What is the caching mechanism of RUNPOD docker image?
our Docker image is stored in AWS ECR. We've noticed that every time we update the Docker template on the runpod, our ECR costs increase rapidly. We've identified that this is due to using 80 runpod instances, and these runpod instances concurrently pull the image. We would like to inquire about the image caching mechanism of runpod. If we let one runpod instance pull the Docker image completely first before starting other instances, will the other instances pull the image from your cache instead of AWS ECR?
6 Replies
Images aren't pulled from the cache unless its already cached on the same machine. By "80 runpod instances", do you mean serverless workers or Pods?
I don't recommend using AWS for images that need to be used on RunPod. I use Dockerhub and its completely free.
80 runpod instances means serverless workers
Incorrect, some people refer to pods as "instances". Workers are workers, pods are pods, there is no such concept as "instances".
I use serverless. I have one task (use one endpoint, from one template), and this endpoint have 80 workers. when I update my docker image version in template, each of these 80 workers will pull the new docker image.
Yes thats normal, there is no cache on RunPod side unless the workers are on the same machine.
I use AWS but I don't use ECR for my RunPod serverless workers because its too expensive.
OK，I get it, thx