init commit

dags/connection-check.py

@@ -0,0 +1,43 @@
+from airflow import DAG
+from airflow.decorators import task
+import pendulum
+
+TARGET_HOST = "airflow-k8s-0140-worker-gpu-0.airflow-k8s-0140-worker-gpu"
+
+with DAG(
+    dag_id='debug_gpu_network_check',
+    # UPDATED: New argument name
+    schedule=None,
+    # UPDATED: Explicit pendulum date
+    start_date=pendulum.datetime(2024, 1, 1, tz="UTC"),
+    catchup=False
+) as dag:
+
+    @task(queue='gpu')
+    def check_network():
+        import subprocess
+        import socket
+        
+        print(f"I am running on hostname: {socket.gethostname()}")
+        print(f"Attempting to resolve and ping: {TARGET_HOST}")
+        
+        # 1. DNS Check
+        try:
+            ip = socket.gethostbyname(TARGET_HOST)
+            print(f"DNS Resolution Success: {TARGET_HOST} -> {ip}")
+        except Exception as e:
+            print(f"DNS Resolution Failed: {e}")
+            raise
+
+        # 2. Ping Check
+        cmd = ["ping", "-c", "4", TARGET_HOST]
+        result = subprocess.run(cmd, capture_output=True, text=True)
+        
+        if result.returncode == 0:
+            print("Ping Success!")
+            print(result.stdout)
+        else:
+            print("Ping Failed (might be firewall, check if TCP port 29500 works instead)")
+            print(result.stderr)
+
+    check_network()