Why might ros1_bridge stop forwarding lidar PointCloud2 messages?

Question

I have a large ROS network comprised of both ROS1 and ROS2 nodes. The nodes are bridged by the ros1_bridge dynamic_bridge.

Two of the ROS2 nodes publish VLP-32C lidar data as sensor_msgs/PointCloud2 onto two different lidar topics at ~10Hz. When the system starts up the ros1_bridge will correctly identify these topics and start forwarding these messages. However, after a few seconds at least one of the lidar streams will stop bridging and that lidar's data will no longer be available on the ROS1 network. It appears the ros1_bridge is still connected on both ends, but no lidar data is coming out. The ROS2 lidar data is still publishing and other messages are still being bridged.

I'm a bit at a loss for what might cause this. Does anyone have any suggestions on how to investigate.

I'm using Ubuntu 20.04 with ROS2 Foxy, ROS1 Noetic, and Cyclone DDS.

---

Originally posted by msmcconnell on ROS Answers with karma: 268 on 2022-03-22

Post score: 0

Answer 1

Seems this was caused by some known limitations in the way Cyclone DDS uses the kernal network. I followed the suggestions here to resolve it https://docs.ros.org/en/rolling/How-To-Guides/DDS-tuning.html#cyclone-dds-tuning

Specifically the command

sudo sysctl -w net.core.rmem_max=2147483647

---

Originally posted by msmcconnell with karma: 268 on 2022-03-23

This answer was ACCEPTED on the original site

Post score: 0