Docker layer caching for base image

Suppose we have our base container image as cimg/ruby:2.5 and we want to install some additional packages on top of that and then use that built environment for many subsequent parallel testing containers.
Is it possible to utilize docker layer caching for that? The documentation was not super clear on that point. It seemed to indicate that if we’re building a docker image as part of our our ci process, that it would be quite beneficial to use DLC, however we don’t need the image for any external purpose, only for our ci testing. Furthermore since our environment is inside a docker container, I wasn’t sure whether DLC was something we could utilize for a base image.

My understanding is that the setup would actually be

  • job: build ~ using base image cimg/something (doesn’t matter)
    => invoke setup_remote_docker
    => invoke “docker build”
    DockerFile
    → from cimg/ruby:2.5
    → install packages and setup environment
    => save the docker image as “custom”
  • job: test ~ using cimg/something (doesn’t matter)
    => setup_remote_docker
    => run the “custom” image in remote_docker
    => trigger tests within the remote_docker ??? (or something like that)

What I’m envisioning is something like:

  • job: build ~ using base image cimg/ruby:2.5
    => install packages and setup environment
    => Save current container as “custom”
  • job: test ~ using base image “custom” from the previous build
    => run tests

Is the only way to achieve this, by uploading our compiled image to dockerhub (or equivalent)? Because we absolutely do not need the image available externally, and plus it seems that pushing/pulling instances from an external source adds significant time. This seems like what workspaces should be for, right?

Hello

The video at the bottom of our enabling docker layer caching documentation linked below shows an example of what you are looking to do which is linked below.

You would not need to create a custom image as this is defined within your dockerfile which is invoked by the docker build . command

If you have any questions I will be more than happy to help.

Kind Regards
Owen Oliver

Hi,
I’ve seen that video before, but let me just clarify.
So currently we are running using a docker executor. We specify a base image for this executor, and then define several steps to install packages and setup the environment.
We do not currently use “docker build” or a docker file for anything within our circleci jobs.

So my current understanding is that in order to take advantage of DLC, we would need to:

  1. Move many of our steps out of circleci/config.yml and into a dockerfile
  2. Enable docker layer caching
  3. Run step “docker build” within each job in the workflow.

What I don’t currently understand is how we would then utilize whatever is built within that docker image for our rails app testing?
Maybe you could elaborate on how to convert a circleci/config.yml file that does everything as separate inline steps into something using DLC and a docker file?

Thanks
– Matt