fix(sandbox): make interactive connect resilient on stopped/resumed sandboxes

.changeset/fix-interactive-connect-resume.md

@@ -0,0 +1,5 @@
+---
+"sandbox": patch
+---
+
+Fix `sandbox connect` hanging or failing on a stopped/resumed sandbox. The interactive shell now surfaces `attach()` failures instead of swallowing them once the connection handshake lands, always stops the spinner on teardown (so a failure can no longer hang the process), and includes the in-sandbox server's stderr when the interactive server exits early. The in-sandbox `vc-interactive-server` also health-checks a reused server before trusting a leftover config file, so a stale `/tmp/vercel/interactive/config.json` restored from a snapshot no longer causes it to connect to a dead socket.

packages/pty-tunnel-server/modes/remote.go

@@ -33,10 +33,46 @@ var _ Bootstrapper = (*ExternalProcessBootstrapper)(nil)
 // GetOrCreateServer implements Bootstrapper.
 func (e *ExternalProcessBootstrapper) GetOrCreateServer() (info config.ServerInfo, err error) {
 	info, err = config.VerifyConnection(e.ConfigPath)
+	if err == nil {
+		// A live PID is not sufficient evidence that the server is usable. Across
+		// a snapshot/resume the config file is restored from the snapshot while
+		// the original server process is gone: the recorded PID may have been
+		// reused by an unrelated process, or a memory-restored daemon may no
+		// longer be serving.
+		if healthErr := e.pingServer(info.Port); healthErr == nil {
+			return info, nil
+		} else {
+			e.Logger.Info(
+				"Existing server config is stale (failed health check), spawning a new server",
+				"port", info.Port,
+				"pid", info.PID,
+				"error", healthErr,
+			)
+		}
+	}
+	return e.spawnServer()
+}
+
+func (e *ExternalProcessBootstrapper) pingServer(port int) error {
+	ctx, cancel := context.WithTimeout(context.Background(), 500*time.Millisecond)
+	defer cancel()
+
+	url := fmt.Sprintf("http://localhost:%d/health", port)
+	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
 	if err != nil {
-		return e.spawnServer()
+		return err
 	}
-	return
+
+	res, err := http.DefaultClient.Do(req)
+	if err != nil {
+		return err
+	}
+	defer res.Body.Close()
+
+	if res.StatusCode != http.StatusOK {
+		return fmt.Errorf("unexpected status %d", res.StatusCode)
+	}
+	return nil
 }
 
 func (e *ExternalProcessBootstrapper) spawnServer() (info config.ServerInfo, err error) {
@@ -66,6 +102,11 @@ func (e *ExternalProcessBootstrapper) spawnServer() (info config.ServerInfo, err
 	basename := path.Join(os.TempDir(), fmt.Sprintf("pty-tunnel-server-%d", time.Now().Nanosecond()))
 	e.Logger.Debug("Creating temporary files for server stdout/stderr", "basename", basename)
 
+	// Remove any leftover config before starting the new server.
+	if rmErr := os.Remove(e.ConfigPath); rmErr != nil && !os.IsNotExist(rmErr) {
+		e.Logger.Debug("Could not remove stale server config", "path", e.ConfigPath, "error", rmErr)
+	}
+
 	e.Logger.Info("Spawning new pty-tunnel-server process", "args", cmd.Args)
 	cmd.SysProcAttr = &syscall.SysProcAttr{
 		Setpgid: true, // new process group

packages/sandbox/src/interactive-shell/interactive-shell.ts

@@ -193,7 +193,7 @@ export async function startInteractiveShell(options: {
 
   using progress = acquireRelease(
     () => ora({ discardStdin: false }).start(),
-    (s) => s.clear(),
+    (s) => s.stop(),
   );
 
   progress.text = "Setting up sandbox environment";
@@ -226,7 +226,12 @@ export async function startInteractiveShell(options: {
   });
 
   await Promise.all([
-    throwIfCommandPrematurelyExited(command, waitForProcess.signal),
+    // `throwIfCommandPrematurelyExited` rejects with an abort error once the
+    // connection is established; swallow only that interruption (a genuine
+    // premature exit is thrown before the abort and still propagates).
+    throwIfCommandPrematurelyExited(command, waitForProcess.signal).catch(
+      waitForProcess.ignoreInterruptions,
+    ),
     attach({
       sandbox: options.sandbox,
       progress,
@@ -237,28 +242,40 @@ export async function startInteractiveShell(options: {
           printCommand(options.execution[0], options.execution.slice(1)),
         ),
     }),
-  ]).catch(waitForProcess.ignoreInterruptions);
+  ]);
 }
 
 async function throwIfCommandPrematurelyExited(
   command: Command,
   signal: AbortSignal,
 ) {
+  let exitCode: number;
   try {
-    const { exitCode } = await command.wait({ signal });
-    throw new Error(
-      [
-        `Interactive shell failed to start (exit code: ${exitCode}).`,
-        `${chalk.bold("hint:")} The sandbox may have timed out or encountered an error.`,
-        "╰▶ Check sandbox status with `sandbox list` or view logs for details.",
-      ].join("\n"),
-    );
+    ({ exitCode } = await command.wait({ signal }));
   } catch (err) {
     if (signal.aborted) {
       return;
     }
     throw err;
   }
+
+  // The interactive server process exited before a connection was established.
+  // Surface its stderr.
+  let serverError = "";
+  try {
+    serverError = (await command.stderr({ signal })).trim();
+  } catch {
+    // Best-effort: never let reading the failure output mask the real error.
+  }
+
+  throw new Error(
+    [
+      `Interactive shell failed to start (exit code: ${exitCode}).`,
+      `${chalk.bold("hint:")} The sandbox may have timed out or encountered an error.`,
+      ...(serverError ? [chalk.dim(serverError)] : []),
+      "╰▶ Check sandbox status with `sandbox list` or view logs for details.",
+    ].join("\n"),
+  );
 }
 
 async function attach({

packages/vercel-sandbox/src/sandbox.ts

@@ -938,6 +938,10 @@ export class Sandbox {
       resume: true,
       signal,
     });
+    // Do not update the session if it was not resumed.
+    if (!response.json.resumed && this.session) {
+      return;
+    }
     this.session = new Session({
       client,
       routes: response.json.routes,