By Ryan Blunden on August 17, 2018
You may be thinking “I use Kubernetes, doesn’t this only affect people executing
docker image pull? Well yes, but that is what Kubernetes does when it doesn’t find a requested image locally.
Below is the advice we’ve given our customers to help them work around the outage and we encourage you to share this with anyone who uses Kubernetes to help them have an incident and outage free weekend.
We use the imagePullPolicy default value (IfNotPresent), meaning the image will not be pulled if available locally.
The idea is to not give a Kubernetes worker node a reason to pull an image from Docker Hub.
While this is somewhat out of our control (if a node dies), we recommend to:
This reduces the chances of Kubernetes re-scheduling Pods onto different worker nodes (where the required image may not have been downloaded) and hence, triggering a pull from Docker Hub.
We hope this post has given you some ideas about how you can limit the impact the Docker Hub outage has on your Kubernetes and Docker deployments.
We know this must be a stressful operation for Docker’s Ops folks and we wish them all the best!