Skip to content

Self-hosted runner (check runner status) #1537

Self-hosted runner (check runner status)

Self-hosted runner (check runner status) #1537

name: Self-hosted runner (check runner status)
# Note that each job's dependencies go into a corresponding docker file.
#
# For example for `run_all_tests_torch_cuda_extensions_gpu` the docker image is
# `huggingface/transformers-pytorch-deepspeed-latest-gpu`, which can be found at
# `docker/transformers-pytorch-deepspeed-latest-gpu/Dockerfile`
on:
repository_dispatch:
schedule:
# run per hour
- cron: "0 */1 * * *"
env:
TRANSFORMERS_IS_CI: yes
jobs:
check_runner_status:
name: Check Runner Status
runs-on: ubuntu-latest
outputs:
offline_runners: ${{ steps.set-offline_runners.outputs.offline_runners }}
steps:
- name: Checkout transformers
uses: actions/checkout@v3
with:
fetch-depth: 2
- name: Check Runner Status
run: python utils/check_self_hosted_runner.py --target_runners single-gpu-ci-runner-docker,multi-gpu-ci-runner-docker,single-gpu-scheduled-ci-runner-docker,multi-scheduled-scheduled-ci-runner-docker,single-gpu-doctest-ci-runner-docker --token ${{ secrets.ACCESS_REPO_INFO_TOKEN }}
- id: set-offline_runners
name: Set output for offline runners
if: ${{ always() }}
run: |
offline_runners=$(python3 -c 'fp = open("offline_runners.txt"); failed = fp.read(); fp.close(); print(failed)')
echo "offline_runners=$offline_runners" >> $GITHUB_OUTPUT
send_results:
name: Send results to webhook
runs-on: ubuntu-latest
needs: check_runner_status
if: ${{ failure() }}
steps:
- name: Preliminary job status
shell: bash
run: |
echo "Runner availability: ${{ needs.check_runner_status.result }}"
- uses: actions/checkout@v3
- uses: actions/download-artifact@v3
- name: Send message to Slack
env:
CI_SLACK_BOT_TOKEN: ${{ secrets.CI_SLACK_BOT_TOKEN }}
CI_SLACK_CHANNEL_ID: ${{ secrets.CI_SLACK_CHANNEL_ID }}
CI_SLACK_CHANNEL_ID_DAILY: ${{ secrets.CI_SLACK_CHANNEL_ID_DAILY }}
CI_SLACK_CHANNEL_DUMMY_TESTS: ${{ secrets.CI_SLACK_CHANNEL_DUMMY_TESTS }}
CI_SLACK_REPORT_CHANNEL_ID: ${{ secrets.CI_SLACK_CHANNEL_ID_DAILY }}
ACCESS_REPO_INFO_TOKEN: ${{ secrets.ACCESS_REPO_INFO_TOKEN }}
CI_EVENT: runner status check
RUNNER_STATUS: ${{ needs.check_runner_status.result }}
OFFLINE_RUNNERS: ${{ needs.check_runner_status.outputs.offline_runners }}
# We pass `needs.setup.outputs.matrix` as the argument. A processing in `notification_service.py` to change
# `models/bert` to `models_bert` is required, as the artifact names use `_` instead of `/`.
run: |
pip install slack_sdk
python utils/notification_service.py