This message was deleted.

This may not answer your question directly but here goes I use a release branch strategy and do a git clone (dev/test/prod)-branch within my docker build In Pulumi, I get the hash of the branch and pass it as a build arg to docker build I then do

RUN echo ${BUILD_GIT_HASH} > /tmp/git_cache_buster.txt && \
    git clone ${BUILD_GIT_URL} --branch ${BUILD_GIT_BRANCH} ./ ...<snip>...

this process means that if the branch hash is the same, the image will be the same and therefore use all cached layers Git hash is generated by

git rev-parse --short=8 [origin]/component-name-dev

This is managed in a dynamic provider HTH

thank you for your receipt your idea is to build each time but speed up builds by using cache layers - sounds good in my case I also want to avoid unnecessary pushing looks like I need to implement language-specific tools to support the workflow, was thinking if pulumi had some universal helpers to do it

I should probably also mention that if it's fully cached, nothing is really being sent in the push, just metadata so the registry can check if it already has the layer