refactor: improve config and ssh core and add iac integration tests

infra_cli/config.py

@@ -9,17 +9,23 @@ class Config:
         self.data = self._load()
 
     def _load(self):
-        if not os.path.exists(self.path):
-            # Fallback to local config if exists
-            if os.path.exists("config.yaml"):
-                self.path = os.path.abspath("config.yaml")
-            else:
-                raise FileNotFoundError(f"Config file not found at {self.path}. Please create it based on config.yaml.example")
+        data = {}
+        if os.path.exists(self.path):
+            with open(self.path, 'r') as f:
+                data = yaml.safe_load(f) or {}
+        elif os.path.exists("config.yaml"):
+            self.path = os.path.abspath("config.yaml")
+            with open(self.path, 'r') as f:
+                data = yaml.safe_load(f) or {}
         
-        with open(self.path, 'r') as f:
-            return yaml.safe_load(f)
+        return data
 
     def get(self, key, default=None):
+        # Support ENV overrides: INFRA_PROXMOX_USER -> proxmox.user
+        env_key = "INFRA_" + key.upper().replace('.', '_')
+        if env_key in os.environ:
+            return os.environ[env_key]
+
         parts = key.split('.')
         val = self.data
         for part in parts:
@@ -30,18 +36,7 @@ class Config:
         return val
 
     def get_node(self, node_name):
-        """Helper to get proxmox node details by name or default to first if none provided"""
         nodes = self.get('proxmox.nodes', {})
         if not nodes:
-            # Fallback for old single-host config if present
-            host = self.get('proxmox.host')
-            if host:
-                return {"host": host, "pass": self.get('proxmox.password')}
             return None
-        
-        if not node_name:
-            # Default to first node found
-            return next(iter(nodes.values()))
-        
-        return nodes.get(node_name)
-
+        return nodes.get(node_name)

infra_cli/ssh.py

@@ -1,15 +1,23 @@
 import subprocess
 import os
+import sys
 
 class SSHClient:
-    def __init__(self, host, user="root", key_path=None, password=None):
+    def __init__(self, host, user="root", key_path=None, password=None, timeout=30):
         self.host = host
         self.user = user
         self.key_path = os.path.expanduser(key_path) if key_path else None
         self.password = password
+        self.timeout = timeout
 
     def run(self, cmd, capture=True):
-        ssh_cmd = ["ssh", "-o", "StrictHostKeyChecking=no", "-o", "UserKnownHostsFile=/dev/null"]
+        ssh_cmd = [
+            "ssh", 
+            "-o", "StrictHostKeyChecking=no", 
+            "-o", "UserKnownHostsFile=/dev/null",
+            "-o", f"ConnectTimeout={self.timeout}",
+            "-o", "BatchMode=yes" if not self.password else "BatchMode=no"
+        ]
         
         if self.key_path:
             ssh_cmd += ["-i", self.key_path]
@@ -17,16 +25,26 @@ class SSHClient:
         target = f"{self.user}@{self.host}"
         
         if self.password:
+            # sshpass is required for password auth
             full_cmd = ["sshpass", "-p", self.password] + ssh_cmd + [target, cmd]
         else:
             full_cmd = ssh_cmd + [target, cmd]
 
-        result = subprocess.run(
-            full_cmd,
-            capture_output=capture,
-            text=True
-        )
-        return result
+        try:
+            result = subprocess.run(
+                full_cmd,
+                capture_output=capture,
+                text=True,
+                timeout=self.timeout + 10
+            )
+            return result
+        except subprocess.TimeoutExpired:
+            print(f"Error: SSH command timed out after {self.timeout}s on {self.host}", file=sys.stderr)
+            # Create a mock result for timeout
+            return subprocess.CompletedProcess(full_cmd, 1, "", "Timeout expired")
+        except Exception as e:
+            print(f"Error: SSH execution failed on {self.host}: {e}", file=sys.stderr)
+            return subprocess.CompletedProcess(full_cmd, 1, "", str(e))
 
     def scp_to(self, local_path, remote_path):
         scp_cmd = ["scp", "-o", "StrictHostKeyChecking=no", "-o", "UserKnownHostsFile=/dev/null"]
@@ -40,4 +58,4 @@ class SSHClient:
         else:
             full_cmd = scp_cmd + [local_path, target]
 
-        return subprocess.run(full_cmd, capture_output=True)
+        return subprocess.run(full_cmd, capture_output=True, timeout=self.timeout + 60)