vrnetlab
vrnetlab copied to clipboard
Reduce size of vMX by "sparsing" the vFPC image
Looks like the vFPC image is 2.2GB on recent JUNOS. It's mostly empty and if it's converted to a sparse .qcow2 it takes like 100MB.
Problem is when can we do this? As soon as we copy the vfpc image its committed to a layer and then the data is there. It need to be done before the docker build but I don't want to rely on qemu-img being available to us.
I see docker has a new multi-stage build that could probably help us by allowing us to the run the build itself inside of a docker image while still not implicitly including all layers in the final output image.