Support for ARM64 Docker Images

[romga]

Currently, Databricks only provides Docker images for x86_64 (AMD64) architectures. With the increasing adoption of ARM64-based cloud instances (such as AWS Graviton) and ARM-based development environments (like Apple Silicon Macs), it would be beneficial for Databricks to provide official ARM64 Docker images.

[takomyr]

Here is the change that allow building images for ARM64 (like Apple Silicon Macs)

diff --git a/ubuntu/minimal/Dockerfile b/ubuntu/minimal/Dockerfile
index db52e0b..79f442e 100644
--- a/ubuntu/minimal/Dockerfile
+++ b/ubuntu/minimal/Dockerfile
@@ -36,7 +36,7 @@ RUN apt-get install -y zulu17=$JDK17_VERSION zulu17-jre=$JDK17_VERSION zulu17-jr
         zulu17-jdk-headless=$JDK17_VERSION zulu17-doc=$JDK17_VERSION zulu17-ca=$JDK17_VERSION zulu17-ca-jre=$JDK17_VERSION zulu17-ca-jre-headless=$JDK17_VERSION \
         zulu17-ca-jdk=$JDK17_VERSION zulu17-ca-jdk-headless=$JDK17_VERSION zulu17-ca-doc=$JDK17_VERSION

-RUN update-java-alternatives -s zulu8-ca-amd64
+RUN update-java-alternatives -s zulu8-ca-$(dpkg --print-architecture)

 # This will install the cert store provided by ubuntu openjdk in case it's needed
 # it's installed under /etc/ssl/certs/java/cacerts

E.g. to get databricksruntime/python:15.4-LTS built for Mac

• clone the repo
• checkout release-15.4-LTS branch
• apply the patch
• in ubuntu/minimal docker build -t databricksruntime/minimal:15.4-LTS .
• in ubuntu/python docker build -t databricksruntime/python:15.4-LTS .

[tkeller-moxe]

Would also love to see this get pushed up to mainstream. We have started to build our own images to support our developers workflow and would much prefer a static base to pull from

[jamie-cho]

I would also like to see this merged in as well. It is not just for developer convenience, but can also be a significant cost savings on larger workloads.

[sshpuntoff]

Hi - @xinzhao-db / @justindu-db
I see that you two are the most recently active maintainers on this repository.
Are you guys the right people to tag, or is there a better point of contact about getting this issue prioritized?

[pareenaverma]

Hi @xinzhao-db @justindu-db I’m from Arm and came across this open issue while looking into Arm enablement for container images.
Given the increasing adoption of Arm-based infrastructure, we’re interested in understanding the status of Arm support for these images. Are there currently any technical blockers preventing Arm builds from being added to the repository, or is it mainly a prioritization question?
If there are specific challenges around building or validating multi-arch images, we’d be happy to follow up and see how we might help.
Thanks!