Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[k8s] GPU Image debloat #3665

Merged
merged 6 commits into from
Jun 18, 2024
Merged

[k8s] GPU Image debloat #3665

merged 6 commits into from
Jun 18, 2024

Conversation

romilbhardwaj
Copy link
Collaborator

@romilbhardwaj romilbhardwaj commented Jun 14, 2024

Optimizes the Kubernetes GPU image for >2x faster provisioning (87s vs 201s). The new image is 1.96G compressed vs 6.39G compressed for the previous image.

# Master
sky launch -y -c test --cloud kubernetes --gpus T4:1 -- nvidia-smi  7.72s user 4.54s system 6% cpu 3:21.66 total

# This image
sky launch -y -c test2 --cloud kubernetes --gpus T4:1 -- nvidia-smi  4.74s user 4.30s system 9% cpu 1:27.87 total

Also adds kubectl to the dependencies pre-installed in the image for faster controller setup.

Closes #3435 and #2461.

Related - skypilot-org/skypilot-catalog#72

Tested:

  • Kubernetes smoke tests (in progress)

@romilbhardwaj
Copy link
Collaborator Author

romilbhardwaj commented Jun 14, 2024

Updated CUDA image to 12.1.1 since that will be supported till Feb 2025, and is at parity with cu121 offered on GCP.

Running smoke tests again.

EDIT - smoke tests pass - pytest -v tests/test_smoke.py --kubernetes

Copy link
Collaborator

@Michaelvll Michaelvll left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for reducing the GPU image @romilbhardwaj! LGTM.

Dockerfile_k8s_gpu Show resolved Hide resolved
Dockerfile_k8s_gpu Outdated Show resolved Hide resolved
@romilbhardwaj
Copy link
Collaborator Author

Thanks @Michaelvll! Will merge after skypilot-org/skypilot-catalog#72 is approved and merged.

@romilbhardwaj romilbhardwaj merged commit c73da25 into master Jun 18, 2024
20 checks passed
@romilbhardwaj romilbhardwaj deleted the k8s_image_optimizations branch June 18, 2024 22:34
Michaelvll pushed a commit that referenced this pull request Aug 23, 2024
* optimize GPU image

* revert changes

* add kubectl to cpu image

* add git to dependencies

* fixes and parity b/w cpu and gpu images

* comments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

[k8s] Include kubectl in default dependencies in k8s image
2 participants