On Monday this week, the 16th of September 2024, Oracle released Oracle Database 23ai (23.5) Free on Linux for ARM, downloadable as an Oracle Linux 8 RPM file at https://www.oracle.com/database/free/get-started/

This release was very long-awaited and incredibly important for many, including myself, as it was the missing link to run Oracle Database Free as a container image on Mac Books with ARM chips (M-chips) directly. While one could build Docker images for other architectures for quite a while thanks to emulation technologies like QEUM, these, unfortunately, never quite managed to emulate all the instructions that an Oracle Database requires, leading to unpredictable crashes of the software. Likewise, one could run Oracle Database Free with emulation technologies like colima
but these too, usually resulted in bad performance and sometimes crashes. Thanks to a proper port, emulation is no longer needed. I can now build arm-native Docker images for Oracle Database Free and anybody can run these native images in their ARM environments.
Both container registries, docker.io/gvenzl/oracle-free and ghcr.io/gvenzl/oracle-free, have been updated with the ARM-based images. There are a couple of nuances to this worth knowing:
Multi-Platform images
Going forward, the gvenzl/oracle-free repositories will container multi-platform images for all the popular tags that were there before: slim
, slim-faststart
, latest
, latest-faststart
, full
, full-faststart
, 23-slim
, 23-slim-faststart
, 23
, 23-faststart
, 23-full
, 23-full-faststart
That means that you can just continue to use, e.g. docker pull gvenzl/oracle-free:slim
and the container runtime will automatically detect your platform and identify the correct image to pull. In other words, you have to change nothing in what you have been doing so far – it just works.

Architecture-specific images
A disclaimer upfront: I’m not sure how well this is going to work going forward, time will tell, so this might be subject to change!
Within the registries, you will also find architecture-specific images. All tags that you have known are present with an appended -arm64
or -amd64
qualifier, e.g.:


As the average user, this should not concern you at all. But I thought it might be worthwhile knowing about it if, for whatever reason, you want to pull a specific architecture directly.
Last but not least, I want to say thank you to all of you who have eagerly and patiently waited for this moment to come. I know that certainly for myself, this is one of the moments that makes me happy, relieved, proud, and so much more.

Pingback: Running Oracle Database 23ai Free on Raspberry Pi 5 via Podman – Gerald on IT
Hi, the image works fine on a M3 as long as I use the typical SQL workload. If I try to load a ONNX model to use the AI feature, I get this error:
BEGIN
DBMS_VECTOR.LOAD_ONNX_MODEL(
directory => ‘DM_DUMP’,
file_name => ‘all_MiniLM_L12_v2.onnx’,
model_name => ‘ALL_MINILM_L12_V2’);
END;
/
BEGIN
*
ERROR in Zeile 1:
ORA-00600: Interner Fehlercode, Argumente: [kgsnxInitCtx_lazy:library.load.fail], [], [], [], [], [], [], [], [], [], [], []
ORA-06512: in “SYS.DBMS_DATA_MINING”, Zeile 5767
ORA-06512: in “SYS.DBMS_VECTOR”, Zeile 1804
ORA-06512: in Zeile 2
Did you encounter such a behavior, and if yes how did you solve it?
Thanks in advance
Hi Andreas,
Thanks for using these images!
Yes, I did encounter this and this is currently a product restriction, see https://forums.oracle.com/ords/apexds/post/vector-search-not-supported-on-23ai-db-on-mac-4664
Unfortunately, the only option is to wait for a newer version of Oracle Database Free where this functionality is provided.
Thanks,