constellation nodes after cluster initialization. In particular, it is responsible for updating the OS images of nodes by replacing nodes running old images with new nodes.

High level goals

  • Admin or constellation init can create custom resources for node related components
  • The operator will manage nodes in the cluster by trying to ensure every node has the specified image
  • If a node uses an outdated image, it will be replaced by a new node
  • Admin can update the specified image at any point in time which will trigger a rolling upgrade through the cluster
  • Nodes are replaced safely (cordon, drain, preservation of node labels)

Description

The operator has multiple controllers with corresponding custom resource definitions (CRDs) that are responsible for the following high level tasks:

NodeVersion

NodeVersion is the only user controlled CRD. The spec allows an administrator to update the desired image and trigger a rolling update.

Example for GCP:

apiVersion: update.edgeless.systems/v1alpha1
kind: NodeVersion
metadata:
  name: constellation-version
spec:
  image: "projects/constellation-images/global/images/<image-name>"

Example for Azure:

apiVersion: update.edgeless.systems/v1alpha1
kind: NodeVersion
metadata:
  name: constellation-version
spec:
  image: "/subscriptions/<subscription-id>/resourceGroups/CONSTELLATION-IMAGES/providers/Microsoft.Compute/galleries/Constellation/images/<image-definition-name>/versions/<image-version>"

AutoscalingStrategy

AutoscalingStrategy is used and modified by the NodeVersion controller to pause the cluster-autoscaler while an image update is in progress.

Example:

apiVersion: update.edgeless.systems/v1alpha1
kind: AutoscalingStrategy
metadata:
  name: autoscalingstrategy
spec:
  enabled: true
  deploymentName: "cluster-autoscaler"
  deploymentNamespace: "kube-system"

ScalingGroup

ScalingGroup represents one scaling group at the CSP. Constellation uses one scaling group for worker nodes and one for control-plane nodes. The scaling group controller will automatically set the image used for newly created nodes to be the image set in the NodeVersion Spec. On cluster creation, one instance of the ScalingGroup resource per scaling group at the CSP is created. It does not need to be updated manually.

Example for GCP:

apiVersion: update.edgeless.systems/v1alpha1
kind: ScalingGroup
metadata:
  name: scalinggroup-worker
spec:
  nodeImage: "constellation-version"
  groupId: "projects/<project-id>/zones/<zone>/instanceGroupManagers/<instance-group-name>"
  autoscaling: true

Example for Azure:

apiVersion: update.edgeless.systems/v1alpha1
kind: ScalingGroup
metadata:
  name: scalinggroup-worker
spec:
  nodeImage: "constellation-version"
  groupId: "/subscriptions/<subscription-id>/resourceGroups/<resource-group>/providers/Microsoft.Compute/virtualMachineScaleSets/<scale-set-name>"
  autoscaling: true

PendingNode

PendingNode represents a node that is either joining or leaving the cluster. These are nodes that are not part of the cluster (they do not have a corresponding node object). Instead, they are used to track the creation and deletion of nodes. This resource is automatically managed by the operator. For joining nodes, the deadline is used to delete the pending node if it fails to join before the deadline ends.

Example for GCP:

apiVersion: update.edgeless.systems/v1alpha1
kind: PendingNode
metadata:
  name: pendingnode-sample
spec:
  providerID: "gce://<project-id>/<zone>/<instance-name>"
  groupID: "projects/<project-id>/zones/<zone>/instanceGroupManagers/<instance-group-name>"
  nodeName: "<kubernetes-node-name>"
  goal: Join
  deadline: "2022-07-04T08:33:18+00:00"

Example for Azure:

apiVersion: update.edgeless.systems/v1alpha1
kind: PendingNode
metadata:
  name: pendingnode-sample
spec:
  providerID: "azure:///subscriptions/<subscription-id>/resourceGroups/<resource-group>/providers/Microsoft.Compute/virtualMachineScaleSets/<scale-set-name>/virtualMachines/<instance-id>"
  groupID: "/subscriptions/<subscription-id>/resourceGroups/<resource-group>/providers/Microsoft.Compute/virtualMachineScaleSets/<scale-set-name>"
  nodeName: "<kubernetes-node-name>"
  goal: Join
  deadline: "2022-07-04T08:33:18+00:00"

Getting Started

Youll need a Kubernetes cluster to run against. You can use KIND to get a local cluster for testing, or run against a remote cluster. Note: Your controller will automatically use the current context in your kubeconfig file (i.e. whatever cluster kubectl cluster-info shows).

Running on the cluster

  1. Install Instances of Custom Resources:
kubectl apply -f config/samples/
  1. Build and push your image to the location specified by IMG:
make docker-build docker-push IMG=<some-registry>/constellation/node-operator:tag
  1. Deploy the controller to the cluster with the image specified by IMG:
make deploy IMG=<some-registry>/constellation/node-operator:tag

Uninstall CRDs

To delete the CRDs from the cluster:

make uninstall

Undeploy controller

UnDeploy the controller to the cluster:

make undeploy

How it works

This project aims to follow the Kubernetes Operator pattern

It uses Controllers which provides a reconcile function responsible for synchronizing resources until the desired state is reached on the cluster

Test It Out

  1. Install the CRDs into the cluster:
make install
  1. Run your controller (this will run in the foreground, so switch to a new terminal if you want to leave it running):
make run

NOTE: You can also run this in one step by running: make install run

Modifying the API definitions

If you are editing the API definitions, generate the manifests such as CRs or CRDs using:

make manifests

NOTE: Run make --help for more information on all potential make targets

More information can be found via the Kubebuilder Documentation

Production deployment

The operator is deployed automatically during constellation-init. Prerequisite for this is that cert-manager is installed. cert-manager is also installed during constellation-init. To deploy you can use the Helm chart at /cli/internal/helm/charts/edgeless/operators/constellation-operator.