Resource Migration
Objectives
Assuming you have a single kubernetes cluster which already has many native resource installed, furthermore, you want to migrate the existing resource to Karmada and then achieve multi-cluster management.
So, this section will guide you to cover:
- Migrate all the existing resource from original cluster to Karmada based on resource granularity.
- Apply higher priority
PropagationPolicy
to meet more propagate demands based on application granularity.
Prerequisites
Karmada with multi cluster has been installed
Step 1: Run the command
$ git clone https://github.com/karmada-io/karmada
$ cd karmada
$ hack/local-up-karmada.sh
$ export KUBECONFIG=~/.kube/karmada.config:~/.kube/members.config
Note:
Before guide started, we should install at least three kubernetes clusters, one is for Karmada control plane, the other two for member clusters. For convenience, we use hack/local-up-karmada.sh script to quickly prepare the above clusters.
After the above command executed, you will see Karmada control plane installed with multi member clusters.
Enable PropagationPolicyPreemption in karmada-controller-manager
Step 2: Run the command
$ kubectl --context karmada-host get deploy karmada-controller-manager -n karmada-system -o yaml | sed '/- --failover-eviction-timeout=30s/{n;s/- --v=4/- --feature-gates=PropagationPolicyPreemption=true\n &/g}' | kubectl --context karmada-host replace -f -
Note:
The feature
PropagationPolicy Priority and Preemption
was introduced in v1.7, and it is controlled by the feature gatePropagationPolicyPreemption
which is disabled by default.You can just execute the above one command to enable this feature gate. Or, if you want to use a more cautious approach, you can do like this:
- execute
kubectl --context karmada-host edit deploy karmada-controller-manager -n karmada-system
- check if feature gate
--feature-gates=PropagationPolicyPreemption=true
is existed inspec.template.spec.containers[0].command
field.- If not, you shall add
--feature-gates=PropagationPolicyPreemption=true
into that field.
Preset resource in a member cluster
To simulate resources already exist in member cluster, we deploy some simple Deployments and Services to member1
cluster.
Step 3: Write the code
Create new file /tmp/deployments-and-services.yaml
and copy text below to it:
apiVersion: apps/v1
kind: Deployment
metadata:
name: nginx-deploy
spec:
selector:
matchLabels:
app: nginx
replicas: 2
template:
metadata:
labels:
app: nginx
spec:
containers:
- name: nginx
image: nginx:latest
ports:
- containerPort: 80
---
apiVersion: v1
kind: Service
metadata:
name: nginx-svc
spec:
selector:
app: nginx
type: NodePort
ports:
- port: 80
nodePort: 30000
targetPort: 80
Step 4: Run the command
$ kubectl --context member1 apply -f /tmp/deployments-and-services.yaml
deployment.apps/nginx-deploy created
service/nginx-svc created
deployment.apps/hello-deploy created
service/hello-svc created
Thus, we can use member1
as the cluster with existing resources, while member2
as a bare cluster.
Tutorials
Migrate all the resources to Karmada
Step 1: Run the command
$ kubectl --context karmada-apiserver apply -f /tmp/deployments-and-services.yaml
deployment.apps/nginx-deploy created
service/nginx-svc created
deployment.apps/hello-deploy created
service/hello-svc created
Note:
Same Deployments and Services should be deployed to Karmada control plane as ResourceTemplate.
Step 2: Write the code
Create new file /tmp/pp-for-migrating-deployments-and-services.yaml
and copy text below to it:
apiVersion: policy.karmada.io/v1alpha1
kind: PropagationPolicy
metadata:
name: migrate-pp
spec:
conflictResolution: Overwrite
placement:
clusterAffinity:
clusterNames:
- member1
priority: 0
resourceSelectors:
- apiVersion: apps/v1
kind: Deployment
- apiVersion: v1
kind: Service
schedulerName: default-scheduler
Note:
You should pay attention to two fields:
spec.conflictResolution: Overwrite
:the value must be Overwrite.spec.resourceSelectors
:selecting resources to migrate, you can define your custom ResourceSelector.
Step 3: Run the command
Apply this PropagationPolicy to Karmada control plane.
$ kubectl --context karmada-apiserver apply -f /tmp/pp-for-migrating-deployments-and-services.yaml
propagationpolicy.policy.karmada.io/migrate-pp created
Step 4: Verification
$ kubectl --context karmada-apiserver get deploy
NAME READY UP-TO-DATE AVAILABLE AGE
nginx-deploy 2/2 2 2 38s
$ kubectl --context karmada-apiserver get rb
NAME SCHEDULED FULLYAPPLIED AGE
nginx-deploy-deployment True True 13s
nginx-svc-service True True 13s
You shall see the Deployments in Karmada are all ready and the aggregatedStatus
of ResourceBinding
is applied, which means the existing resources in member1
cluster has been taken over by Karmada.
So far, you have finished the migration, isn’t it so easy?
Apply higher priority PropagationPolicy
Step 5: Write the code
Create new file /tmp/pp-for-nginx-app.yaml
and copy text below to it:
apiVersion: policy.karmada.io/v1alpha1
kind: PropagationPolicy
metadata:
name: nginx-pp
spec:
conflictResolution: Overwrite
placement:
clusterAffinity:
clusterNames:
- member1
- member2 ## propagate to more clusters other than member1
priority: 10 ## priority greater than above PropagationPolicy (10 > 0)
preemption: Always ## preemption should equal to Always
resourceSelectors:
- apiVersion: apps/v1
kind: Deployment
name: nginx-deploy
- apiVersion: v1
kind: Service
name: nginx-svc
schedulerName: default-scheduler
Step 6: Run the command
Apply this higher priority PropagationPolicy to Karmada control plane.
$ kubectl --context karmada-apiserver apply -f /tmp/pp-for-nginx-app.yaml
propagationpolicy.policy.karmada.io/nginx-pp created
Step 7: Verification
$ kubectl --context member2 get deploy -o wide
NAME READY UP-TO-DATE AVAILABLE AGE CONTAINERS IMAGES SELECTOR
nginx-deploy 2/2 2 2 5m24s nginx nginx:latest app=nginx
$ kubectl --context member2 get svc -o wide
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE SELECTOR
nginx-svc NodePort 10.13.161.255 <none> 80:30000/TCP 54s app=nginx
...
As you see, you shall find nginx
application related resource are all propagated to member2
cluster, which means the higher priority PropagationPolicy
does work.