fix: 修复任务生命周期与调度就绪，优化执行和构建开销 (#3069)

* fix: harden task lifecycle and scheduler readiness

* fix: confine log writes to the configured log directory

* fix: verify complete build artifacts and untracked inputs

* fix: reconcile scheduler state and make stop win startup races

* fix: isolate cron generations and serialize scheduler recovery

back/schedule/addCron.ts

@@ -69,6 +69,15 @@ const addCron = (
     return;
   }
 
+  // Recovery replaces the whole snapshot, including deletions and disabled jobs.
+  // Validation above must finish before touching the previous schedule.
+  if (call.request.replace) {
+    for (const jobs of scheduleStacks.values()) {
+      for (const job of jobs) job?.cancel();
+    }
+    scheduleStacks.clear();
+  }
+
   // ===== 第二遍：注册所有任务 =====
   for (const item of call.request.crons) {
     const { id, schedule, command, extra_schedules, name } = item;

back/schedule/client.ts

@@ -1,4 +1,4 @@
-import { credentials } from '@grpc/grpc-js';
+import { credentials, status, Metadata } from '@grpc/grpc-js';
 import {
   AddCronRequest,
   AddCronResponse,
@@ -9,7 +9,41 @@ import {
 import config from '../config';
 import { getGrpcCerts } from '../config/grpcCerts';
 
+import { HealthService } from '../protos/health';
+import { SchedulerReadiness } from '../shared/schedulerReadiness';
+
 class Client {
+  readonly readiness = new SchedulerReadiness(() => this.probe());
+
+  private async waitForReady(timeoutMs: number) {
+    try {
+      await new Promise<void>((resolve, reject) => {
+        this.client.waitForReady(Date.now() + timeoutMs, (err) =>
+          err ? reject(err) : resolve()
+        );
+      });
+    } catch (error) {
+      this.readiness.invalidate();
+      throw Object.assign(
+        error instanceof Error ? error : new Error(String(error)),
+        { status: 503 }
+      );
+    }
+  }
+
+  private async probe(): Promise<void> {
+    await this.waitForReady(1000);
+    await new Promise<void>((resolve, reject) => {
+      this.client.makeUnaryRequest(
+        HealthService.check.path,
+        HealthService.check.requestSerialize,
+        HealthService.check.responseDeserialize,
+        { service: 'scheduler' },
+        { deadline: Date.now() + 1000 },
+        (err, res) => err ? reject(err) : res?.status === 1 ? resolve() : reject(new Error('Scheduler unavailable')),
+      );
+    });
+  }
   private _client: CronClient | null = null;
 
   private get client(): CronClient {
@@ -28,22 +62,40 @@ class Client {
     return this._client;
   }
 
-  addCron(request: AddCronRequest['crons']): Promise<AddCronResponse> {
+  async addCron(
+    request: AddCronRequest['crons'],
+    replace = false
+  ): Promise<AddCronResponse> {
+    await this.waitForReady(2000);
     return new Promise((resolve, reject) => {
-      this.client.addCron({ crons: request }, (err, res) => {
-        if (err) {
-          reject(err);
+      this.client.addCron(
+        { crons: request, replace },
+        new Metadata(),
+        { deadline: Date.now() + 5000 },
+        (err, res) => {
+          if (err) {
+            if (err.code === status.UNAVAILABLE) {
+              this.readiness.invalidate();
+              Object.assign(err, { status: 503 });
+            }
+            return reject(err);
+          }
+          resolve(res);
         }
-        resolve(res);
-      });
+      );
     });
   }
 
-  delCron(request: DeleteCronRequest['ids']): Promise<DeleteCronResponse> {
+  async delCron(request: DeleteCronRequest['ids']): Promise<DeleteCronResponse> {
+    await this.waitForReady(2000);
     return new Promise((resolve, reject) => {
-      this.client.delCron({ ids: request }, (err, res) => {
+      this.client.delCron({ ids: request }, new Metadata(), { deadline: Date.now() + 5000 }, (err, res) => {
         if (err) {
-          reject(err);
+          if (err.code === status.UNAVAILABLE) {
+            this.readiness.invalidate();
+            Object.assign(err, { status: 503 });
+          }
+          return reject(err);
         }
         resolve(res);
       });

back/schedule/health.ts

@@ -52,6 +52,9 @@ const check = async (
   callback: sendUnaryData<HealthCheckResponse>,
 ) => {
   switch (call.request.service) {
+    // Local scheduler liveness only: never call HTTP from this probe.
+    case 'scheduler':
+      return callback(null, { status: 1 });
     case 'cron': {
       const healthUrl = `http://localhost:${config.port}${
         config.baseUrl || ''