Skip to content

Self-hosted, Self-Improving, Federated Everything App — dev management, agent orchestration, persistent mind, digital twin, second brain, and creative pipeline from a single dashboard over Tailscale to port the user into the OS

fix: transition-log CoS status probe in system health snapshot

[._.]/ Adam Eivy5a041d43d2c9
diff --git a/server/routes/health.test.js b/server/routes/health.test.js
index b3977c9cda..4853e12883 100644
--- a/server/routes/health.test.js
+++ b/server/routes/health.test.js
@@ -271,6 +271,43 @@ describe('System Health Routes', () => {
}
});
+ it('transition-logs CoS status probe failure once on entry and once on recovery without poll noise', async () => {
+ // Ensure healthy initial state
+ getStatus.mockResolvedValueOnce({ running: true, paused: false, activeAgents: 0 });
+ await request(app).get('/api/system/health/details');
+
+ const errorSpy = vi.spyOn(console, 'error').mockImplementation(() => {});
+ const logSpy = vi.spyOn(console, 'log').mockImplementation(() => {});
+
+ try {
+ // First failure logs unavailable once
+ getStatus.mockRejectedValueOnce(Object.assign(new Error('connection refused'), { code: 'ECONNREFUSED' }));
+ const first = await request(app).get('/api/system/health/details');
+ expect(first.status).toBe(200);
+ expect(first.body.cos).toBeNull();
+ expect(errorSpy).toHaveBeenCalledWith('❌ CoS health probe unavailable (source=cos-status, code=ECONNREFUSED)');
+
+ errorSpy.mockClear();
+
+ // Second consecutive failure does NOT emit an error (deduplicated / throttled)
+ getStatus.mockRejectedValueOnce(Object.assign(new Error('connection refused'), { code: 'ECONNREFUSED' }));
+ const second = await request(app).get('/api/system/health/details');
+ expect(second.status).toBe(200);
+ expect(second.body.cos).toBeNull();
+ expect(errorSpy).not.toHaveBeenCalled();
+
+ // Recovery logs once
+ getStatus.mockResolvedValueOnce({ running: true, paused: false, activeAgents: 1 });
+ const third = await request(app).get('/api/system/health/details');
+ expect(third.status).toBe(200);
+ expect(third.body.cos).toMatchObject({ running: true });
+ expect(logSpy).toHaveBeenCalledWith('✅ CoS health probe recovered (source=cos-status)');
+ } finally {
+ errorSpy.mockRestore();
+ logSpy.mockRestore();
+ }
+ });
+
it('reports an unreadable PM2 as unavailable rather than zero processes (#8431)', async () => {
listProcessesStrict.mockResolvedValueOnce(null);
checkHealth.mockResolvedValueOnce({ connected: true, hasSchema: true });
diff --git a/server/services/systemHealthSnapshot.js b/server/services/systemHealthSnapshot.js
index 39ee04821a..9c2b73f238 100644
--- a/server/services/systemHealthSnapshot.js
+++ b/server/services/systemHealthSnapshot.js
@@ -67,7 +67,7 @@ async function loadHealthSettings() {
};
}
-// Last observed availability per CoS read ('cos-queue' / 'cos-agents'), so a
+// Last observed availability per CoS read ('cos-status' / 'cos-queue' / 'cos-agents'), so a
// persistent failure logs once on entry and once on recovery, not on every read.
const cosProbeAvailability = new Map();
@@ -94,10 +94,10 @@ export async function getSystemHealthSnapshot() {
console.error(`❌ App health probe failed (operation=getAppStatusSummary, code=${code})`);
return failedProbe;
}),
- cos.getStatus().catch((error) => {
- console.error(`❌ Chief of Staff health probe failed: ${error.message}`);
- return failedProbe;
- }),
+ cos.getStatus().then(
+ (status) => { noteCosProbe('cos-status', true); return status; },
+ (error) => { noteCosProbe('cos-status', false, error); return failedProbe; }
+ ),
// Queue depth is read here rather than taken off `getStatus()`, which has no
// such field — `cosStatus.queueLength` never existed, so the widget's
// "N queued" was dead and always rendered 0. Both reads ride the same