gitarsenal-cli 1.9.72 → 1.9.74

package/python/test_container.py

@@ -3,15 +3,20 @@ import time
 import secrets
 import string
 import modal
+import sys
 
 def generate_random_password(length=16):
     """Generate a random password for SSH access"""
+    print(f"[DEBUG] Generating random password of length {length}")
     alphabet = string.ascii_letters + string.digits + "!@#$%^&*"
     password = ''.join(secrets.choice(alphabet) for i in range(length))
+    print(f"[DEBUG] Password generated successfully (length: {len(password)})")
     return password
 
+print("[DEBUG] Starting container setup...")
+
 image = (
-    modal.Image.from_registry("nvidia/cuda:12.4.0-devel-ubuntu22.04", add_python="3.11")
+    modal.Image.from_registry("nvidia/cuda:12.2.0-devel-ubuntu22.04", add_python="3.11")
     .apt_install("openssh-server", "sudo", "curl", "wget", "git", "vim", "htop", "tmux", "nvtop")
     .pip_install("cupy-cuda12x", "torch", "transformers")
     .run_commands(
@@ -23,32 +28,118 @@ image = (
     )
 )
 
+print("[DEBUG] Image configuration completed")
+
 app = modal.App("cuda-ssh-container", image=image)
+print("[DEBUG] Modal app created")
 
 @app.function(gpu="A10G", timeout=3600)
 def start_ssh():
+    print("[DEBUG] Starting SSH function...")
+    
     # Generate SSH password
-    password = generate_random_password()
-    subprocess.run(["bash", "-c", f"echo 'root:{password}' | chpasswd"], check=True)
+    print("[DEBUG] Step 1: Generating SSH password")
+    try:
+        password = generate_random_password()
+        print(f"[DEBUG] Password generation successful")
+    except Exception as e:
+        print(f"[ERROR] Failed to generate password: {e}")
+        return
+    
+    # Set root password
+    print("[DEBUG] Step 2: Setting root password")
+    try:
+        result = subprocess.run(["bash", "-c", f"echo 'root:{password}' | chpasswd"], 
+                              check=True, capture_output=True, text=True)
+        print(f"[DEBUG] Password set successfully. Return code: {result.returncode}")
+    except subprocess.CalledProcessError as e:
+        print(f"[ERROR] Failed to set password: {e}")
+        print(f"[ERROR] stdout: {e.stdout}")
+        print(f"[ERROR] stderr: {e.stderr}")
+        return
+    except Exception as e:
+        print(f"[ERROR] Unexpected error setting password: {e}")
+        return
     
     # Start SSH server
-    subprocess.Popen(["/usr/sbin/sshd", "-D"])
-    time.sleep(2)
+    print("[DEBUG] Step 3: Starting SSH server")
+    try:
+        ssh_process = subprocess.Popen(["/usr/sbin/sshd", "-D"])
+        print(f"[DEBUG] SSH server started with PID: {ssh_process.pid}")
+        time.sleep(2)
+        print("[DEBUG] Waited 2 seconds for SSH server to initialize")
+    except Exception as e:
+        print(f"[ERROR] Failed to start SSH server: {e}")
+        return
     
     # Test CUDA
-    subprocess.run(["nvidia-smi"])
-    subprocess.run(["nvcc", "--version"])
+    print("[DEBUG] Step 4: Testing CUDA availability")
+    try:
+        print("[DEBUG] Running nvidia-smi...")
+        nvidia_result = subprocess.run(["nvidia-smi"], capture_output=True, text=True, timeout=30)
+        print(f"[DEBUG] nvidia-smi return code: {nvidia_result.returncode}")
+        if nvidia_result.stdout:
+            print(f"[DEBUG] nvidia-smi output:\n{nvidia_result.stdout}")
+        if nvidia_result.stderr:
+            print(f"[DEBUG] nvidia-smi stderr:\n{nvidia_result.stderr}")
+    except subprocess.TimeoutExpired:
+        print("[ERROR] nvidia-smi timed out")
+    except Exception as e:
+        print(f"[ERROR] nvidia-smi failed: {e}")
+    
+    try:
+        print("[DEBUG] Running nvcc --version...")
+        nvcc_result = subprocess.run(["nvcc", "--version"], capture_output=True, text=True, timeout=30)
+        print(f"[DEBUG] nvcc return code: {nvcc_result.returncode}")
+        if nvcc_result.stdout:
+            print(f"[DEBUG] nvcc output:\n{nvcc_result.stdout}")
+        if nvcc_result.stderr:
+            print(f"[DEBUG] nvcc stderr:\n{nvcc_result.stderr}")
+    except subprocess.TimeoutExpired:
+        print("[ERROR] nvcc timed out")
+    except Exception as e:
+        print(f"[ERROR] nvcc failed: {e}")
     
     # Forward SSH port
-    with modal.forward(port=22, unencrypted=True) as tunnel:
-        hostname, port = tunnel.tcp_socket
-        print(f"SSH: ssh -p {port} root@{hostname}")
-        print(f"Password: {password}")
-        
-        # Keep alive
-        while True:
-            time.sleep(60)
+    print("[DEBUG] Step 5: Setting up SSH port forwarding")
+    try:
+        with modal.forward(port=22, unencrypted=True) as tunnel:
+            hostname, port = tunnel.tcp_socket
+            print(f"[DEBUG] Tunnel established successfully")
+            print(f"[DEBUG] Hostname: {hostname}, Port: {port}")
+            print(f"\n" + "="*50)
+            print(f"SSH CONNECTION INFO:")
+            print(f"SSH: ssh -p {port} root@{hostname}")
+            print(f"Password: {password}")
+            print(f"="*50 + "\n")
+            
+            # Keep alive with periodic status updates
+            print("[DEBUG] Starting keep-alive loop...")
+            counter = 0
+            while True:
+                counter += 1
+                print(f"[DEBUG] Keep-alive cycle {counter} - Container still running")
+                time.sleep(60)
+                
+    except Exception as e:
+        print(f"[ERROR] Failed to set up port forwarding: {e}")
+        import traceback
+        traceback.print_exc()
+        return
 
 if __name__ == "__main__":
-    with app.run():
-        start_ssh.remote()
+    print("[DEBUG] Script starting from main...")
+    print(f"[DEBUG] Python version: {sys.version}")
+    print(f"[DEBUG] Modal version: {modal.__version__}")
+    
+    try:
+        print("[DEBUG] Starting modal app...")
+        with app.run():
+            print("[DEBUG] Modal app context established")
+            start_ssh.remote()
+    except Exception as e:
+        print(f"[ERROR] Failed to run modal app: {e}")
+        import traceback
+        traceback.print_exc()
+    
+    print("[DEBUG] Script completed")