Hi all,

I have a Python node that queries information on startup via a service call. Recently, I tried to convert the node to a lifecycle node where the queries are performed in on_configure() (I also tried on_activate()) but execution gets stuck when trying to receive the service response. Probably this is due to the node not being in an active event loop in that state.

Here's a minimal example showing the problem (you need the lifecycle-py package to be installed):

from typing import Optional

import rclpy

from rclpy.lifecycle import Node
from rclpy.lifecycle import Publisher
from rclpy.lifecycle import State
from rclpy.lifecycle import TransitionCallbackReturn
from rclpy.timer import Timer

import std_msgs.msg

from example_interfaces.srv import AddTwoInts


class ManagedNode(Node):

    def __init__(self, node_name, **kwargs):
        self._count: int = 0
        self._pub: Optional[Publisher] = None
        self._timer: Optional[Timer] = None
        self._cli = None
        super().__init__(node_name, **kwargs)

    def publish(self):
        msg = std_msgs.msg.String()
        msg.data = "Lifecycle HelloWorld #" + str(self._count);
        self._count += 1

        if self._pub is None or not self._pub.is_activated:
            self.get_logger().info('Lifecycle publisher is currently inactive. Messages are not published.');
        else:
            self.get_logger().info(f'Lifecycle publisher is active. Publishing: [{msg.data}]');

        if self._pub is not None:
            self._pub.publish(msg)

    def send_request(self, a, b):
        req = AddTwoInts.Request()
        req.a = a
        req.b = b
        future = self._cli.call_async(req)
        rclpy.spin_until_future_complete(self, future)
        print(f'result: {future}')

    def on_configure(self, state: State) -> TransitionCallbackReturn:
        self._pub = self.create_lifecycle_publisher(std_msgs.msg.String, "lifecycle_chatter", 10);
        self._timer = self.create_timer(1.0, self.publish)

        self._cli = self.create_client(AddTwoInts, 'add_two_ints')
        while not self._cli.wait_for_service(timeout_sec=1.0):
            self.get_logger().info('service not available, waiting again...')
        self.send_request(2, 3)

        self.get_logger().info("on_configure() is called.")
        return TransitionCallbackReturn.SUCCESS

    def on_cleanup(self, state: State) -> TransitionCallbackReturn:
        self.destroy_timer(self._timer)
        self.destroy_publisher(self._pub)

        self.get_logger().info('on_cleanup() is called.')
        return TransitionCallbackReturn.SUCCESS

    def on_shutdown(self, state: State) -> TransitionCallbackReturn:
        self.destroy_timer(self._timer)
        self.destroy_publisher(self._pub)

        self.get_logger().info('on_shutdown() is called.')
        return TransitionCallbackReturn.SUCCESS


def main():
    rclpy.init()

    executor = rclpy.executors.SingleThreadedExecutor()
    node = ManagedNode('managed_node')
    executor.add_node(node)
    try:
        executor.spin()
    except (KeyboardInterrupt, rclpy.executors.ExternalShutdownException):
        node.destroy_node()


if __name__ == '__main__':
    main()

Steps to reproduce:

1. Start the node: python3 managed_node.py
2. Start the service: ros2 run examples_rclcpp_minimal_service service_main
3. Perform transition on node: ros2 lifecycle set /managed_node configure
4. The node is stuck. If send_request() is commented out everything is working fine

Does anyone know if/how it's possible to make service calls in the early stages of a managed node?

Thanks! Thomas