Enhance SSH configuration and security group management

- Added optional SSH security-group CIDR configuration in `env.example` to allow broader access if needed.
- Improved `guess_operator_cidr` function to support multiple IP retrieval services and allow overriding CIDR via environment variable.
- Refactored security group management to ensure idempotency and proper logging of security group rules.
- Introduced helper function `_get_volume_or_none` for better volume management and error handling.
- Updated `wait_ssh` function to enhance SSH connection handling with improved timeout settings and logging.
- Added state reset logic in `cmd_up` to handle scenarios where the server is not found in the cloud.

src/gpu_rent/ssh_ops.py

@@ -28,12 +28,20 @@ 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 wait_ssh(cfg: Config, host: str, timeout: float = 420.0) -> None:
-    wait_tcp(host, 22, timeout=min(timeout, 240))
+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
+
+    logging.getLogger("paramiko").setLevel(logging.CRITICAL)
+    logging.getLogger("paramiko.transport").setLevel(logging.CRITICAL)
+
+    wait_tcp(host, 22, timeout=min(timeout, 300))
     deadline = time.time() + timeout
     key = str(cfg.ssh_private_key_path)
     last = None
+    attempt = 0
     while time.time() < deadline:
+        attempt += 1
         client = paramiko.SSHClient()
         client.set_missing_host_key_policy(paramiko.AutoAddPolicy())
         try:
@@ -41,9 +49,11 @@ def wait_ssh(cfg: Config, host: str, timeout: float = 420.0) -> None:
                 hostname=host,
                 username=cfg.ssh_user,
                 key_filename=key,
-                timeout=12,
-                banner_timeout=12,
-                auth_timeout=12,
+                timeout=20,
+                banner_timeout=60,
+                auth_timeout=30,
+                allow_agent=False,
+                look_for_keys=False,
             )
             client.close()
             return
@@ -53,8 +63,15 @@ def wait_ssh(cfg: Config, host: str, timeout: float = 420.0) -> None:
                 client.close()
             except Exception:
                 pass
-            time.sleep(5)
-    raise CloudError(f"SSH {cfg.ssh_user}@{host} не принял ключ ({last})")
+            if attempt == 1 or attempt % 6 == 0:
+                # Progress without paramiko traceback spam.
+                print(f"жду SSH {cfg.ssh_user}@{host}… ({type(exc).__name__})", flush=True)
+            time.sleep(8)
+    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. "
+        "Если SG узкий (VPN/WARP): в .env GPU_RENT_SSH_CIDR=0.0.0.0/0 и снова up."
+    )
 
 
 def ssh_argv(cfg: Config, host: str, remote: list[str] | None = None) -> list[str]: