Created singleton Kubernetes API clients avoid memory leak

Author: apple

agent/tasks.py

@@ -1,4 +1,5 @@
 import logging
+import threading
 
 import requests
 from celery import shared_task
@@ -13,19 +14,59 @@
 K8S_API_TIMEOUT = 30
 HTTP_REQUEST_TIMEOUT = 30
 
+# Kubernetes client cache (singleton pattern)
+_k8s_clients_cache = {}
+_k8s_config_lock = threading.Lock()
+_k8s_config_loaded = False
 
-def get_pod_details(namespace='drdroid'):
-    # Import kubernetes here to avoid loading at module import time
-    from kubernetes import client, config
 
-    try:
-        # Try to load in-cluster config first, then local config
-        try:
-            config.load_incluster_config()
-        except config.ConfigException:
-            config.load_kube_config()
+def _get_k8s_clients():
+    """
+    Get or create singleton Kubernetes API clients.
+    Reuses the same client instances to prevent connection leaks.
+    Thread-safe for Celery workers.
+    """
+    global _k8s_config_loaded, _k8s_clients_cache
+
+    # Fast path: clients already initialized
+    if _k8s_clients_cache:
+        return _k8s_clients_cache['v1'], _k8s_clients_cache['custom']
+
+    # Slow path: initialize clients (thread-safe)
+    with _k8s_config_lock:
+        # Double-check after acquiring lock
+        if _k8s_clients_cache:
+            return _k8s_clients_cache['v1'], _k8s_clients_cache['custom']
+
+        from kubernetes import client, config
+
+        # Load Kubernetes config only once
+        if not _k8s_config_loaded:
+            try:
+                config.load_incluster_config()
+                logger.info("Loaded in-cluster Kubernetes config")
+            except config.ConfigException:
+                config.load_kube_config()
+                logger.info("Loaded local Kubernetes config")
+            _k8s_config_loaded = True
 
-        v1 = client.CoreV1Api()
+        # Create singleton API clients
+        _k8s_clients_cache['v1'] = client.CoreV1Api()
+        _k8s_clients_cache['custom'] = client.CustomObjectsApi()
+
+        logger.info("Initialized Kubernetes API clients (singleton)")
+
+    return _k8s_clients_cache['v1'], _k8s_clients_cache['custom']
+
+
+def get_pod_details(namespace='drdroid'):
+    """
+    Fetch pod details from Kubernetes API.
+    Uses cached API clients to prevent connection leaks.
+    """
+    try:
+        # Get singleton clients (reuses existing instances)
+        v1, custom_api = _get_k8s_clients()
 
         # Get all pods in the namespace with timeout
         pods = v1.list_namespaced_pod(namespace, _request_timeout=K8S_API_TIMEOUT)
@@ -34,7 +75,6 @@ def get_pod_details(namespace='drdroid'):
         # Get pod metrics for CPU/memory usage
         pod_metrics = {}
         try:
-            custom_api = client.CustomObjectsApi()
             metrics = custom_api.list_namespaced_custom_object(
                 group="metrics.k8s.io",
                 version="v1beta1",