ci(review): keep four runners on review through the daytime (#15769)

The review pool is label-switched: the whole hk1/hk2 fleet carries
`ecs-review` overnight and none of it does between 05:00 and 17:00, so a
review pushed during the day waits until the evening switch. Today that
backlog is 65 queued review runs at 13:45.

Give the daytime a small dedicated pool instead of moving the whole fleet:
the morning switch now runs a new `day` mode, which keeps a named handful of
runners on `ecs-review` and returns everyone else to `ecs-qwen`. Four hk2
runners are the default; a repository variable overrides the list. A name
that matches no registered runner fails the tick instead of silently
shrinking the pool, and an offline kept runner stays untouched like every
other offline runner.

The 17:00 switch still moves the whole fleet to review, so the four merge
into the full pool at night.

.github/scripts/review-runner-schedule.mjs

@@ -6,6 +6,8 @@
  */
 
 // Switch all online hk1/hk2 runners between review and CI, twice a day.
+// `day` keeps a named handful on `ecs-review` so reviews start during the
+// daytime too, and returns the rest of the fleet to CI.
 // Offline runners are skipped; their labels stay unchanged.
 // Running jobs finish normally; labels only control which job starts next.
 
@@ -20,14 +22,38 @@ const MANAGED_LABELS = ['ecs-review', 'ecs-qwen'];
 // copies that can drift would reintroduce the silent no-op it prevents.
 const FLEET_NAME = /^ecs-qwen-hk[12]-\d+$/;
 
-export function planLabels(runners, mode) {
-  if (!['review', 'ci'].includes(mode)) {
-    throw new Error('mode must be review or ci');
+const MODES = ['review', 'day', 'ci'];
+
+// `dayKeep` is the runner-name list that stays on `ecs-review` in `day` mode.
+// Every other managed runner then serves CI, so the daytime review capacity is
+// exactly those runners and never the whole fleet.
+export function planLabels(runners, mode, dayKeep = []) {
+  if (!MODES.includes(mode)) {
+    throw new Error('mode must be review, day or ci');
   }
-  const target = mode === 'review' ? 'ecs-review' : 'ecs-qwen';
+  if (mode === 'day') {
+    if (!dayKeep.length) {
+      throw new Error('mode day needs at least one review-day runner name');
+    }
+    // Typo defence: a name that matches no registered hk1/hk2 runner would
+    // silently shrink the daytime pool every morning. Offline runners stay
+    // allowed — they keep their labels and rejoin when they come back.
+    const registered = new Set(
+      runners.filter((r) => FLEET_NAME.test(r.name)).map((r) => r.name),
+    );
+    const unknown = dayKeep.filter((name) => !registered.has(name));
+    if (unknown.length) {
+      throw new Error(
+        `review-day runner names match no registered hk1/hk2 runner: ${unknown.join(', ')}`,
+      );
+    }
+  }
+  const keep = new Set(mode === 'day' ? dayKeep : []);
   return runners
     .filter((r) => r.status === 'online' && FLEET_NAME.test(r.name))
     .flatMap((r) => {
+      const target =
+        mode === 'review' || keep.has(r.name) ? 'ecs-review' : 'ecs-qwen';
       const labels = r.labels.map((l) => l.name);
       const remove = labels.filter(
         (l) => MANAGED_LABELS.includes(l) && l !== target,
@@ -48,14 +74,21 @@ async function gh(args) {
 }
 
 async function main() {
-  const [repo, mode] = process.argv.slice(2);
+  const [repo, mode, ...rest] = process.argv.slice(2);
   if (!/^[\w.-]+\/[\w.-]+$/.test(repo ?? '')) {
     throw new Error(
-      'usage: review-runner-schedule.mjs <owner/repo> <review|ci>',
+      'usage: review-runner-schedule.mjs <owner/repo> <review|day|ci> [review-day runner names]',
     );
   }
-  // Validate the mode before spending an authenticated call on it.
-  planLabels([], mode);
+  const dayKeep = rest.join(' ').split(/\s+/).filter(Boolean);
+  // Validate the mode (and the day list shape) before spending an
+  // authenticated call on it.
+  if (!MODES.includes(mode)) {
+    throw new Error('mode must be review, day or ci');
+  }
+  if (mode === 'day' && !dayKeep.length) {
+    throw new Error('mode day needs at least one review-day runner name');
+  }
   if (!process.env.RUNNER_ADMIN_TOKEN) {
     throw new Error(
       'RUNNER_ADMIN_TOKEN is empty (needs Administration: write)',
@@ -73,7 +106,7 @@ async function main() {
   if (!runners.some((r) => FLEET_NAME.test(r.name))) {
     throw new Error('no ecs-qwen-hk1-<n> or ecs-qwen-hk2-<n> runner found');
   }
-  const actions = planLabels(runners, mode);
+  const actions = planLabels(runners, mode, dayKeep);
   const results = await Promise.all(
     actions.map(async ({ id, name, add, remove }) => {
       try {
@@ -112,6 +145,7 @@ async function main() {
   );
   const summary = [
     `### hk1/hk2 runner pool: ${mode}`,
+    ...(mode === 'day' ? [`- keeping on review: ${dayKeep.join(', ')}`] : []),
     `- ${actions.length} runners to switch`,
     ...results.map((r) => r.message),
   ].join('\n');

.github/scripts/review-runner-schedule.test.mjs

@@ -82,7 +82,78 @@ describe('review runner schedule', () => {
       [],
     );
     assert.deepEqual(planLabels([runner(1, ['ecs-qwen'])], 'ci'), []);
-    assert.throws(() => planLabels([], 'invalid'), /mode must be review or ci/);
+    assert.throws(
+      () => planLabels([], 'invalid'),
+      /mode must be review, day or ci/,
+    );
+  });
+
+  it('keeps the named runners on review and returns the rest to CI in day mode', () => {
+    assert.deepEqual(
+      planLabels(
+        [
+          runner(1, ['ecs-qwen', 'ecs-autofix'], { name: 'ecs-qwen-hk2-1' }),
+          runner(2, ['ecs-qwen'], { name: 'ecs-qwen-hk2-2' }),
+          runner(3, ['ecs-review'], { name: 'ecs-qwen-hk1-7' }),
+          runner(4, ['ecs-qwen'], { name: 'ecs-qwen-hk1-8' }),
+          runner(5, ['ecs-qwen'], { name: 'ecs-qwen-hk3-9' }),
+        ],
+        'day',
+        ['ecs-qwen-hk2-1'],
+      ),
+      [
+        {
+          id: 1,
+          name: 'ecs-qwen-hk2-1',
+          add: ['ecs-review'],
+          remove: ['ecs-qwen'],
+        },
+        {
+          id: 3,
+          name: 'ecs-qwen-hk1-7',
+          add: ['ecs-qwen'],
+          remove: ['ecs-review'],
+        },
+      ],
+    );
+  });
+
+  it('is idempotent for day mode and rejects a review-day name the fleet lacks', () => {
+    assert.deepEqual(
+      planLabels(
+        [runner(1, ['ecs-review'], { name: 'ecs-qwen-hk2-1' })],
+        'day',
+        ['ecs-qwen-hk2-1'],
+      ),
+      [],
+    );
+    // A typo would otherwise shrink the daytime pool on every morning tick
+    // while the run stays green.
+    assert.throws(
+      () => planLabels([runner(1, ['ecs-qwen'])], 'day', ['ecs-qwen-hk9-1']),
+      /match no registered hk1\/hk2 runner/,
+    );
+    assert.throws(
+      () => planLabels([], 'day', []),
+      /needs at least one review-day runner name/,
+    );
+  });
+
+  it('leaves an offline review-day runner untouched', () => {
+    assert.deepEqual(
+      planLabels(
+        [
+          runner(1, ['ecs-review'], {
+            name: 'ecs-qwen-hk2-1',
+            status: 'offline',
+          }),
+          runner(2, ['ecs-qwen'], { name: 'ecs-qwen-hk2-2' }),
+        ],
+        'day',
+        ['ecs-qwen-hk2-1'],
+      ),
+      [],
+    );
   });
 
   it('wires two daily UTC switches and manual pool selection', () => {
@@ -101,11 +172,20 @@ describe('review runner schedule', () => {
     );
     assert.ok(
       schedule.includes(
-        "github.event.schedule == '0 9 * * *' && 'review' || 'ci'",
+        "github.event.schedule == '0 9 * * *' && 'review' || 'day'",
       ),
     );
     assert.ok(schedule.includes('"$GITHUB_REPOSITORY" "$POOL"'));
     assert.ok(schedule.includes('secrets.RUNNER_ADMIN_PAT'));
+    // The daytime pool. Four names stay on review after the 05:00 switch; an
+    // unset override falls back to them, and a name the fleet does not have
+    // fails the tick instead of shrinking the pool silently.
+    assert.ok(schedule.includes("- 'day'"));
+    assert.match(
+      schedule,
+      /REVIEW_DAY_HOSTS: .*'ecs-qwen-hk2-1 ecs-qwen-hk2-2 ecs-qwen-hk2-3 ecs-qwen-hk2-4'/,
+    );
+    assert.ok(schedule.includes('"${day_hosts[@]}"'));
     // The in-repo half of the fence on the admin PAT. Losing it is silent:
     // a dispatch from any other branch still runs, now executing that
     // branch's copy of the planner with the token in its environment. The

.github/workflows/.size-baseline

@@ -45,7 +45,7 @@
 107075 qwen-fleet-shepherd.yml
 22680 qwen-issue-followup-bot.yml
 5760 qwen-pr-safety-precheck.yml
-2518 qwen-review-runner-schedule.yml
+3243 qwen-review-runner-schedule.yml
 27648 qwen-triage-finalize.yml
 362377 qwen-triage.yml
 15171 release-chrome-extension.yml

.github/workflows/qwen-review-runner-schedule.yml

@@ -1,9 +1,17 @@
 name: 'Qwen Review Runner Schedule'
 
 # Switch every online hk1/hk2 runner at 17:00 / 05:00 Asia/Shanghai.
-# GitHub cron uses UTC. Manual runs can select either pool explicitly.
+# 17:00 puts the whole fleet on review; 05:00 returns it to CI but keeps
+# REVIEW_DAY_HOSTS on review, so four reviews can start during the daytime
+# instead of queueing until the evening. GitHub cron uses UTC.
+# Manual runs can select a pool explicitly.
 # Running jobs finish; only the labels for the next job change.
 #
+# REVIEW_DAY_HOSTS is a space-separated list of hk1/hk2 runner names that
+# stay on `ecs-review` in the `day` pool. Unset falls back to the four hk2
+# runners below; the script fails loudly if a name matches no registered
+# runner, so a typo cannot silently shrink the daytime pool.
+#
 # RUNNER_ADMIN_PAT needs runner-label write access: repo scope for a
 # classic PAT, or repository Administration: write for a fine-grained PAT.
 # Store it in the environment below, restricted to `main`.
@@ -11,8 +19,8 @@ name: 'Qwen Review Runner Schedule'
 
 on:
   schedule:
-    - cron: '0 9 * * *' # 17:00 Asia/Shanghai: enable reviews
-    - cron: '0 21 * * *' # 05:00 Asia/Shanghai: return to CI
+    - cron: '0 9 * * *' # 17:00 Asia/Shanghai: whole fleet on review
+    - cron: '0 21 * * *' # 05:00 Asia/Shanghai: CI, REVIEW_DAY_HOSTS on review
   workflow_dispatch:
     inputs:
       pool:
@@ -22,6 +30,7 @@ on:
         default: 'ci'
         options:
           - 'review'
+          - 'day'
           - 'ci'
 
 permissions:
@@ -59,5 +68,8 @@ jobs:
       - name: 'Switch runner pool'
         env:
           RUNNER_ADMIN_TOKEN: '${{ secrets.RUNNER_ADMIN_PAT }}'
-          POOL: "${{ github.event_name == 'workflow_dispatch' && inputs.pool || (github.event.schedule == '0 9 * * *' && 'review' || 'ci') }}"
-        run: 'node .github/scripts/review-runner-schedule.mjs "$GITHUB_REPOSITORY" "$POOL"'
+          POOL: "${{ github.event_name == 'workflow_dispatch' && inputs.pool || (github.event.schedule == '0 9 * * *' && 'review' || 'day') }}"
+          REVIEW_DAY_HOSTS: "${{ vars.QWEN_REVIEW_DAY_HOSTS || 'ecs-qwen-hk2-1 ecs-qwen-hk2-2 ecs-qwen-hk2-3 ecs-qwen-hk2-4' }}"
+        run: |
+          read -r -a day_hosts <<< "$REVIEW_DAY_HOSTS"
+          node .github/scripts/review-runner-schedule.mjs "$GITHUB_REPOSITORY" "$POOL" "${day_hosts[@]}"