Workload Checks Suite #211
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
# Workload Checks Suite | |
# | |
# Runs Vector Workload Checks. | |
# | |
# Runs on: | |
# - scheduled UTC midnight Tues-Sat | |
# - on demand by a PR comment matching either of: | |
# - '/ci-run-regression' | |
# - '/ci-run-all' | |
# (the comment issuer must be a member of the Vector GH team) | |
# | |
# This workflow runs the collection of our workload checks, using the latest Vector nightly image, | |
# which depends on when the workflow is invoked. | |
# | |
# The goal is to establish a baseline of check results for a variety of cases | |
# and visualize trends for important Vector use cases. | |
name: Workload Checks Suite | |
on: | |
workflow_call: | |
workflow_dispatch: | |
schedule: | |
# At midnight UTC Tue-Sat | |
- cron: '0 0 * * 2-6' | |
env: | |
SINGLE_MACHINE_PERFORMANCE_API: ${{ secrets.SINGLE_MACHINE_PERFORMANCE_API }} | |
jobs: | |
compute-metadata: | |
name: Compute metadata | |
runs-on: ubuntu-latest | |
outputs: | |
replicas: ${{ steps.experimental-meta.outputs.REPLICAS }} | |
warmup-seconds: ${{ steps.experimental-meta.outputs.WARMUP_SECONDS }} | |
total-samples: ${{ steps.experimental-meta.outputs.TOTAL_SAMPLES }} | |
smp-version: ${{ steps.experimental-meta.outputs.SMP_CRATE_VERSION }} | |
lading-version: ${{ steps.experimental-meta.outputs.LADING_VERSION }} | |
target-sha: ${{ steps.git-metadata.outputs.TARGET_SHA }} | |
steps: | |
- uses: actions/checkout@v3 | |
- name: Get git metadata | |
id: git-metadata | |
env: | |
GH_TOKEN: ${{ secrets.GITHUB_TOKEN }} | |
run: | | |
export TARGET_SHA=$(git rev-parse HEAD) | |
echo "TARGET_SHA=${TARGET_SHA}" >> $GITHUB_OUTPUT | |
echo "target sha is: ${TARGET_SHA}" | |
if [ "${TARGET_SHA}" = "" ] ; then | |
echo "TARGET_SHA not found, exiting." | |
exit 1 | |
fi | |
- name: Setup experimental metadata | |
id: experimental-meta | |
run: | | |
export WARMUP_SECONDS="45" | |
export REPLICAS="10" | |
export TOTAL_SAMPLES="600" | |
export SMP_CRATE_VERSION="0.10.0" | |
export LADING_VERSION="0.19.1" | |
echo "warmup seconds: ${WARMUP_SECONDS}" | |
echo "replicas: ${REPLICAS}" | |
echo "total samples: ${TOTAL_SAMPLES}" | |
echo "smp crate version: ${SMP_CRATE_VERSION}" | |
echo "lading version: ${LADING_VERSION}" | |
echo "WARMUP_SECONDS=${WARMUP_SECONDS}" >> $GITHUB_OUTPUT | |
echo "REPLICAS=${REPLICAS}" >> $GITHUB_OUTPUT | |
echo "TOTAL_SAMPLES=${TOTAL_SAMPLES}" >> $GITHUB_OUTPUT | |
echo "SMP_CRATE_VERSION=${SMP_CRATE_VERSION}" >> $GITHUB_OUTPUT | |
echo "LADING_VERSION=${LADING_VERSION}" >> $GITHUB_OUTPUT | |
submit-job: | |
name: Submit workload checks job | |
runs-on: ubuntu-latest | |
needs: | |
- compute-metadata | |
steps: | |
- name: Check status, in-progress | |
env: | |
GH_TOKEN: ${{ secrets.GITHUB_TOKEN }} | |
run: | | |
gh api \ | |
--method POST \ | |
-H "Accept: application/vnd.github+json" \ | |
/repos/${{ github.repository }}/statuses/${{ needs.compute-metadata.outputs.target-sha }} \ | |
-f state='pending' \ | |
-f description='Experiments submitted to the Workload Checks cluster.' \ | |
-f context='Workload Checks Suite / submission' \ | |
-f target_url=${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }} | |
- uses: actions/checkout@v3 | |
- name: Configure AWS Credentials | |
uses: aws-actions/[email protected] | |
with: | |
aws-access-key-id: ${{ secrets.SINGLE_MACHINE_PERFORMANCE_BOT_ACCESS_KEY_ID }} | |
aws-secret-access-key: ${{ secrets.SINGLE_MACHINE_PERFORMANCE_BOT_SECRET_ACCESS_KEY }} | |
aws-region: us-west-2 | |
- name: Login to Amazon ECR | |
id: login-ecr | |
uses: aws-actions/amazon-ecr-login@v2 | |
- name: Download SMP binary | |
run: | | |
aws s3 cp s3://smp-cli-releases/v${{ needs.compute-metadata.outputs.smp-version }}/x86_64-unknown-linux-gnu/smp ${{ runner.temp }}/bin/smp | |
- name: Submit job | |
env: | |
RUST_LOG: info | |
run: | | |
CURRENT_DATE=$(date --utc '+%Y_%m_%d') | |
RUST_LOG_DEBUG="debug,aws_config::profile::credentials=error" | |
chmod +x ${{ runner.temp }}/bin/smp | |
RUST_BACKTRACE=1 RUST_LOG="${RUST_LOG_DEBUG}" ${{ runner.temp }}/bin/smp \ | |
--team-id ${{ secrets.SINGLE_MACHINE_PERFORMANCE_TEAM_ID }} \ | |
job submit-workload \ | |
--lading-version ${{ needs.compute-metadata.outputs.lading-version }} \ | |
--total-samples ${{ needs.compute-metadata.outputs.total-samples }} \ | |
--warmup-seconds ${{ needs.compute-metadata.outputs.warmup-seconds }} \ | |
--replicas ${{ needs.compute-metadata.outputs.replicas }} \ | |
--target-image timberio/vector:nightly-debian \ | |
--target-sha ${{ needs.compute-metadata.outputs.target-sha }} \ | |
--target-config-dir ${{ github.workspace }}/workload-checks \ | |
--target-name vector \ | |
--target-command "/usr/bin/vector" \ | |
--target-environment-variables "DD_HOSTNAME=smp-workload-checks,DD_DD_URL=http://127.0.0.1:9092,DD_API_KEY=00000001" \ | |
--tags smp_status=nightly,client_team="vector",tag_date="${CURRENT_DATE}" \ | |
--submission-metadata ${{ runner.temp }}/submission-metadata | |
- uses: actions/upload-artifact@v3 | |
with: | |
name: vector-submission-metadata | |
path: ${{ runner.temp }}/submission-metadata | |
- name: Await job | |
timeout-minutes: 120 | |
env: | |
RUST_LOG: info | |
run: | | |
chmod +x ${{ runner.temp }}/bin/smp | |
${{ runner.temp }}/bin/smp --team-id ${{ secrets.SINGLE_MACHINE_PERFORMANCE_TEAM_ID }} \ | |
job status \ | |
--wait \ | |
--wait-delay-seconds 60 \ | |
--wait-timeout-minutes 90 \ | |
--submission-metadata ${{ runner.temp }}/submission-metadata | |
- name: Handle cancellation if necessary | |
if: ${{ cancelled() }} | |
env: | |
RUST_LOG: info | |
run: | | |
chmod +x ${{ runner.temp }}/bin/smp | |
${{ runner.temp }}/bin/smp --team-id ${{ secrets.SINGLE_MACHINE_PERFORMANCE_TEAM_ID }} job cancel \ | |
--submission-metadata ${{ runner.temp }}/submission-metadata | |
- name: Check status, cancelled | |
if: ${{ cancelled() }} | |
env: | |
GH_TOKEN: ${{ secrets.GITHUB_TOKEN }} | |
run: | | |
gh api \ | |
--method POST \ | |
-H "Accept: application/vnd.github+json" \ | |
/repos/${{ github.repository }}/statuses/${{ needs.compute-metadata.outputs.target-sha }} \ | |
-f state='failure' \ | |
-f description='Experiments submitted to the Workload Checks cluster cancelled.' \ | |
-f context='Workload Checks Suite / submission' \ | |
-f target_url=${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }} | |
- name: Check status, success | |
env: | |
GH_TOKEN: ${{ secrets.GITHUB_TOKEN }} | |
run: | | |
gh api \ | |
--method POST \ | |
-H "Accept: application/vnd.github+json" \ | |
/repos/${{ github.repository }}/statuses/${{ needs.compute-metadata.outputs.target-sha }} \ | |
-f state='success' \ | |
-f description='Experiments submitted to the Workload Checks cluster successfully.' \ | |
-f context='Workload Checks Suite / submission' \ | |
-f target_url=${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }} | |
- name: Check status, failure | |
if: ${{ failure() }} | |
env: | |
GH_TOKEN: ${{ secrets.GITHUB_TOKEN }} | |
run: | | |
gh api \ | |
--method POST \ | |
-H "Accept: application/vnd.github+json" \ | |
/repos/${{ github.repository }}/statuses/${{ needs.compute-metadata.outputs.target-sha }} \ | |
-f state='success' \ | |
-f description='Experiments submitted to the Workload Checks Suite failed.' \ | |
-f context='Workload Checks Suite / submission' \ | |
-f target_url=${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }} |