# 1.9.0 image change

**URL:** <https://discuss.ray.io/t/1-9-0-image-change/4384>\
**Category:** Ray Clusters\
**Created:** [December 7, 2021, 4:49pm UTC](https://discuss.ray.io/t/1-9-0-image-change/4384 "2021-12-07T16:49:57Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![rdier](https://avatars.discourse-cdn.com/v4/letter/r/b3f665/32.png) [@rdier](https://discuss.ray.io/u/rdier)\
**Post date:** [December 7, 2021, 4:49pm UTC](https://discuss.ray.io/t/1-9-0-image-change/4384/1 "2021-12-07T16:49:57Z")

</div>

I’m not sure if this is the correct area for this, so feel free to move/point me to the correct section for this.

I have been helping my users make use of the ray-ml images in docker and k8s clusters, and was going to build a 1.9.0 image for them to begin testing with, however the 1.{X\<9}.Y have been built using an Ubuntu Focal (20.04) base image,`BASE_IMAGE=ubuntu:focal`, where as the 1.X.Y-gpu builds have been based off of the nvidia cuda images, and specifically the bionic (18.04) version of the images, `BASE_IMAGE=nvidia/cuda:11.2.0-cudnn8-devel-ubuntu18.04`.

The issue I am now facing is that I have internal libraries and dependencies built for focal, that this change to **only** have a bionic image arbitrarily hinders my ability to deploy `ray-ml:1.9.0`.

I know that the [build matrix](https://github.com/ray-project/ray/pull/19505) is now getting a bit unwieldy,

> ray: {3.6, 3.7, 3.8, 3.9} x {cpu, cu101, cu102, cu110, cu111, cu112}.  
> ray-ml: {3.6, 3.7, 3.8} x {cu112}

However, I think it would be extremely beneficial for users like myself, and I have to imagine others, if there were a focal/bionic modifier, similar to the python and cuda/cpu/gpu modifier, for focal or bionic.

I would think that you would want the `ray` and `ray-ml` branches to be roughly similar, but now there is a sizeable gulf between the two images.  
And given that there are focal images available from nvidia to base off of:  
`BASE_IMAGE=nvidia/cuda:11.2.0-cudnn8-devel-ubuntu18.04`  
`BASE_IMAGE=nvidia/cuda:11.2.0-cudnn8-devel-ubuntu20.04`

It shouldn’t be too difficult to either add this to the build matrix, or forklift the cuda images from the bionic/18 branches to the focal/20 branches.