File permission error when doing cleanup post-step on `gpu-t4-4-core` runner

[qthequartermasterman]

Contributing guidelines

• I've read the contributing guidelines and wholeheartedly agree

I've found a bug, and:

• The documentation does not mention anything about my problem
• There are no open or closed issues that are related to my problem

Description

When I attempt to use this action on a gpu-t4-4-core, I get this error in the Post Set up Docker (cleanup) step:

Post job cleanup.
Docker daemon logs
Error: EACCES: permission denied, open '/home/runner/setup-docker-action/run-2eab05d5/dockerd.log'
    at Object.runDir [as readFileSync] (node:fs:449:20)
    at Install.__importDefault (/home/runner/work/_actions/docker/setup-docker-action/v4/webpack:/docker-setup-docker/node_modules/@docker/actions-toolkit/lib/docker/install.js:529:1)
    at Generator.next (<anonymous>)
    at /home/runner/work/_actions/docker/setup-docker-action/v4/webpack:/docker-setup-docker/node_modules/@docker/actions-toolkit/lib/docker/install.js:59:1
    at new Promise (<anonymous>)
    at core (/home/runner/work/_actions/docker/setup-docker-action/v4/webpack:/docker-setup-docker/node_modules/@docker/actions-toolkit/lib/docker/install.js:53:1)
    at exports (/home/runner/work/_actions/docker/setup-docker-action/v4/webpack:/docker-setup-docker/node_modules/@docker/actions-toolkit/lib/docker/install.js:529:1)
    at Object.exports (/home/runner/work/_actions/docker/setup-docker-action/v4/webpack:/docker-setup-docker/node_modules/@actions/core/lib/core.js:288:1)
    at Generator.next (<anonymous>)
    at /home/runner/work/_actions/docker/setup-docker-action/v4/webpack:/docker-setup-docker/node_modules/@actions/core/lib/core.js:34:1

This is during the cleanup of the action. Everything before this final step works as expected. I only see it when using the official Github Actions gpu-t4-4-core runner (i.e. others like ubuntu-latest works fine).

Because the cleanup fails, my CI jobs are marked as failures, even though all of the main actions worked as expected.

Expected behaviour

Clean up succeeds on gpu-t4-4-core.

Actual behaviour

Clean up fails with Error: EACCES: permission denied, open '/home/runner/setup-docker-action/run-2eab05d5/dockerd.log'

Repository URL

No response

Workflow run URL

No response

YAML workflow

This is a redacted version of the workflow that has sensitive information. This isn't the exact workflow I'm running. 


---
name: Build Image

on:
  workflow_call:
    inputs:

permissions:
  id-token: write
  contents: read

jobs:
  build_image:
    timeout-minutes: 40
    strategy:
      fail-fast: false
      matrix:
        os: ["gpu-t4-4-core"]
        python-version: ["3.12"]
    runs-on: ${{ matrix.os }}
    services:
      registry:
        image: registry:3
        ports:
          - 5000:5000
    steps:
      - name: Set up Docker
        uses: docker/setup-docker-action@v4
        with:
          daemon-config: |
            {
              "debug": true,
              "features": {
                "containerd-snapshotter": true
              }
            }
      # Set up QEMU so that we can build arm images on x86 runners
      - name: Set up QEMU
        uses: docker/setup-qemu-action@v3
      - name: Set up Docker Buildx
        uses: docker/setup-buildx-action@v3
      - name: Lint Dockerfile
        uses: hadolint/hadolint-action@v3.3.0
        with:
          dockerfile: Dockerfile

Workflow logs

No response

Docker daemon logs

Additional info

I have had similar issues in the past with the gpu-t4-4-core runner and other actions. This is an example: Jimver/cuda-toolkit#373

The solution in that case was for the action to ensure that users other than root can read the log files.

[qthequartermasterman]

I think this is distinct from #152 because this image does have sudo available, so building still works. It's just reading the log file in the cleanup that happens without root permission.