microros reconnection failed.

[chrisjaxy]

Describe the bug
microros reconnection failed.

To Reproduce
Steps to reproduce the behaviour:
I found that when the microros connection fails or times out, it cannot be reconnected again. I tested pub, sub, client, services and added the log.








I found that the reason for the failure is the function get_memory. For failure, I made the following modifications to the code to make it reconnect successfully.










My code can refer to the demo program given by this link micro-ROS/micro_ros_setup#434. I created 26 pubs, 22 subs, 4 services and 5 clients. The method of reproduction is to restart the microros during the process of rclc_service_init_best_effort, rclc_client_init_best_effort, rclc_publisher_init_best_effort, rclc_subscription_init_best_effort.
The configuration of my colcon meta is as follows:

{
    "names": {
        "tracetools": {
            "cmake-args": [
                "-DTRACETOOLS_DISABLED=ON",
                "-DTRACETOOLS_STATUS_CHECKING_TOOL=OFF"
            ]
        },
        "rosidl_typesupport": {
            "cmake-args": [
                "-DROSIDL_TYPESUPPORT_SINGLE_TYPESUPPORT=ON"
            ]
        },
        "rcl": {
            "cmake-args": [
                "-DBUILD_TESTING=OFF",
                "-DRCL_COMMAND_LINE_ENABLED=OFF",
                "-DRCL_LOGGING_ENABLED=OFF"
            ]
        }, 
        "rcutils": {
            "cmake-args": [
                "-DENABLE_TESTING=OFF",
                "-DRCUTILS_NO_FILESYSTEM=ON",
                "-DRCUTILS_NO_THREAD_SUPPORT=ON",
                "-DRCUTILS_NO_64_ATOMIC=ON",
                "-DRCUTILS_AVOID_DYNAMIC_ALLOCATION=ON"
            ]
        },
        "microxrcedds_client": {
            "cmake-args": [
                "-DUCLIENT_PIC=OFF",
                "-DUCLIENT_PROFILE_UDP=OFF",
                "-DUCLIENT_PROFILE_TCP=OFF",
                "-DUCLIENT_PROFILE_DISCOVERY=OFF",
                "-DUCLIENT_PROFILE_SERIAL=OFF",
                "-UCLIENT_PROFILE_STREAM_FRAMING=ON",
                "-DUCLIENT_PROFILE_MULTITHREAD=ON",
                "-DUCLIENT_PROFILE_CUSTOM_TRANSPORT=ON"
            ]
        },
        "rmw_microxrcedds": {
            "cmake-args": [
                "-DRMW_UXRCE_MAX_NODES=1",
                "-DRMW_UXRCE_MAX_PUBLISHERS=30",
                "-DRMW_UXRCE_MAX_SUBSCRIPTIONS=25",
                "-DRMW_UXRCE_MAX_SERVICES=10",
                "-DRMW_UXRCE_MAX_CLIENTS=5",
                "-DRMW_UXRCE_MAX_HISTORY=2",
                "-DRMW_UXRCE_TRANSPORT=custom",
                "-DRMW_UXRCE_ENTITY_DESTROY_TIMEOUT=0",
                "-DRMW_UXRCE_TOPIC_NAME_MAX_LENGTH=100",
            ]
        }
    }
}

If you have successfully reproduced here and have better modifications, you can notify or tell me.

Expected behaviour
After microros reconnection fails or times out, it can still reconnect successfully.

System information (please complete the following information):

• OS: Ubuntu 20.04, freertos
• ROS 2 foxy
• Version lastest

[pablogs9]

Are you destroying your entities when the disconnect is detected? Could you share your micro-ROS app code?

Check this: https://github.com/micro-ROS/micro_ros_arduino/blob/39107e4c81affd714db98d7b63655b843905b99a/examples/micro-ros_reconnection_example/micro-ros_reconnection_example.ino#L88-L92

[chrisjaxy]

Are you destroying your entities when the disconnect is detected? Could you share your micro-ROS app code?

Check this: https://github.com/micro-ROS/micro_ros_arduino/blob/39107e4c81affd714db98d7b63655b843905b99a/examples/micro-ros_reconnection_example/micro-ros_reconnection_example.ino#L88-L92

When the connection fails, I destroy the entities.I refer to the previous program on micro-ROS/micro_ros_setup#434

rclc_support_t test_support;
rcl_node_t test_node;
rclc_executor_t test_executor;
rcl_allocator_t test_allocator;
rcl_publisher_t test_publisher;
rcl_subscription_t test_subscription;

std_msgs__msg__UInt8 pile_heartbeat;
sensor_msgs__msg__BatteryState battery;
bool micro_ros_init_successful = false;
bool micro_ros_destory_flag = false;

void battery_info(const void * msgin)
{
	const sensor_msgs__msg__BatteryState * msg = (const sensor_msgs__msg__BatteryState *)msgin;

	LOG_DEBUG("battert current : %1f ,percentage : %3f, status: %d", msg->current, msg->percentage, msg->power_supply_status);
}

bool create_entities()
{	
	test_allocator = rcl_get_default_allocator();

	// create init_options
	RCCHECK(rclc_support_init(&test_support, 0, NULL, &test_allocator));

	// create node
	RCCHECK(rclc_node_init_default(&test_node, "test", "", &test_support));

	// create publisher
	RCCHECK(rclc_publisher_init_default(
		&test_publisher,
		&test_node,
		ROSIDL_GET_MSG_TYPE_SUPPORT(std_msgs, msg, UInt8),
		"std_msgs_msg_Int8"));

	RCCHECK(rclc_subscription_init_best_effort(
		&test_subscription,
		&test_node,
		ROSIDL_GET_MSG_TYPE_SUPPORT(sensor_msgs, msg, BatteryState),
		"/battery"));

	// create executor
	test_executor = rclc_executor_get_zero_initialized_executor();
	RCCHECK(rclc_executor_init(&test_executor, &test_support.context, 1, &test_allocator));
	RCCHECK(rclc_executor_add_subscription(&test_executor, &test_subscription, &battery, battery_info, ON_NEW_DATA));

	micro_ros_init_successful = true;

	return true;
}

void destroy_entities()
{
	rcl_subscription_fini(&test_subscription, &test_node);
	rcl_publisher_fini(&test_publisher, &test_node);
	rcl_node_fini(&test_node);
	rclc_executor_fini(&test_executor);
	rclc_support_fini(&test_support);

	micro_ros_init_successful = false;
}

void StartDefaultTask(void *argument)
{
	bool pingFlag = false;
	MicroRosChannelConfig();

	while (1)
	{
		osDelay(5);
                if ((micro_ros_init_successful == false) && !pingFlag)
		{
			if (rmw_uros_ping_agent(50, 1) != RMW_RET_OK)
		 	{
				LOG_DEBUG("ping error!");
				osDelay(200);
				continue;
			}

			LOG_DEBUG("ping ok!");
			pingFlag = true;
		}

		if (!micro_ros_init_successful) {
			LOG_DEBUG("start create_entities!");
			if (create_entities() == false) {
				micro_ros_destory_flag = true;
				LOG_DEBUG("create_entities fail, destroy_entities!");
			} else {
				LOG_DEBUG("create_entities finish");
			}
		} else {
			rclc_executor_spin_some(&test_executor, RCL_MS_TO_NS(1));
		}

		if ((micro_ros_init_successful == true) && (micro_ros_destory_flag == true)){
			LOG_DEBUG("destroy_entities!");
			destroy_entities();
			micro_ros_destory_flag = false;
		}
        }
}

[pablogs9]

Make sure that you are destroying entities correctly because this get_memory() failure is related to this entity's lack of destruction.

Here you have an approach using a state machine: https://github.com/micro-ROS/micro_ros_arduino/blob/a7fd2280324e9344180512b41d0cfab1cc072bff/examples/micro-ros_reconnection_example/micro-ros_reconnection_example.ino#L104-L127

[chrisjaxy]

Make sure that you are destroying entities correctly because this get_memory() failure is related to this entity's lack of destruction.

Here you have an approach using a state machine: https://github.com/micro-ROS/micro_ros_arduino/blob/a7fd2280324e9344180512b41d0cfab1cc072bff/examples/micro-ros_reconnection_example/micro-ros_reconnection_example.ino#L104-L127

I'm sure that I am destroying entities correctly. When I have destroyed entities, I check the size of xAvailableHeapSpaceInBytes in freertos, it's exactly the same as before connection. You can try to reproduce this problem. The method of reproduction is to restart the microros agent during the process of rclc_service_init_best_effort, rclc_client_init_best_effort, rclc_publisher_init_best_effort, rclc_subscription_init_best_effort.

[pablogs9]

We will take a look

[chrisjaxy]

We will take a look

If you succeed or fail to reproduce this problem, please let me know.

[pablogs9]

I have not been able to replicate this. I see that you show in your original message a bunch of images with changes in the rmw_microxrcedds. Is it possible to create a PR with those changes so we can analyze them?

[chrisjaxy]

I have not been able to replicate this. I see that you show in your original message a bunch of images with changes in the rmw_microxrcedds. Is it possible to create a PR with those changes so we can analyze them?

This is the modification we tried. https://github.com/micro-ROS/rmw_microxrcedds/pull/265. We created a number of pub, sub, clent and services, and then restarted the agent to reproduce the connection process. The connection process is fast, making it difficult to restart the agent just in the middle of the connection process. I think it may be easy to reproduce if the baud rate of the communication is reduced.

[bjsowa]

@pablogs9 I believe I'm facing the same issue. My code uses the state machine approach very similar to the example you provided. If the agent successfully connects and later disconnects, all of the entities are correctly destroyed. But, if the agent disconnects during entity creation, some of the entities are not correctly destroyed, leaving allocated items in the mempool.

I applied #265 from @chrisjaxy to humble branch which I'm currently using and that seems to fix the issue for me. Please consider applying these changes.

[Acuadros95]

This should be fixed with #280.

Closing, reopen if the issue persist.