You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on Jun 25, 2024. It is now read-only.
Repository navigation
This repository was archived by the owner on Jun 25, 2024. It is now read-only.
Tried to run this on an Apple silicon Mac and it doesn't play nicely as the image runs in emulation mode. When you adapt it for ARM it works great on Apple silicon. Any chance you could do a multi-arch build to support other platforms?
basically, wanted to share with you all here @giqbal@tseaward , that DB connect will probably not be supported for a long time any more. But it will get a much better replacement called Spark Connect. It will be more lightweight client which will execute all the code on the cluster (rather then now executing python locally and pyspark on the cluster).
@alexeyegorov I added a comment in the PR, this might not be as straightforward as first thought with issues in building the image on a non-ARM machine. I will explore Spark Connect, thanks!
Activity
tseaward commentedon Oct 3, 2022
Yes, thanks for this container, I find it very useful 馃憤.
I had the same issue. The only line I needed to change was the following:
databricksConnectDocker/Dockerfile
Line 30 in ac7b50c
Replaced
x86_64with$(ARCH)whereARCH=aarch64.Perhaps
$(uname -m)would also work?giqbal commentedon Oct 3, 2022
@tseaward yes had to make the same change
alexeyegorov commentedon Dec 7, 2022
Would you mind to do this as a MR? @tseaward @giqbal
giqbal commentedon Dec 8, 2022
@alexeyegorov I don't have access to Apple Silicon machine atm. Will do next time I am near one.
augusten commentedon Dec 12, 2022
Hi all! I created a PR in order to support ARM64 here
alexeyegorov commentedon Dec 13, 2022
@augusten thanks for your PR. :)
basically, wanted to share with you all here @giqbal @tseaward , that DB connect will probably not be supported for a long time any more. But it will get a much better replacement called Spark Connect. It will be more lightweight client which will execute all the code on the cluster (rather then now executing python locally and pyspark on the cluster).
augusten commentedon Dec 14, 2022
@alexeyegorov I added a comment in the PR, this might not be as straightforward as first thought with issues in building the image on a non-ARM machine. I will explore Spark Connect, thanks!