@ddtcorex/dsh-maestro-supervisor 0.6.6 → 0.6.7

package/lib/debug-agent.js

@@ -217,20 +217,40 @@ function defaultWriteFile(p, c) {
     writeFileSync(p, c, 'utf-8');
 }
 async function defaultDryBoot() {
+    const { execSync } = await import('node:child_process');
+    const { resolveDeepseekHarnessDir } = await import('./paths.js');
+    const { buildKillStalePortsCommand } = await import('./restart-guards.js');
+    const port = Math.floor(19000 + Math.random() * 1000);
+    let tmp = '';
     try {
-        const { execSync } = await import('node:child_process');
-        const { resolveDeepseekHarnessDir } = await import('./paths.js');
         const deepseekDir = resolveDeepseekHarnessDir();
-        const tmp = execSync('mktemp -d', { encoding: 'utf-8' }).trim();
-        const port = Math.floor(19000 + Math.random() * 1000);
+        tmp = execSync('mktemp -d', { encoding: 'utf-8' }).trim();
         // Use spawn-based dry-boot to avoid nested quoting hell (prev \\$! / \\$(seq) caused syntax error)
-        const out = execSync(`timeout 8 bash -c 'DSH_HOME=${tmp} pnpm --dir ${deepseekDir} dsh web --port ${port} --no-open >${tmp}/dsh.log 2>&1 & pid=$!; for i in $(seq 1 5); do sleep 1; if curl -s -o /dev/null -w "%{http_code}" http://127.0.0.1:${port}/ 2>&1 | grep -q 200; then kill $pid 2>/dev/null || true; wait $pid 2>/dev/null || true; echo ok; exit 0; fi; done; kill $pid 2>/dev/null || true; wait $pid 2>/dev/null || true; echo fail; exit 1'`, { encoding: 'utf-8', timeout: 12000 });
-        execSync(`rm -rf ${tmp}`);
+        const out = execSync(`timeout 8 bash -c 'DSH_HOME=${tmp} pnpm --dir ${deepseekDir} dsh web --port ${port} --no-open >${tmp}/dsh.log 2>&1 & pid=$!; for i in $(seq 1 5); do sleep 1; if curl -s -o /dev/null -w "%{http_code}" http://127.0.0.1:${port}/ 2>&1 | grep -q 200; then echo ok; exit 0; fi; done; echo fail; exit 1'`, { encoding: 'utf-8', timeout: 12000 });
         return out.includes('ok');
     }
     catch {
         return false;
     }
+    finally {
+        // `pnpm dsh web` is a pnpm→sh→node tree (see AGENTS.md Known Issues); the
+        // backgrounded job's `$!` above is only the pnpm wrapper, so `kill $pid`
+        // never reached the real node process — it kept running detached,
+        // holding the ephemeral port. Confirmed live 2026-08-31: 3 failed
+        // attempts left 3 orphaned "MainThread" node processes eating 6.4G RAM
+        // for 7+ hours. Resolve the real pid by the port it is actually
+        // listening on instead (same technique as the live-restart kill step).
+        try {
+            execSync(buildKillStalePortsCommand([port]), { timeout: 5000, stdio: 'pipe' });
+        }
+        catch { }
+        if (tmp) {
+            try {
+                execSync(`rm -rf ${tmp}`);
+            }
+            catch { }
+        }
+    }
 }
 async function defaultFetchLLM(prompt) {
     const cfg = await resolveLLMConfig();

package/lib/health-poller.js

@@ -198,6 +198,30 @@ export async function pollHealth(opts = {}) {
                 logTail: logContent.slice(-5000),
             };
         }
+        // Corroborate with a cheap port-liveness check before declaring a crash.
+        // The HTTP fetch shares dsh-web's own event loop, so a busy-but-alive
+        // process (e.g. heavy GitLab-webhook-triggered review work) times out
+        // the same way a genuinely dead one does — but a real crash always frees
+        // the port, while `ss` does not depend on the contended event loop to
+        // answer. Downgrade to DEGRADED (still visible, still escalates after
+        // repeated ticks) instead of forcing an immediate rollback + restartWeb
+        // that would kill an in-flight review for nothing.
+        let portAlive = false;
+        try {
+            portAlive = await psAliveFn();
+        }
+        catch {
+            portAlive = false;
+        }
+        if (portAlive) {
+            return {
+                up: true,
+                httpCode,
+                error: logError ? `${fetchError} + ${logError}` : fetchError,
+                degraded: true,
+                logTail: logContent.slice(-5000),
+            };
+        }
         return {
             up: false,
             httpCode,

package/lib/supervisor.js

@@ -197,7 +197,7 @@ export class Supervisor {
                 return cfg.downThreshold;
         }
         catch { }
-        return 3;
+        return 5;
     }
     async findInterruptedRecent(withinMs) {
         const ms = withinMs ?? this.getResumeWithinMs();

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ddtcorex/dsh-maestro-supervisor",
-  "version": "0.6.6",
+  "version": "0.6.7",
   "description": "Supervisor daemon for DSH Web resilience — auto-detect crashes, rollback to LKG, report",
   "type": "module",
   "main": "./lib/index.js",