QDR Operator
A Kubernetes operator for managing Layer 7 (e.g. Application Layer) addressing and routing within and across clusters. The operator manages interior and edge QDR deployments automating resource creation and administration.
Introduction
This operator provides an Interconnect
Custom Resource Definition
(CRD) that describes a deployment of Apache Qpid Dispatch Router allowing developers to expertly
deploy routers for their infrastructure and middle-ware oriented messaging requirements. The number of messaging routers, the deployment topology, address
semantics and other desired options can be specified through the CRD.
Usage
Deploy the QDR Operator into the Kubernetes cluster where it will manage requests for the Interconnect
resource. The QDR Operator will watch for create, update and delete resource requests and perform the necessary steps to ensure the present cluster state matches the desired state.
Deploy QDR Operator
The deploy
directory contains the manifests needed to properly install the
Operator.
Create the service account for the operator.
$ kubectl create -f deploy/service_account.yaml
Create the RBAC role and role-binding that grants the permissions
necessary for the operator to function.
$ kubectl create -f deploy/role.yaml
$ kubectl create -f deploy/role_binding.yaml
Deploy the CRD to the cluster that defines the Interconnect resource.
$ kubectl create -f deploy/crds/interconnectedcloud_v1alpha1_interconnect_crd.yaml
Next, deploy the operator into the cluster.
$ kubectl create -f deploy/operator.yaml
This step will create a pod on the Kubernetes cluster for the QDR Operator.
Observe the qdr-operator
pod and verify it is in the running state.
$ kubectl get pods -l name=qdr-operator
If for some reason, the pod does not get to the running state, look at the
pod details to review any event that prohibited the pod from starting.
$ kubectl describe pod -l name=qdr-operator
You will be able to confirm that the new CRD has been registered in the cluster and you can review its details.
$ kubectl get crd
$ kubectl describe crd interconnects.interconnectedcloud.github.io
To create a router deployment, you must create a Interconnect
resource representing the desired specification of the deployment. For example, to create a 3-node router mesh deployment you may run:
$ cat <<EOF | kubectl create -f -
apiVersion: interconnectedcloud.github.io/v1alpha1
kind: Interconnect
metadata:
name: example-interconnect
spec:
# Add fields here
deploymentPlan:
image: quay.io/interconnectedcloud/qdrouterd:1.9.0
role: interior
size: 3
placement: Any
EOF
The operator will create a deployment of three router instances, all connected together with default address semantics. It will also create a service through which the interior router mesh can be accessed. It will configure a default set of listeners and connectors as described below. You will be able to confirm that the instance has been created in the cluster and you can review its details. To view the Interconnect instance, the deployment it manages and the associated pods that are deployed:
$ kubectl describe interconnect example-interconnect
$ kubectl describe deploy example-interconnect
$ kubectl describe svc example-interconnect
$ kubectl get pods -o yaml
Deployment Plan
The CRD Deployment Plan defines the attributes for an Interconnect instance.
Role and Placement
The Deployment Plan Role defines the mode of operation for the routers in a topology.
-
interior - This role creates an interconnect of auto-meshed routers for concurrent connection capacity and resiliency.
Connectivity between the routers will be defined by InterRouterListeners and InterRouterConnectors. Downlink
connectivity with edge routers will be via EdgeListeners.
-
edge - This role creates a set of stand-alone routers. The connectivity from the edge to interior routers
will be via EdgeConnectors.
The Deployment Plan Placement defines the deployment resource and the associated scheduling of the pods in the cluster.
-
Any - There is no constraint on pod placement. The operator will manage a Deployment resource where the number of
pods scheduled will be up to Deployment Plan Size.
-
Every - Router pods will be placed on each node in the cluster. The operator will manage a DaemonSet resource where the
number of pods scheduled will correspond to the number of nodes in the cluster. Note the Deployment Plan Size is
not used.
-
Anti-Affinity - This constrains scheduling and prevents multiple router pods from running on the same node in the
cluster. The operator will manage a Deployment resource with number of pods up to Deployment Plan Size.
Note if Deployment Plan Size is greater than the number of nodes in the cluster, the excess pods that cannot be
scheduled will remain in the pending state.
Connectivity
The connectivity between routers in a deployment is via the declared listeners and connectors. There are three types of listeners supported by the operator.
-
Listeners - A listener for accepting normal messaging client connections. The operator supports this listener for
both interior and edge routers.
-
InterRouterListeners - A listener for accepting connections from peer interior routers. The operator
support this listener for interior routers only.
-
EdgeListeners - A listener for accepting connections from downlink edge routers. The operator supports this
listener for interior routers only.
There are three types of connectors supported by the operator.
-
Connectors - A connector for connecting to an external messaging intermediary. The operator supports this connector
for both interior and edge routers.
-
InterRouterConnectors - A connector for establishing connectivity to peer interior routers. The operator
supports this connector for interior routers only.
-
EdgeConnector - A connector for establishing up-link connectivity from edge to interior routers. The operator
supports this connector for edge routers only.
Development
This Operator is built using the Operator SDK. Follow the Quick Start instructions to checkout and install the operator-sdk CLI.
Local development may be done with minikube or minishift.
Source Code
Clone this repository to a location on your workstation such as $GOPATH/src/github.com/ORG/REPO
. Navigate to the repository and install the dependencies.
$ cd $GOPATH/src/github.com/ORG/REPO/qdr-operator
$ dep ensure && dep status
Run Operator Locally
Ensure the service account, role, role bindings and CRD are added to the local cluster.
$ kubectl create -f deploy/service_account.yaml
$ kubectl create -f deploy/role.yaml
$ kubectl create -f deploy/role_binding.yaml
$ kubectl create -f deploy/crds/interconnectedcloud_v1alpha1_interconnect_crd.yaml
Start the operator locally for development.
$ operator-sdk up local
Create a minimal Interconnect resource to observe and test your changes.
$ cat <<EOF | kubectl create -f -
apiVersion: interconnectedcloud.github.io/v1alpha1
kind: Interconnect
metadata:
name: example-interconnect
spec:
deploymentPlan:
image: quay.io/interconnectedcloud/qdrouterd:1.9.0
role: interior
size: 3
placement: Any
EOF
As you make local changes to the code, restart the operator to enact the changes.
Build
The Makefile will do the dependency check, operator-sdk generate k8s, run local test, and finally the operator-sdk build. Please ensure any local docker server is running.
make
Test
Before submitting PR, please test your code.
File or local validation.
$ make test
Cluster-based test.
Ensure there is a cluster running before running the test.
$ make cluster-test
Manage the operator using the Operator Lifecycle Manager
Ensure the Operator Lifecycle Manager is installed in the local cluster. By default, the catalog-source.sh
will install the operator catalog resources in operator-lifecycle-manager
namespace. You may also specify different namespace where you have the Operator Lifecycle Manager installed.
$ ./hack/catalog-source.sh [namespace]
$ oc apply -f deploy/olm-catalog/qdr-operator/0.1.0/catalog-source.yaml