feat(doctor): warn about in-flight cron jobs

masatohoshino · web-flow · commit 5ada3acb5a1a · 2026-07-01T06:41:37.000-07:00
Squashed from PR #98620 after updating branch with current main and passing CI.
diff --git a/src/commands/doctor/cron/index.test.ts b/src/commands/doctor/cron/index.test.ts
@@ -392,6 +392,66 @@ describe("maybeRepairLegacyCronStore", () => {
     expectNoteContaining("Examples: alias-pinned -> gpt", "Cron");
   });
 
+  describe("in-flight cron job advisory", () => {
+    const RUNNING_AT_MS = Date.parse("2026-05-01T00:00:00.000Z");
+
+    it("warns about jobs still marked in-flight without touching the store", async () => {
+      const storePath = await makeTempStorePath();
+      await writeCurrentCronStore(storePath, [
+        createCurrentCronJob({ id: "running-job", state: { runningAtMs: RUNNING_AT_MS } }),
+      ]);
+      const prompter = makePrompter(true);
+
+      await maybeRepairLegacyCronStore({
+        cfg: createCronConfig(storePath),
+        options: {},
+        prompter,
+      });
+
+      expectNoteContaining("1 cron job is still marked in-flight", "Cron");
+      expectNoteContaining("shows it as `running`", "Cron");
+      expectNoteContaining("marks such runs interrupted the next time it starts", "Cron");
+      expectNoteContaining("openclaw cron show <id>", "Cron");
+
+      // Observer-only: no repair prompt and the running marker is left untouched.
+      expect(prompter.confirm).not.toHaveBeenCalled();
+      const jobs = await readPersistedJobs(storePath);
+      const state = requireRecord(requirePersistedJob(jobs, 0).state, "cron state");
+      expect(state.runningAtMs).toBe(RUNNING_AT_MS);
+      expect(state.lastRunStatus).toBeUndefined();
+    });
+
+    it("pluralizes the advisory when multiple jobs are in-flight", async () => {
+      const storePath = await makeTempStorePath();
+      await writeCurrentCronStore(storePath, [
+        createCurrentCronJob({ id: "running-a", state: { runningAtMs: RUNNING_AT_MS } }),
+        createCurrentCronJob({ id: "running-b", state: { runningAtMs: RUNNING_AT_MS + 1000 } }),
+      ]);
+
+      await maybeRepairLegacyCronStore({
+        cfg: createCronConfig(storePath),
+        options: {},
+        prompter: makePrompter(true),
+      });
+
+      expectNoteContaining("2 cron jobs are still marked in-flight", "Cron");
+      expectNoteContaining("shows them as `running`", "Cron");
+    });
+
+    it("stays silent when no job is marked in-flight", async () => {
+      const storePath = await makeTempStorePath();
+      await writeCurrentCronStore(storePath, [createCurrentCronJob({ id: "idle-job" })]);
+
+      await maybeRepairLegacyCronStore({
+        cfg: createCronConfig(storePath),
+        options: {},
+        prompter: makePrompter(true),
+      });
+
+      expectNoNoteContaining("still marked in-flight", "Cron");
+    });
+  });
+
   it("repairs legacy cron store fields and migrates notify fallback to webhook delivery", async () => {
     const storePath = await makeTempStorePath();
     await writeCronStore(storePath, [createLegacyCronJob()]);
diff --git a/src/commands/doctor/cron/index.ts b/src/commands/doctor/cron/index.ts
@@ -66,6 +66,22 @@ function errorMessage(err: unknown): string {
   return err instanceof Error ? err.message : String(err);
 }
 
+// Count jobs the store still marks in-flight (`state.runningAtMs` is a number).
+// The scheduler sets this while a run is active and clears it on completion, so a
+// leftover marker (gateway killed mid-run) makes `cron list` show the job as
+// `running` while nothing executes it. Startup marks exactly these runs interrupted
+// (`src/cron/service/ops.ts` `start`), so doctor only reports the count here.
+function countInFlightCronJobs(jobs: Array<Record<string, unknown>>): number {
+  return jobs.filter((job) => {
+    const state = job.state;
+    return (
+      typeof state === "object" &&
+      state !== null &&
+      typeof (state as { runningAtMs?: unknown }).runningAtMs === "number"
+    );
+  }).length;
+}
+
 type LegacyCronRepairState = {
   storePath: string;
   quarantinePath: string;
@@ -387,6 +403,19 @@ export async function maybeRepairLegacyCronStore(params: {
   }
   noteCronModelOverrides({ cfg: params.cfg, jobs: rawJobs, storePath });
 
+  const inFlightCount = countInFlightCronJobs(rawJobs);
+  if (inFlightCount > 0) {
+    const subject = inFlightCount === 1 ? "it" : "them";
+    note(
+      [
+        `${pluralize(inFlightCount, "cron job")} ${inFlightCount === 1 ? "is" : "are"} still marked in-flight (\`state.runningAtMs\` is set), so ${formatCliCommand("openclaw cron list")} shows ${subject} as \`running\`.`,
+        `- If no gateway is currently executing ${subject}, the marker is left over from an interrupted run; the gateway marks such runs interrupted the next time it starts.`,
+        `- Review with ${formatCliCommand("openclaw cron list")} or ${formatCliCommand("openclaw cron show <id>")}.`,
+      ].join("\n"),
+      "Cron",
+    );
+  }
+
   const normalized = normalizeStoredCronJobs(rawJobs);
   const notifyCount = rawJobs.filter((job) => job.notify === true).length;
   const dreamingStaleCount = countStaleDreamingJobs(rawJobs);
