databricksConnectDocker
databricksConnectDocker copied to clipboard
Add support for ARM64 in Dockerfile
This change comes due to this issue, as currently the image is not performant for ARM64 platform
The change allows to pass arguments of what Conda and Tini versions to install. The default args remain the same, but we can pass the following two to buld properly for ARM64:
CONDA_ARCH=aarch64
TINI_ARCH=tini-arm64
If I see this correctly, except the errors here (do you have an idea why it is failing?), you replace the image completely with the ARM one? I would suggest building both, an ARM and non-ARM version. For this, it can be a parameter passed to the dockerfile. I would expect in the end to have e.g. 9.1.14 ARM and non-ARM, 10.4.6 ARM and non-ARM images... Does it make sense?
and please, rebase the history, when this is done. 👍
I want to add a pipeline stage to build an ARM image, but leave AMD as default in the pipeline that runs during the PR validation, but it seems the ubuntu-latest VM cannot build ARM images in DevOps Azure Pipelines
I tried a few variations to build with --platform linux/arm64/v8
and base image specifically as arm64v8/openjdk:8-jdk
, but it is failing all the time, I attached a few images below for context, too
It seems there is no way to build ARM with an architecture that is AMD underneath. Based on this issue, it looks like there is currently no DevOps agent that would support ARM easily

