@ddtcorex/dsh-maestro-supervisor 0.1.0 → 0.2.0

package/lib/health-poller.js

@@ -46,13 +46,30 @@ export async function pollHealth(opts = {}) {
             break;
         }
     }
-    // If either fetch failed or log has error, consider down
-    if (fetchError || logError) {
+    // Distinguish FULL (http !=200) vs DEGRADED (http 200 but log has plugin error)
+    if (fetchError) {
         return {
             up: false,
             httpCode,
-            error: logError ?? fetchError,
-            degraded: !!logError && httpCode === 200,
+            error: logError ? `${fetchError} + ${logError}` : fetchError,
+            degraded: false,
+        };
+    }
+    if (logError) {
+        // http 200 but log error → DEGRADED (isolatable), not FULL
+        if (httpCode === 200) {
+            return {
+                up: true,
+                httpCode,
+                error: logError,
+                degraded: true,
+            };
+        }
+        return {
+            up: false,
+            httpCode,
+            error: logError,
+            degraded: false,
         };
     }
     // Also check psAlive as secondary signal — if fetch ok but ps dead, still down
@@ -95,9 +112,16 @@ async function defaultLogTail() {
     try {
         const { readFileSync } = await import('node:fs');
         const { homedir } = await import('node:os');
-        const logPath = `${homedir()}/.dsh.log`;
-        const content = readFileSync(logPath, 'utf-8');
-        return content.slice(-5000);
+        const candidates = [`${homedir()}/.dsh/dsh-web.log`, `${homedir()}/.dsh.log`];
+        for (const logPath of candidates) {
+            try {
+                const content = readFileSync(logPath, 'utf-8');
+                if (content)
+                    return content.slice(-5000);
+            }
+            catch { }
+        }
+        return '';
     }
     catch {
         return '';

package/lib/supervisor.d.ts

@@ -25,6 +25,7 @@ export declare class Supervisor {
     private lastRollback;
     private rollingBack;
     private lastLKGWrite;
+    private lastDegradedNotify;
     private timer;
     constructor(deps: SupervisorDeps);
     tick(): Promise<void>;

package/lib/supervisor.js

@@ -3,12 +3,27 @@ export class Supervisor {
     lastRollback = 0;
     rollingBack = false;
     lastLKGWrite = 0;
+    lastDegradedNotify = 0;
     timer = null;
     constructor(deps) {
         this.deps = deps;
     }
     async tick() {
         const health = await this.deps.pollHealth();
+        // DEGRADED: http 200 but log has plugin error → report, notify, no rollback
+        if (health.degraded) {
+            const now = this.deps.getTime ? this.deps.getTime() : Date.now();
+            if (now - this.lastDegradedNotify < 60000)
+                return;
+            this.lastDegradedNotify = now;
+            try {
+                const ts = new Date().toISOString().replace(/[:.]/g, '-');
+                const reportPath = await this.deps.writeReport({ ts, health, action: `degraded — ${health.error ?? 'plugin'}` }).catch(() => '');
+                await this.deps.notify(`DEGRADED: ${health.error ?? 'plugin'} (report: ${reportPath})`).catch(() => { });
+            }
+            catch { }
+            return;
+        }
         if (health.up) {
             // Throttle LKG writes to at most once per 5 minutes
             const now = this.deps.getTime ? this.deps.getTime() : Date.now();

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ddtcorex/dsh-maestro-supervisor",
-  "version": "0.1.0",
+  "version": "0.2.0",
   "description": "Supervisor daemon for DSH Web resilience — auto-detect crashes, rollback to LKG, report",
   "type": "module",
   "bin": {