Avoid FD spike after retrying KafkaAdminClient

A caller might call kafka.KafkaAdminClient repeatedly and handle
kafka.errors.NoBrokersAvailable if the broker is not available.

However, each retry will cause 3 extra FD being used. Depends on how
long the caller wait before retry, the FD usage can reach 300~700 before
Python garbage collector collecting those FD.

This commit close those FD early.

kafka/admin/client.py

@@ -209,14 +209,20 @@ def __init__(self, **configs):
             metric_group_prefix='admin',
             **self.config
         )
-        self._client.check_version(timeout=(self.config['api_version_auto_timeout_ms'] / 1000))
-
-        # Get auto-discovered version from client if necessary
-        if self.config['api_version'] is None:
-            self.config['api_version'] = self._client.config['api_version']
-
-        self._closed = False
-        self._refresh_controller_id()
+        try:
+            self._client.check_version(timeout=(self.config['api_version_auto_timeout_ms'] / 1000))
+
+            # Get auto-discovered version from client if necessary
+            if self.config['api_version'] is None:
+                self.config['api_version'] = self._client.config['api_version']
+
+            self._closed = False
+            self._refresh_controller_id()
+        except Exception:
+            self._metrics.close()
+            self._client.close()  # prevent FD leak
+            self._closed = True
+            raise
         log.debug("KafkaAdminClient started.")
 
     def close(self):