-
Notifications
You must be signed in to change notification settings - Fork 688
build: OPS-810: add dynamo frontend image w/EPP support #4150
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
Conversation
Signed-off-by: Tushar Sharma <tusharma@nvidia.com>
Signed-off-by: Tushar Sharma <tusharma@nvidia.com>
Signed-off-by: Tushar Sharma <tusharma@nvidia.com>
Signed-off-by: Tushar Sharma <tusharma@nvidia.com>
Signed-off-by: Tushar Sharma <tusharma@nvidia.com>
Signed-off-by: Tushar Sharma <tusharma@nvidia.com>
Signed-off-by: Tushar Sharma <tusharma@nvidia.com>
Signed-off-by: Tushar Sharma <tusharma@nvidia.com>
saturley-hall
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I am not seeing significant deviations from what is in the dev container here. Other than the EPP additions why are we reinstalling everything versus taking the environment and copying the nixl installs versus just building atop the dev image?
|
@saturley-hall, We don't have to reinstall everything, we could copy the virtual environment from the dynamo base container. My thought was that we eventually want to move away from some of the dev/test dependencies we install in the virtual environment so I went with this approach. I think these deps are pretty minimal so I'm just going to copy the venv instead and we can revisit this later if our dev dependencies become significant. Also, the reason to not build on top of the dev images is due to the several build dependencies from that container. This container doesn't need CUDA, cargo, rust or any dependent packages required to build those pacakges so we use a minimal ubuntu image. |
Signed-off-by: Tushar Sharma <tusharma@nvidia.com>
Signed-off-by: Tushar Sharma <tusharma@nvidia.com>
Signed-off-by: Tushar Sharma <tusharma@nvidia.com>
Overview:
This pull request introduces a Dockerfile for the frontend container. This container can be used for quick dynamo prototyping with mock workers and leveraging EPP for GAIE. This frontend image uses a CUDA-less base image, takes the dynamo built artifacts from the dynamo-base container, and installs the EPP binary on top of it. In total, the size of the container is approximately 4.53 GB. To avoid adding complexity to our current build.sh logic and our base dockerfiles (used for CI + additional efforts), We've created a seperate
Dockerfile.frontendwhich gets built via docker buildx. Since EPP is built itself in it's own Dockerfile, we have to avoid building EPP in our base dynamo dockerfile itself. Another reason is that this increases build time by adding additional dependencies which aren't required. The README.md is also updated for instructions on how to build the frontend container.Details:
Summary by CodeRabbit