为什么 EKS pod 引用旧的人工图像地址？

Question

We migrated our docker image artifactory from the self-hosted Nexus on EC2 to the ECR repository and updated the new image reference in all the deployments. But many times we observed that when the pod is recreated or deployment is restarted the pod fails with the imagepullbackoff error because it is referring to the old image reference which does not exist whereas its deployment contains the new image reference. Then we have to manually update the new image reference in the pod which is not the correct way. Does anybody know why this behavior is happening?

Answer 1

when changing the container image name or tag you would just scale your application to 0 and back to the original size means all pods will be deleted and recreated so image tags will be updated in every pod you can do it by:

kubectl scale --replicas=0 deployment my-deployment 

Ensure you set ImagePullPolicy: Always and kill each pod and have the deployment recreate it. this step is not required for local images.

Now update the image simple way is by using following command:

kubectl set image deployment/my-deployment mycontainer=myimage

Now scale up the deployment using the below command ensure you have matched the replica count before and after updating image

kubectl scale --replicas=1 deployment application

Now check if imagepullbackoff error repeats now.