Skip to content

Build the docker image on an OSDC runner - #3970

Draft
huydhn wants to merge 1 commit into
mainfrom
osdc/docker-builds
Draft

huydhn wants to merge 1 commit into
mainfrom
osdc/docker-builds

Conversation

@huydhn

@huydhn huydhn commented Sep 25, 2026

Copy link
Copy Markdown
before after
runner [self-hosted, linux.2xlarge] mt-l-x86iavx512-8-64
build docker build on the runner, then push docker buildx build --push against the out-of-cluster BuildKit pool

The OSDC runner is a container, so it has no Docker daemon to build with. Builds go to the shared BuildKit pool instead, via docker-build-remote-buildkit — the same path torchtitan and ExecuTorch use. build.sh gains a REMOTE_BUILDKIT switch for it; without the variable set it behaves exactly as before, so local runs are unaffected.

That also drops the steps that only made sense on an EC2 host: Clean workspace, setup-ssh, setup-linux, teardown-linux, and calculate-docker-image.

Pushing to ECR needs id-token: write plus an OIDC login as role/arc. The image already lives in 308535385114.dkr.ecr.us-east-1.amazonaws.com/tutorials/, so no new repository is involved — but the role's resource list has not been confirmed to cover it.

_build-tutorials-base.yml's manager job is still on [self-hosted, linux.2xlarge] and is not part of this change.

Authored with Claude Code.

The OSDC runner is a container and has no Docker daemon, so the build goes to
the shared BuildKit pool instead. build.sh gains a REMOTE_BUILDKIT switch for
it; without the variable set it behaves exactly as before.

That also drops the steps that only made sense on an EC2 host: Clean
workspace, setup-ssh, setup-linux and teardown-linux.
@pytorch-bot

pytorch-bot Bot commented Sep 25, 2026 •

Copy link
Copy Markdown

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/tutorials/3970

Note: Links to docs will display an error until the docs builds have been completed.

❌ 1 New Failure

As of commit 0cb9858 with merge base 65e2e1a (image):

NEW FAILURE - The following job has failed:

This comment was automatically generated by Dr. CI and updates every 15 minutes.

@meta-cla meta-cla Bot added the cla signed label Sep 25, 2026

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant