We have a lot docker images that we need to upload to k8s, it works however when we start a new node this process can take a lot of time.

We try to change the images to use "multi-stage" builds however still this process take a lot of time and we didn't saw significant change on the start time on new node.

Now we try the following: (however we not sure how much impact we will see ...)

1. Try (as much as we can) to use the same docker images from , example all the images which use alpine to use the exact same version (3.11) and not some of using version 3.10 and some 3.10.1 and some 3.11 etc , this will prevent the reuse mechanism of docker to layers that already cashed ...

2. Reduce the layers numbers , with combining the RUN commands to one instead of many.(we need to do it for many docker files...) , not sure how docker mechanism will help here to achieve our goals

All this process to change both of using the same version and reduce layers can take a lot of time to do, are we are wasting our time or there or this is something that can reduce the load time ? is there any other idea which may help us ?

We have many docker images based on Golang, NodeJS, java, etc.