A next-generation deployment mechanism founded on the operator pattern allows Kong Gateways to be provisioned in a dynamic and Kubernetes-native way, as well as enabling automation of Kong cluster operations and management of the Gateway lifecycle.
Before deployment, Kong and Gateway API CRDs need to be deployed:
kubectl kustomize https://github.com/Kong/kubernetes-ingress-controller/config/crd | kubectl apply -f -
kubectl kustomize "https://github.com/kubernetes-sigs/gateway-api/config/crd?ref=v0.5.1" | kubectl apply -f -
If using OpenShift, install the operator from the OperatorHub UI. Otherwise, deploy the operator with the following one-liner:
kubectl kustomize "https://github.com/kong/gateway-operator-docs/config/default?submodules=false" | kubectl apply -f -
Optionally, you can wait for the operator with:
kubectl -n kong-system wait --for=condition=Available=true --timeout=120s deployment/gateway-operator-controller-manager
After deployment, usage is driven primarily via the Gateway resource.
You can deploy a Gateway
resource to the cluster, which will result in the underlying control plane (the Kong Kubernetes Ingress Controller) and the data plane (the Kong Gateway).
For example, deploy the following GatewayClass
:
echo '
kind: GatewayClass
apiVersion: gateway.networking.k8s.io/v1beta1
metadata:
name: kong
spec:
controllerName: konghq.com/gateway-operator
' | kubectl apply -f -
and Gateway
:
echo '
kind: Gateway
apiVersion: gateway.networking.k8s.io/v1beta1
metadata:
name: kong
spec:
gatewayClassName: kong
listeners:
- name: http
protocol: HTTP
port: 80
' | kubectl apply -f -
Wait for the Gateway
to be Ready
:
kubectl wait --for=condition=Ready=true gateways.gateway.networking.k8s.io/kong
Once Ready
, you'll be able to receive the default IP address of the Gateway
:
$ kubectl get gateway kong
NAME CLASS ADDRESS READY AGE
kong kong 172.18.0.240 True 97s
The Gateway
is now accessible via that IP:
$ curl -s -w '\n' http://172.18.0.240
{"message":"no Route matched with those values"}
Note: if your cluster can not provision
LoadBalancer
typeServices
then the IP you receive may only be routable from within the cluster.
Note: the
no Route matched
result is normal for aGateway
with no configuration. CreateIngress
,HTTPRoute
, and other resources to start routing traffic to your applications. See the Ingress Controller Guides for more information).
A Gateway
resource has subcomponents such as a ControlPlane
and a DataPlane,
which are created and managed on its behalf.
At a deeper technical level, ControlPlane
corresponds with the Kong Kubernetes Ingress Controller (KIC) and DataPlane
corresponds with the Kong Gateway.
While not required for primary usage, it is possible to provide configuration for these subcomponents using the GatewayConfiguration
API.
That configuration can include the container image and image version to use for the subcomponents, as well as environment and volume mount overrides will be passed down toPods
created for that component.
For example:
kind: GatewayConfiguration
apiVersion: gateway-operator.konghq.com/v1alpha1
metadata:
name: kong
namespace: default
spec:
controlPlaneDeploymentOptions:
containerImage: kong/kubernetes-ingress-controller
version: 2.8.1
env:
- name: TEST_VAR
value: TEST_VAL
dataPlaneDeploymentOptions:
containerImage: kong/kong
version: 3.1.1
env:
- name: TEST_VAR
value: TEST_VAL
Configurations like the above can be created on the API, but won't be active until referenced by a GatewayClass
:
kind: GatewayClass
apiVersion: gateway.networking.k8s.io/v1beta1
metadata:
name: kong
spec:
controllerName: konghq.com/gateway-operator
parametersRef:
group: gateway-operator.konghq.com
kind: GatewayConfiguration
name: kong
namespace: default
With the parametersRef
in the above GatewayClass
being used to attach the GatewayConfiguration
, that configuration will start applying to all Gateway
resources created for that class, and will retroactively apply to any Gateway
resources previously created.
The GatewayConfiguration
API can be used to provide the image and the image version desired for either the ControlPlane
or DataPlane
component of the Gateway
e.g.:
kind: GatewayConfiguration
apiVersion: gateway-operator.konghq.com/v1alpha1
metadata:
name: kong
namespace: default
spec:
dataPlaneDeploymentOptions:
containerImage: kong/kong
version: 2.7.0
controlPlaneDeploymentOptions:
containerImage: kong/kubernetes-ingress-controller
version: 2.4.2
The above configuration will deploy all DataPlane
resources connected to the
GatewayConfiguration
(by way of GatewayClass
) using kong/kong:2.7.0
and any ControlPlane
will be deployed with kong/kubernetes-ingress-controller:2.4.2
.
Given the above, a manual upgrade or downgrade can be performed simply by changing the version.
For example: assuming that at least one Gateway
is currently deployed and running using the above GatewayConfiguration
, an upgrade could be performed by running the following:
kubectl edit gatewayconfiguration kong
And updating the dataPlaneDeploymentOptions.version
to 3.1.1
.
The result will be a replacement Pod
will roll out with the new version and once healthy the old Pod
will be terminated.
You can use Kong Enterprise as the data plane by doing as follows:
Create a secret with the Kong license in the namespace you intend to use for deploying the gateway.
kubectl create secret generic kong-enterprise-license --from-file=license=<license-file> -n <your-namespace>
Create a GatewayConfiguration
specifying the enterprise container image and the environment variable referencing the license secret.
The operator will use the image and the environment variable specified in the GatewayConfiguration
to customize the dataplane.
As the result, the dataplane will use kong/kong-gateway:3.1.1
as the image and mount the license secret.
kind: GatewayConfiguration
apiVersion: gateway-operator.konghq.com/v1alpha1
metadata:
name: kong
namespace: <your-namespace>
spec:
dataPlaneDeploymentOptions:
containerImage: kong/kong-gateway:3.1.1
env:
- name: KONG_LICENSE_DATA
valueFrom:
secretKeyRef:
key: license
name: kong-enterprise-license
Create a GatewayClass
that references the GatewayConfiguration
above.
kind: GatewayClass
apiVersion: gateway.networking.k8s.io/v1beta1
metadata:
name: kong
spec:
controllerName: konghq.com/gateway-operator
parametersRef:
group: gateway-operator.konghq.com
kind: GatewayConfiguration
name: kong
namespace: <your-namespace>
And finally create a Gateway that uses the GatewayClass
above:
kind: Gateway
apiVersion: gateway.networking.k8s.io/v1beta1
metadata:
name: kong
namespace: <your-namespace>
spec:
gatewayClassName: kong
listeners:
- name: http
protocol: HTTP
port: 80
Wait for the Gateway
to be Ready
:
kubectl wait --for=condition=Ready=true gateways.gateway.networking.k8s.io/kong
Check that the data plane is using the enterprise image:
$ kubectl get deployment -l konghq.com/gateway-operator=dataplane -o jsonpath='{.items[0].spec.template.spec.containers[0].image}'
kong/kong-gateway:3.1.1
A log message should describe the status of the provided license.
$ kubectl logs $(kubectl get po -l app=$(kubectl get dataplane -o=jsonpath='{.items[0].metadata.name}') -o=jsonpath="{.items[0].metadata.name}") | grep license_helpers.lua
2022/08/29 10:50:55 [error] 2111#0: *8 [lua] license_helpers.lua:194: log_license_state(): The Kong Enterprise license will expire on 2022-09-20. Please contact <support@konghq.com> to renew your license., context: ngx.timer
Note: the license secret, the
GatewayConfiguration
, and theGateway
MUST be created in the same namespace.
The latest supported version of Gateway API is v1beta1
released in v0.5.1.
The following matrix summarizes support for the Gateway API resources:
CRD Name | Supported | Notes |
---|---|---|
v1beta1.GatewayClass | yes | |
v1beta1.Gateway | yes | |
v1beta1.HTTPRoute | yes | |
v1alpha2.TLSRoute | no | |
v1alpha2.TCPRoute | no | |
v1alpha2.UDPRoute | no |