ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸ ê·¸
2 minute read
Deploy a gardenlet Via gardener-operator
The gardenlet can automatically be deployed by gardener-operator
into existing Kubernetes clusters in order to register them as seeds.
Prerequisites
Using this method only works when gardener-operator
is managing the garden cluster.
If you have used the gardener/controlplane
Helm chart for the deployment of the Gardener control plane, please refer to this document.
💡 Tip
The initial seed cluster can be the garden cluster itself, but for better separation of concerns, it is recommended to only register other clusters as seeds.
Deployment of gardenlets
Using this method, gardener-operator
is only taking care of the very first deployment of gardenlet.
Once running, the gardenlet leverages the self upgrade strategy in order to keep itself up-to-date.
Concretely, gardener-operator
only acts when there is no respective Seed
resource yet.
In order to request a gardenlet deployment, create following resource in the (virtual) garden cluster:
apiVersion: seedmanagement.gardener.cloud/v1alpha1
kind: Gardenlet
metadata:
name: local
namespace: garden
spec:
deployment:
replicaCount: 1
revisionHistoryLimit: 2
helm:
ociRepository:
ref: <url-to-gardenlet-chart-repository>:v1.97.0
config:
apiVersion: gardenlet.config.gardener.cloud/v1alpha1
kind: GardenletConfiguration
controllers:
shoot:
reconcileInMaintenanceOnly: true
respectSyncPeriodOverwrite: true
shootState:
concurrentSyncs: 0
featureGates:
ShootManagedIssuer: true
etcdConfig:
featureGates:
UseEtcdWrapper: true
logging:
enabled: true
vali:
enabled: true
shootNodeLogging:
shootPurposes:
- infrastructure
- production
- development
- evaluation
seedConfig:
apiVersion: core.gardener.cloud/v1beta1
kind: Seed
metadata:
labels:
base: kind
spec:
backup:
provider: local
region: local
secretRef:
name: backup-local
namespace: garden
dns:
provider:
secretRef:
name: internal-domain-internal-local-gardener-cloud
namespace: garden
type: local
ingress:
controller:
kind: nginx
domain: ingress.local.seed.local.gardener.cloud
networks:
nodes: 172.18.0.0/16
pods: 10.1.0.0/16
services: 10.2.0.0/16
shootDefaults:
pods: 10.3.0.0/16
services: 10.4.0.0/16
provider:
region: local
type: local
zones:
- "0"
settings:
excessCapacityReservation:
enabled: false
scheduling:
visible: true
verticalPodAutoscaler:
enabled: true
This causes gardener-operator
to deploy gardenlet to the same cluster where it is running.
Once it comes up, gardenlet will create a Seed
resource with the same name and uses the Gardenlet
resource for self-upgrades (see this document).
Remote Clusters
If you want gardener-operator
to deploy gardenlet into some other cluster, create a kubeconfig Secret
and reference it in the Gardenlet
resource:
apiVersion: v1
kind: Secret
metadata:
name: remote-cluster-kubeconfig
namespace: garden
type: Opaque
data:
kubeconfig: base64(kubeconfig-to-remote-cluster)
---
apiVersion: seedmanagement.gardener.cloud/v1alpha1
kind: Gardenlet
metadata:
name: local
namespace: garden
spec:
kubeconfigSecretRef:
name: remote-cluster-kubeconfig
# ...
After successful deployment of gardenlet, gardener-operator
will delete the remote-cluster-kubeconfig
Secret
and set .spec.kubeconfigSecretRef
to nil
.
This is because the kubeconfig will never ever be needed anymore (gardener-operator
is only responsible for initial deployment, and gardenlet updates itself with an in-cluster kubeconfig).