[CI] Extract wait-for-jobs composite action and gate on stage-a-cpu-only (#20641)

This commit is contained in:
Liangsheng Yin
2026-03-15 13:52:48 -07:00
committed by GitHub
parent 23c191afb6
commit 8397a8a301
2 changed files with 141 additions and 120 deletions

View File

@@ -336,48 +336,13 @@ jobs:
outputs:
stage_a_result: ${{ steps.wait.outputs.result }}
steps:
- name: Wait for stage-a-test-1 to complete
- uses: actions/checkout@v4
- uses: ./.github/actions/wait-for-jobs
id: wait
uses: actions/github-script@v7
with:
script: |
const maxWaitMinutes = 240;
const pollIntervalSeconds = 120; // 2 minutes to reduce GH API calls
const maxAttempts = (maxWaitMinutes * 60) / pollIntervalSeconds;
for (let attempt = 0; attempt < maxAttempts; attempt++) {
const jobs = await github.paginate(github.rest.actions.listJobsForWorkflowRun, {
owner: context.repo.owner,
repo: context.repo.repo,
run_id: context.runId,
per_page: 100,
});
const stageAJob = jobs.find(job => job.name === 'stage-a-test-1');
if (stageAJob) {
console.log(`stage-a-test-1 status: ${stageAJob.status}, conclusion: ${stageAJob.conclusion}`);
if (stageAJob.status === 'completed') {
if (stageAJob.conclusion === 'success' || stageAJob.conclusion === 'skipped') {
core.setOutput('result', stageAJob.conclusion === 'success' ? 'success' : 'skipped');
return;
} else {
core.setOutput('result', 'failure');
core.setFailed(`stage-a-test-1 ${stageAJob.conclusion}`);
return;
}
}
} else {
console.log('stage-a-test-1 job not found yet');
}
console.log(`Waiting ${pollIntervalSeconds}s... (attempt ${attempt + 1}/${maxAttempts})`);
await new Promise(resolve => setTimeout(resolve, pollIntervalSeconds * 1000));
}
core.setFailed('Timeout waiting for stage-a-test-1');
core.setOutput('result', 'timeout');
stage-name: stage-a
jobs: '["stage-a-test-1", "stage-a-cpu-only"]'
max-wait-minutes: '240'
wait-for-stage-b:
needs: [check-changes, call-gate, wait-for-stage-a]
@@ -396,88 +361,19 @@ jobs:
outputs:
stage_b_result: ${{ steps.wait.outputs.result }}
steps:
- name: Wait for stage-b jobs to complete
- uses: actions/checkout@v4
- uses: ./.github/actions/wait-for-jobs
id: wait
uses: actions/github-script@v7
with:
script: |
const maxWaitMinutes = 480;
const pollIntervalSeconds = 120; // 2 minutes to reduce GH API calls
const maxAttempts = (maxWaitMinutes * 60) / pollIntervalSeconds;
// Stage-b jobs to wait for
const stageBJobs = [
{ prefix: 'stage-b-test-small-1-gpu', expectedCount: 8 }, // partitions 0-7
{ prefix: 'stage-b-test-large-1-gpu', expectedCount: 14 }, // partitions 0-13
{ prefix: 'stage-b-test-large-2-gpu', expectedCount: 4 }, // partitions 0-3
{ prefix: 'stage-b-test-4-gpu-b200', expectedCount: 1 },
];
const totalExpectedJobs = stageBJobs.reduce((sum, j) => sum + j.expectedCount, 0); // 27 total
// Helper to match job names exactly (prefix alone or prefix + " (N)" for matrix jobs)
const matchesPrefix = (jobName, prefix) => {
return jobName === prefix || jobName.startsWith(prefix + ' (');
};
for (let attempt = 0; attempt < maxAttempts; attempt++) {
const jobs = await github.paginate(github.rest.actions.listJobsForWorkflowRun, {
owner: context.repo.owner,
repo: context.repo.repo,
run_id: context.runId,
per_page: 100,
});
let allCompleted = true;
let anyFailed = false;
let failedJobs = [];
let completedCount = 0;
let totalCount = 0;
for (const { prefix, expectedCount } of stageBJobs) {
const matchingJobs = jobs.filter(job => matchesPrefix(job.name, prefix));
// Check existing jobs for failures first (fail fast)
for (const job of matchingJobs) {
totalCount++;
console.log(`${job.name}: status=${job.status}, conclusion=${job.conclusion}`);
if (job.status !== 'completed') {
allCompleted = false;
} else {
completedCount++;
if (job.conclusion !== 'success' && job.conclusion !== 'skipped') {
anyFailed = true;
failedJobs.push(job.name);
}
}
}
if (matchingJobs.length < expectedCount) {
console.log(`${prefix}: found ${matchingJobs.length}/${expectedCount} jobs (waiting for more)`);
allCompleted = false;
}
}
console.log(`Progress: ${completedCount}/${totalCount} jobs completed (expected ${totalExpectedJobs})`);
// Fail fast if any jobs failed (don't wait for all jobs to be created)
if (anyFailed) {
core.setOutput('result', 'failure');
core.setFailed(`Stage-b jobs failed: ${failedJobs.join(', ')}`);
return;
}
if (allCompleted && totalCount >= totalExpectedJobs) {
core.setOutput('result', 'success');
return;
}
console.log(`Waiting ${pollIntervalSeconds}s... (attempt ${attempt + 1}/${maxAttempts})`);
await new Promise(resolve => setTimeout(resolve, pollIntervalSeconds * 1000));
}
core.setFailed('Timeout waiting for stage-b jobs');
core.setOutput('result', 'timeout');
stage-name: stage-b
jobs: |
[
{"prefix": "stage-b-test-small-1-gpu", "expected_count": 8},
{"prefix": "stage-b-test-large-1-gpu", "expected_count": 14},
{"prefix": "stage-b-test-large-2-gpu", "expected_count": 4},
{"prefix": "stage-b-test-4-gpu-b200", "expected_count": 1}
]
max-wait-minutes: '480'
# =============================================== PR Gate ====================================================
call-gate: