Enhance SSH handling in session management

- Introduced `probe_ssh` function for quick SSH checks, improving the handling of server bootstrap scenarios.
- Updated `cmd_up` to utilize `probe_ssh`, enhancing the logic for managing server states based on SSH availability.
- Improved logging for server state transitions and error handling during SSH connection attempts.
- Adjusted timeout settings in `wait_ssh` for better performance and reliability in SSH key acceptance checks.

src/gpu_rent/ssh_ops.py

@@ -28,6 +28,48 @@ def wait_tcp(host: str, port: int, timeout: float = 300.0) -> None:
     raise CloudError(f"TCP {host}:{port} не открылся за {int(timeout)} с ({last})")
 
 
+def probe_ssh(cfg: Config, host: str, attempts: int = 3) -> str:
+    """Quick SSH check. Returns 'ok' | 'auth' | 'down'."""
+    import logging
+
+    logging.getLogger("paramiko").setLevel(logging.CRITICAL)
+    logging.getLogger("paramiko.transport").setLevel(logging.CRITICAL)
+    try:
+        with socket.create_connection((host, 22), timeout=8):
+            pass
+    except OSError:
+        return "down"
+
+    key = str(cfg.ssh_private_key_path)
+    auth_seen = False
+    for _ in range(max(1, attempts)):
+        client = paramiko.SSHClient()
+        client.set_missing_host_key_policy(paramiko.AutoAddPolicy())
+        try:
+            client.connect(
+                hostname=host,
+                username=cfg.ssh_user,
+                key_filename=key,
+                timeout=12,
+                banner_timeout=20,
+                auth_timeout=15,
+                allow_agent=False,
+                look_for_keys=False,
+            )
+            client.close()
+            return "ok"
+        except Exception as exc:
+            name = type(exc).__name__
+            if "Authentication" in name or "Authentication" in str(exc):
+                auth_seen = True
+            try:
+                client.close()
+            except Exception:
+                pass
+            time.sleep(2)
+    return "auth" if auth_seen else "down"
+
+
 def wait_ssh(cfg: Config, host: str, timeout: float = 900.0) -> None:
     """Wait until sshd accepts our key. Paramiko banner noise is muted."""
     import logging
@@ -70,17 +112,13 @@ def wait_ssh(cfg: Config, host: str, timeout: float = 900.0) -> None:
             if attempt == 1 or attempt % 6 == 0:
                 print(f"жду SSH {cfg.ssh_user}@{host}… ({name})", flush=True)
             # Key never injected (boot-from-volume): don't burn full timeout.
-            if auth_fails >= 8:
+            if auth_fails >= 3:
                 raise CloudError(
                     f"SSH {cfg.ssh_user}@{host}: ключ отклонён (AuthenticationException). "
                     "Nova keypair не попал в authorized_keys при boot-from-volume. "
-                    "Останови up (Ctrl+C), затем снова: gpu-rent up --yes "
-                    "(create теперь с config_drive + user_data). "
-                    "Или в консоли панели (root): "
-                    f"добавь содержимое {cfg.ssh_private_key_path}.pub в "
-                    f"/home/{cfg.ssh_user}/.ssh/authorized_keys"
+                    "up пересоздаст compute с user_data (Base64), диски оставит."
                 ) from exc
-            time.sleep(8)
+            time.sleep(5)
     raise CloudError(
         f"SSH {cfg.ssh_user}@{host} не принял ключ за {int(timeout)} с ({last}). "
         "Часто cloud-init ещё поднимает sshd на GPU-образе. "