Fix handling of exceptions when osquery goes away (#83)

This resolves an issue where the extension does not shut down if osquery
goes away unexpectedly (without sending a shutdown signal via Thrift).

The scenario could be reliably reproduced by running `osqueryi`,
connecting an extension, and then sending a `SIGKILL` to `osqueryi`. The
exception thrown in the `start_watcher` function would be of type
`thrift.transport.TTransport.TTransportException`, and would cause the
watcher thread to exit without exiting the rest of the program.

This is a possible fix for issues that users have experienced with
extensions reconnecting after the Watchdog kills osquery.

Author: zwass

osquery/management.py

@@ -192,12 +192,21 @@ def start_watcher(client, interval):
         while True:
             status = client.extension_manager_client().ping()
             if status.code != 0:
+                logging.error("Ping received nonzero status: %d", status)
                 break
             time.sleep(interval)
-    except socket.error:
-        # The socket was torn down.
-        pass
-    os._exit(0)
+
+    except socket.error as e:
+        logging.error("Ping received socket.error: %s", e)
+
+    except TTransport.TTransportException as e:
+        logging.error("Ping received thrift.transport.TTransport.TTransportException: %s", e)
+
+    except Exception as e:
+        logging.error("Ping received unknown exception: %s", e)
+
+    finally:
+        os._exit(1)
 
 
 def start_extension(name="<unknown>",