Replicaset doesnot update pods in when pod image is modified - kubernetes

Question

I have created a replicaset with wrong container image with below configuration.
apiVersion: extensions/v1beta1
kind: ReplicaSet
metadata:
name: rs-d33393
namespace: default
spec:
replicas: 4
selector:
matchLabels:
name: busybox-pod
template:
metadata:
labels:
name: busybox-pod
spec:
containers:
- command:
- sh
- -c
- echo Hello Kubernetes! && sleep 3600
image: busyboxXXXXXXX
name: busybox-container
Pods Information:
$ kubectl get pods
NAME READY STATUS RESTARTS AGE
rs-d33393-5hnfx 0/1 InvalidImageName 0 11m
rs-d33393-5rt5m 0/1 InvalidImageName 0 11m
rs-d33393-ngw78 0/1 InvalidImageName 0 11m
rs-d33393-vnpdh 0/1 InvalidImageName 0 11m
After this, i try to edit the image inside replicaset using kubectl edit replicasets.extensions rs-d33393 and update image as busybox.
Now, i am expecting pods to be recreated with proper image as part of replicaset.
This has not been the exact result.
Can someone please explain, why it is so?
Thanks :)

Answer 1

With ReplicaSets directly you have to kill the old pod, so the new ones will be created with the right image.
If you would be using a Deployment, and you should, changing the image would force the pod to be re-created.

Answer 2

Replicaset does not support updates. As long as required number of pods exist matching the selector labels, replicaset's jobs is done. You should use Deployment instead.
https://kubernetes.io/docs/concepts/workloads/controllers/replicaset/
From the docs:
To update Pods to a new spec in a controlled way, use a Deployment, as
ReplicaSets do not support a rolling update directly.

Answer 3

Deployment is a higher-level concept that manages ReplicaSets and provides declarative updates to Pods. Therefore, it is recommend to use Deployments instead of directly using ReplicaSets unless you don’t require updates at all. ( i.e. one may never need to manipulate ReplicaSet objects when using a Deployment)
Its easy to perform rolling updates and rollbacks when deployed using deployments.
$ kubectl create deployment busybox --image=busyboxxxxxxx --dry-run -o yaml > busybox.yaml
$ cat busybox.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
creationTimestamp: null
labels:
app: busybox
name: busybox
spec:
replicas: 1
selector:
matchLabels:
app: busybox
strategy: {}
template:
metadata:
creationTimestamp: null
labels:
app: busybox
spec:
containers:
- image: busyboxxxxxxx
name: busyboxxxxxxx
ubuntu#dlv-k8s-cluster-master:~$ kubectl create -f busybox.yaml --record=true
deployment.apps/busybox created
Check rollout history
ubuntu#dlv-k8s-cluster-master:~$ kubectl rollout history deployment busybox
deployment.apps/busybox
REVISION CHANGE-CAUSE
1 kubectl create --filename=busybox.yaml --record=true
Update image on deployment
ubuntu#dlv-k8s-cluster-master:~$ kubectl set image deployment.app/busybox *=busybox --record
deployment.apps/busybox image updated
ubuntu#dlv-k8s-cluster-master:~$ kubectl rollout history deployment busybox
deployment.apps/busybox
REVISION CHANGE-CAUSE
1 kubectl create --filename=busybox.yaml --record=true
2 kubectl set image deployment.app/busybox *=busybox --record=true
Rollback Deployment
ubuntu#dlv-k8s-cluster-master:~$ kubectl rollout undo deployment busybox
deployment.apps/busybox rolled back
ubuntu#dlv-k8s-cluster-master:~$ kubectl rollout history deployment busybox
deployment.apps/busybox
REVISION CHANGE-CAUSE
2 kubectl set image deployment.app/busybox *=busybox --record=true
3 kubectl create --filename=busybox.yaml --record=true

Answer 4

You could use
k scale rs new-replica-set --replicas=0
and then
k scale rs new-replica-set --replicas=<Your number of replicas>

Answer 5

Edit the replicaset(assuming its called replicaset.yaml) file with command:
kubectl edit rs replicaset
edit the image name in the editor
save the file
exit the editor
Now , you will need to either delete the replica sets or delete the existing pods:
kubectl delete rs new-replica-set
kubectl delete pod pod_1 pod_2 pod_2 pod_4
replicaset should spin up new pods with new image.