fix: retry missing config snapshots before skip (#23343) (thanks @lbo728)

src/gateway/config-reload.test.ts

@@ -1,12 +1,15 @@
-import { afterEach, beforeEach, describe, expect, it } from "vitest";
+import chokidar from "chokidar";
+import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
 import { listChannelPlugins } from "../channels/plugins/index.js";
 import type { ChannelPlugin } from "../channels/plugins/types.js";
+import type { ConfigFileSnapshot } from "../config/config.js";
 import { setActivePluginRegistry } from "../plugins/runtime.js";
 import { createTestRegistry } from "../test-utils/channel-plugins.js";
 import {
   buildGatewayReloadPlan,
   diffConfigPaths,
   resolveGatewayReloadSettings,
+  startGatewayConfigReloader,
 } from "./config-reload.js";
 
 describe("diffConfigPaths", () => {
@@ -163,3 +166,134 @@ describe("resolveGatewayReloadSettings", () => {
     expect(settings.debounceMs).toBe(300);
   });
 });
+
+type WatcherHandler = () => void;
+type WatcherEvent = "add" | "change" | "unlink" | "error";
+
+function createWatcherMock() {
+  const handlers = new Map<WatcherEvent, WatcherHandler[]>();
+  return {
+    on(event: WatcherEvent, handler: WatcherHandler) {
+      const existing = handlers.get(event) ?? [];
+      existing.push(handler);
+      handlers.set(event, existing);
+      return this;
+    },
+    emit(event: WatcherEvent) {
+      for (const handler of handlers.get(event) ?? []) {
+        handler();
+      }
+    },
+    close: vi.fn(async () => {}),
+  };
+}
+
+function makeSnapshot(partial: Partial<ConfigFileSnapshot> = {}): ConfigFileSnapshot {
+  return {
+    path: "/tmp/openclaw.json",
+    exists: true,
+    raw: "{}",
+    parsed: {},
+    resolved: {},
+    valid: true,
+    config: {},
+    issues: [],
+    warnings: [],
+    legacyIssues: [],
+    ...partial,
+  };
+}
+
+describe("startGatewayConfigReloader", () => {
+  beforeEach(() => {
+    vi.useFakeTimers();
+  });
+
+  afterEach(() => {
+    vi.useRealTimers();
+    vi.restoreAllMocks();
+  });
+
+  it("retries missing snapshots and reloads once config file reappears", async () => {
+    const watcher = createWatcherMock();
+    vi.spyOn(chokidar, "watch").mockReturnValue(watcher as unknown as never);
+
+    const readSnapshot = vi
+      .fn<() => Promise<ConfigFileSnapshot>>()
+      .mockResolvedValueOnce(makeSnapshot({ exists: false, raw: null, hash: "missing-1" }))
+      .mockResolvedValueOnce(
+        makeSnapshot({
+          config: {
+            gateway: { reload: { debounceMs: 0 } },
+            hooks: { enabled: true },
+          },
+          hash: "next-1",
+        }),
+      );
+
+    const onHotReload = vi.fn(async () => {});
+    const onRestart = vi.fn();
+    const log = {
+      info: vi.fn(),
+      warn: vi.fn(),
+      error: vi.fn(),
+    };
+
+    const reloader = startGatewayConfigReloader({
+      initialConfig: { gateway: { reload: { debounceMs: 0 } } },
+      readSnapshot,
+      onHotReload,
+      onRestart,
+      log,
+      watchPath: "/tmp/openclaw.json",
+    });
+
+    watcher.emit("unlink");
+    await vi.runOnlyPendingTimersAsync();
+    await vi.advanceTimersByTimeAsync(150);
+
+    expect(readSnapshot).toHaveBeenCalledTimes(2);
+    expect(onHotReload).toHaveBeenCalledTimes(1);
+    expect(onRestart).not.toHaveBeenCalled();
+    expect(log.info).toHaveBeenCalledWith("config reload retry (1/2): config file not found");
+    expect(log.warn).not.toHaveBeenCalledWith("config reload skipped (config file not found)");
+
+    await reloader.stop();
+  });
+
+  it("caps missing-file retries and skips reload after retry budget is exhausted", async () => {
+    const watcher = createWatcherMock();
+    vi.spyOn(chokidar, "watch").mockReturnValue(watcher as unknown as never);
+
+    const readSnapshot = vi
+      .fn<() => Promise<ConfigFileSnapshot>>()
+      .mockResolvedValue(makeSnapshot({ exists: false, raw: null, hash: "missing" }));
+
+    const onHotReload = vi.fn(async () => {});
+    const onRestart = vi.fn();
+    const log = {
+      info: vi.fn(),
+      warn: vi.fn(),
+      error: vi.fn(),
+    };
+
+    const reloader = startGatewayConfigReloader({
+      initialConfig: { gateway: { reload: { debounceMs: 0 } } },
+      readSnapshot,
+      onHotReload,
+      onRestart,
+      log,
+      watchPath: "/tmp/openclaw.json",
+    });
+
+    watcher.emit("unlink");
+    await vi.runAllTimersAsync();
+
+    expect(readSnapshot).toHaveBeenCalledTimes(3);
+    expect(onHotReload).not.toHaveBeenCalled();
+    expect(onRestart).not.toHaveBeenCalled();
+    expect(log.warn).toHaveBeenCalledWith("config reload skipped (config file not found)");
+
+    await reloader.stop();
+  });
+});

src/gateway/config-reload.ts

@@ -44,6 +44,8 @@ const DEFAULT_RELOAD_SETTINGS: GatewayReloadSettings = {
   mode: "hybrid",
   debounceMs: 300,
 };
+const MISSING_CONFIG_RETRY_DELAY_MS = 150;
+const MISSING_CONFIG_MAX_RETRIES = 2;
 
 const BASE_RELOAD_RULES: ReloadRule[] = [
   { prefix: "gateway.remote", kind: "none" },
@@ -268,19 +270,22 @@ export function startGatewayConfigReloader(opts: {
   let running = false;
   let stopped = false;
   let restartQueued = false;
+  let missingConfigRetries = 0;
 
-  const schedule = () => {
+  const scheduleAfter = (wait: number) => {
     if (stopped) {
       return;
     }
     if (debounceTimer) {
       clearTimeout(debounceTimer);
     }
-    const wait = settings.debounceMs;
     debounceTimer = setTimeout(() => {
       void runReload();
     }, wait);
   };
+  const schedule = () => {
+    scheduleAfter(settings.debounceMs);
+  };
 
   const runReload = async () => {
     if (stopped) {
@@ -298,9 +303,18 @@ export function startGatewayConfigReloader(opts: {
     try {
       const snapshot = await opts.readSnapshot();
       if (!snapshot.exists) {
-        opts.log.warn("config reload skipped (config file not found; may be mid-write)");
+        if (missingConfigRetries < MISSING_CONFIG_MAX_RETRIES) {
+          missingConfigRetries += 1;
+          opts.log.info(
+            `config reload retry (${missingConfigRetries}/${MISSING_CONFIG_MAX_RETRIES}): config file not found`,
+          );
+          scheduleAfter(MISSING_CONFIG_RETRY_DELAY_MS);
+          return;
+        }
+        opts.log.warn("config reload skipped (config file not found)");
         return;
       }
+      missingConfigRetries = 0;
       if (!snapshot.valid) {
         const issues = snapshot.issues.map((issue) => `${issue.path}: ${issue.message}`).join(", ");
         opts.log.warn(`config reload skipped (invalid config): ${issues}`);