perf(docker): reduce prebuilt docker image size

[Sharrrrk]

Signed-off-by: Shark Liu shark.liu@autocore.ai

Description

Solve issue #2840

Pre-review checklist for the PR author

The PR author must check the checkboxes below when creating the PR.

• I've confirmed the contribution guidelines.
• The PR follows the pull request guidelines.

In-review checklist for the PR reviewers

The PR reviewers must check the checkboxes below before approval.

• The PR follows the pull request guidelines.

Post-review checklist for the PR author

The PR author must check the checkboxes below before merging.

• There are no open discussions or they are tracked via tickets.

After all checkboxes are checked, anyone who has write access can merge the PR.

[kenji-miyake]

@Sharrrrk Thank you so much for this improvement!
Would it be possible for you to split install_devel-related changes into a separate PR because it's feat?
Like feat(setup): add an option to install dev packages.

[Sharrrrk]

@Sharrrrk Thank you so much for this improvement! Would it be possible for you to split install_devel-related changes into a separate PR because it's feat? Like feat(setup): add an option to install dev packages.

Sure, I'll create a separate PR.

[Sharrrrk]

@kenji-miyake Would it be possible that you could do a test of this script before merging. Thanks.

[kenji-miyake]

@Sharrrrk Yes, I'll trigger GitHub Actions.

[doganulus]

Can we test this PR? What's the status?

[kenji-miyake]

Can we test this PR?

@doganulus What do you mean by the word "test"?
To check the workflow, we can trigger it manually.
To try out the generated image, you can build it locally, or you can fork and try this PR.

What's the status?

I'm waiting for the review comments to be applied.
And I'm planning to check this in more detail, but it's difficult for me to spare time for a while. I'm sorry for that, but please wait for a bit more.

[doganulus]

By testing, I mean using the resulting image, for example, with the AWSIM simulator or the scenario simulator. I think a runtime image as in this PR would be an improvement for these use cases. Thank you for the explanation!

[xmfcx]

Following the discussion here: autowarefoundation/autoware-documentation#270

Is it possible to keep the prebuilt image as it is, and add this runtime/release image separately?

[kenji-miyake]

Is it possible to keep the prebuilt image as it is, and add this runtime/release image separately?

@xmfcx Yes, it's possible. But could you explain some reasons/use cases for that, if possible?
Since Autoware changes every day, the prebuilt artifacts cannot be reused generally.

[xmfcx]

Is it possible to keep the prebuilt image as it is, and add this runtime/release image separately?

@xmfcx Yes, it's possible. But could you explain some reasons/use cases for that, if possible? Since Autoware changes every day, the prebuilt artifacts cannot be reused generally.

Ok, since @cyn-liu agreed to adapt the training video to take this into account in autowarefoundation/autoware-documentation#270 (comment)

We don't need to include build artifacts in the new prebuilt image. I take back my request.

[liuXinGangChina]

Hi, @kenji-miyake .
Can we merge this pr?

[kenji-miyake]

@liuXinGangChina cc @Sharrrrk As we moved to Humble, there are conflicts. Could you fix them?

Also, I didn't notice that there were changes after my last review. I'm sorry for that, but I'd appreciate it if you re-request a review by clicking the button.

So, now do you mean that you've addressed my comments and it's ready to merge?
In that case, I'm sorry for my late work, but I'll manage to make time for a review this week.

[Sharrrrk]

Hi, @kenji-miyake , the conflicts is about cuda_base_image, as it has been set to ubuntu:22.04 on main branch, but nvidia/cuda dit not has docker image match the cuda:11.6 + ubuntu22.04. Thank you so much if you could give some advice.

<<<<<<< 2840-reduce-prebuilt-docker-image-size
base_image=ubuntu:20.04
cuda_base_image=nvidia/cuda:11.6.2-runtime-ubuntu20.04

base_image=ubuntu:22.04
cuda_base_image=ubuntu:22.04

main
cuda_version=11.6
cudnn_version=8.4.1.50-1+cuda11.6
tensorrt_version=8.4.2-1+cuda11.6

[kenji-miyake]

@Sharrrrk I'm sorry to be late.
We need to upgrade the CUDA version to 11.8 or 12.0 to use a CUDA image for the base image. However, since there is no arm64 (sbsa) version provided yet, we can't upgrade the version.
See also #3084 (comment).

So, until then, I think using ubuntu:22.04 for the base image is okay.

[Sharrrrk]

@Sharrrrk I'm sorry to be late. We need to upgrade the CUDA version to 11.8 or 12.0 to use a CUDA image for the base image. However, since there is no arm64 (sbsa) version provided yet, we can't upgrade the version. See also #3084 (comment).

So, until then, I think using ubuntu:22.04 for the base image is okay.

Do you mean we should set both base_image and cuda_base_image to ubuntu:22.04 for now?

[kenji-miyake]

@Sharrrrk Yes, exactly. It will be resolved in #3084, but arm64 TensorRT packages aren't released yet. We have to wait for that.

[kenji-miyake]

@Sharrrrk Thank you for updating the PR, and I'm sorry for my late review.
Added additional comments, could you confirm them, please?

[kenji-miyake]

@Sharrrrk I believe everything is okay. Thank you for your work and patience.