rclcpp::exceptions::RCLError / rcl_wait() failed: error not set

[doisyg]

Bug report

Required Info:

• Operating System:
  • Ubuntu 22.04
• Installation type:
  • bin deb
• Version or commit hash:
  • last Humble sync
• DDS implementation:
  • Cyclone + localhost_only
• Client library (if applicable):
  • rclcpp

Don't know if it is rclcpp or something nav2 specific, already posted here: ros-navigation/navigation2#3097

I get an rclcpp::exceptions::RCLError / rcl_wait() failed: error not set error randomly during endurance testing on a real robot on amcl, controller_server and some other navigation node. All were working fine under galactic.
Struggling to understand the origin of it.

Jul 27 02:14:09 bash[813572]: [controller_server-66] terminate called after throwing an instance of 'rclcpp::exceptions::RCLError'
Jul 27 02:14:09 bash[813572]: [controller_server-66] what(): rcl_wait() failed: error not set
Jul 27 02:14:09 bash[813572]: [controller_server-66] Stack trace (most recent call last) in thread 816228:
Jul 27 02:14:09 bash[813572]: [controller_server-66] ros-navigation/navigation2#14 Object "", at 0xffffffffffffffff, in
Jul 27 02:14:09 bash[813572]: [controller_server-66] ros-navigation/navigation2#13 Object "/usr/lib/x86_64-linux-gnu/libc.so.6", at 0x7fc4eee369ff, in
Jul 27 02:14:10 bash[813572]: [controller_server-66] ros-navigation/navigation2#12 Object "/usr/lib/x86_64-linux-gnu/libc.so.6", at 0x7fc4eeda4b42, in
Jul 27 02:14:10 bash[813572]: [controller_server-66] ros-navigation/navigation2#11 Object "/usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.30", at 0x7fc4ef0342c2, in
Jul 27 02:14:10 bash[813572]: [controller_server-66] ros-navigation/navigation2#10 Object "/opt/ros/humble/lib/librclcpp.so", at 0x7fc4ef44dc80, in rclcpp::executors::SingleThreadedExecutor::spin()
Jul 27 02:14:10 bash[813572]: [controller_server-66] ros-navigation/navigation2#9 Object "/opt/ros/humble/lib/librclcpp.so", at 0x7fc4ef445972, in rclcpp::Executor::get_next_executable(rclcpp::AnyExecutable&, std::chrono::duration<long, std::ratio<1l, 1000000000l> >)
Jul 27 02:14:10 bash[813572]: [controller_server-66] ros-navigation/navigation2#8 Object "/opt/ros/humble/lib/librclcpp.so", at 0x7fc4ef4435cd, in rclcpp::Executor::wait_for_work(std::chrono::duration<long, std::ratio<1l, 1000000000l> >)
Jul 27 02:14:10 bash[813572]: [controller_server-66] ros-navigation/navigation2#7 Object "/opt/ros/humble/lib/librclcpp.so", at 0x7fc4ef43f9c8, in rclcpp::exceptions::throw_from_rcl_error(int, std::__cxx11::basic_string<char, std::char_traits, std::allocator > const&, rcutils_error_state_s const*, void (*)())
Jul 27 02:14:10 bash[813572]: [controller_server-66] ros-navigation/navigation2#6 Object "/usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.30", at 0x7fc4ef00627d, in std::rethrow_exception(std::__exception_ptr::exception_ptr)
Jul 27 02:14:10 bash[813572]: [controller_server-66] ros-navigation/navigation2#5 Object "/usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.30", at 0x7fc4ef0062f6, in std::terminate()
Jul 27 02:14:10 bash[813572]: [controller_server-66] ros-navigation/navigation2#4 Object "/usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.30", at 0x7fc4ef00628b, in
Jul 27 02:14:10 bash[813572]: [controller_server-66] ros-navigation/navigation2#3 Object "/usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.30", at 0x7fc4eeffabfd, in
Jul 27 02:14:10 bash[813572]: [controller_server-66] ros-navigation/navigation2#2 Object "/usr/lib/x86_64-linux-gnu/libc.so.6", at 0x7fc4eed387f2, in abort
Jul 27 02:14:10 bash[813572]: [controller_server-66] ros-navigation/navigation2#1 Object "/usr/lib/x86_64-linux-gnu/libc.so.6", at 0x7fc4eed52475, in raise
Jul 27 02:14:10 bash[813572]: [controller_server-66] #0 Object "/usr/lib/x86_64-linux-gnu/libc.so.6", at 0x7fc4eeda6a7c, in pthread_kill
Jul 27 02:14:10 bash[813572]: [controller_server-66] Aborted (Signal sent by tkill() 814467 1000)
Jul 27 02:14:10 bash[813572]: [ERROR] [controller_server-66]: process has died [pid 814467, exit code -6, cmd '/opt/wyca/elodie/nav2_controller/lib/nav2_controller/controller_server --ros-args --params-file /home/wyca/.wyca/nav2_params/controller_server_params.yaml --params-file /home/wyca/.wyca/nav2_params/local_costmap_params.yaml --params-file /home/wyca/.wyca/nav2_params/costmaps_plugins_params.yaml --params-file /tmp/launch_params_est2liq2 -r /cmd_vel:=/twist_mux_root/twist_nav'].

[liangyuwei]

Ran into a similar situation with the same exception thrown. Not sure why it happened, we simply use a workaround to spin with try-catch for RCLError.

[doisyg]

@liangyuwei thanks for confirming it.

Not sure why it happened, we simply use a workaround to spin with try-catch for RCLError.

Where did you put this try-catch ? In a fork of rclcpp ?

[fujitatomoya]

@liangyuwei do you also use rmw_cyclonedds?

@doisyg this could be related to rmw_wait as implementation, so it would be worth to try rmw_fastrtps to see if the same problem can be observed. (see https://github.com/ros2/rmw_fastrtps#getting-started)

Where did you put this try-catch ? In a fork of rclcpp?

i think user application code to catch the exception from spin.

[doisyg]

@doisyg this could be related to rmw_wait as implementation, so it would be worth to try rmw_fastrtps to see if the same problem can be observed. (see https://github.com/ros2/rmw_fastrtps#getting-started)

@fujitatomoya Currently cannot test with rmw_fastrtps as this related issue ros-navigation/navigation2#3014 is still causing our system to crash at startup.

[SteveMacenski]

rclcpp::Executor::wait_for_work does that give anyone some contextual ideas from the traceback?

[doisyg]

@arslogavitabrevis also reported the issue on the nav2 ticket. Can you confirm OS, ROS version and DDS used ?

[arslogavitabrevis]

We are using Cyclone DDS, Ubuntu 22.04 and Humble from debian binary. We also encontered this error during endurance tests on navigation2 nodes.

[iuhilnehc-ynos]

There are no reproduced steps, I am not sure if the following path is the error case:

rcl_wait -> rcl_timer_get_time_until_next_call -> rcutils_atomic_load_bool (updated 5 months ago) -> return RCL_RET_TIMER_CANCELED without setting a error state. (The log showed failed: error not set)

https://github.com/ros2/rcl/blob/191b13bc2eb8578e40799eb6b50c74b885e7066e/rcl/src/rcl/wait.c#L573-L576

      ret = rcl_timer_get_time_until_next_call(wait_set->timers[i], &timer_timeout);
      if (ret != RCL_RET_OK) {
        return ret;  // The rcl error state should already be set.
      }

It seems we should deal with RCL_RET_TIMER_CANCELED.

[ladianchad]

I had the same error in planner_server. I'm using fast DDS, ubuntu 22.04, and humble.

[planner_server-3] terminate called after throwing an instance of 'rclcpp::exceptions::RCLError'
[planner_server-3]   what():  rcl_wait() failed: error not set
[ERROR] [planner_server-3]: process has died [pid 13889, exit code -6, cmd '/opt/floatic/workspace/dependency_ws/install/nav2_planner/lib/nav2_planner/planner_server --ros-args -r __node:=planner_server -r __ns:=/navigation --params-file /tmp/tmp7i39them --params-file /tmp/launch_params_06b_heu1 -r tf:=/tf -r tf_static:=/tf_static -r cmd_vel:=/nav_vel -r odom:=/odom -r goal_pose:=/goal_pose'].
[INFO] [planner_server-3]: process started with pid [27679]
[planner_server-3] [INFO] [1661392904.274648775] [navigation.planner_server]: 
[planner_server-3]      planner_server lifecycle node launched. 
[planner_server-3]      Waiting on external lifecycle transitions to activate
[planner_server-3]      See https://design.ros2.org/articles/node_lifecycle.html for more information.
[planner_server-3] [INFO] [1661392904.279793060] [navigation.planner_server]: Creating
[planner_server-3] [INFO] [1661392904.282890912] [navigation.global_costmap.global_costmap]: 
[planner_server-3]      global_costmap lifecycle node launched. 
[planner_server-3]      Waiting on external lifecycle transitions to activate
[planner_server-3]      See https://design.ros2.org/articles/node_lifecycle.html for more information.

[ladianchad]

I requested pr to solve this problem at this ticket ros2/rcl#1003

[iuhilnehc-ynos]

@fujitatomoya cc: @ivanpauno

I believe that @ladianchad's PR can solve this issue, but I suspected that rolling back the ros2/rcl#963 is the correct way.
Adding https://github.com/ros2/rcl/blob/191b13bc2eb8578e40799eb6b50c74b885e7066e/rcl/src/rcl/timer.c#L318-L320 can not avoid calling rcl_clock_get_now with a canceled timer. If it could happen(there is a small time window to cancel the timer between the check and getting time), is it still reasonable to add the checking inside?

Even if using rcl_timer_get_time_until_next_call to get the time, it seems necessary to check whether the timer is canceled or not.

The later checking the flag, the more accurate to process the timer event.
I think that's probably why @wjwwood implemented the timer at the beginning.

[liangyuwei]

@liangyuwei thanks for confirming it.

Not sure why it happened, we simply use a workaround to spin with try-catch for RCLError.

Where did you put this try-catch ? In a fork of rclcpp ?

@doisyg Sorry for the late reply, we do this:

while (rclcpp::ok()) {
    try {
        executor.spin();
    } catch (rclcpp::exceptions::RCLError & e) {
        RCLCPP_ERROR_STREAM(
            rclcpp::get_logger("rclcpp"),
            "executor spin error: " << e.what());
    }
}