rmw_zenoh

A ROS 2 RMW implementation based on Zenoh that is written using the zenoh-c bindings.

Design

For information about the Design please visit design page.

Requirements

• ROS 2: Rolling/Jazzy/Iron

Setup

Build rmw_zenoh_cpp

Note: By default, we vendor and compile zenoh-c with a subset of zenoh features. The ZENOHC_CARGO_FLAGS CMake argument may be overwritten with other features included if required. See zenoh_c_vendor/CMakeLists.txt for more details.

mkdir ~/ws_rmw_zenoh/src -p && cd ~/ws_rmw_zenoh/src
git clone https://github.com/ros2/rmw_zenoh.git
cd ~/ws_rmw_zenoh
rosdep install --from-paths src --ignore-src --rosdistro <DISTRO> -y # replace <DISTRO> with ROS 2 distro of choice
source /opt/ros/<DISTRO>/setup.bash # replace <DISTRO> with ROS 2 distro of choice
colcon build --cmake-args -DCMAKE_BUILD_TYPE=Release

Test

Make sure to source the built workspace using the commands below prior to running any other commands.

cd ~/ws_rmw_zenoh
source install/setup.bash

Start the Zenoh router

Note: Manually launching Zenoh router won't be necessary in the future.

# terminal 1
ros2 run rmw_zenoh_cpp rmw_zenohd

Note: Without the Zenoh router, nodes will not be able to discover each other since multicast discovery is disabled by default in the node's session config. Instead, nodes will receive discovery information about other peers via the Zenoh router's gossip functionality. See more information on the session configs below.

Terminate ROS 2 daemon started with another RMW

pkill -9 -f ros && ros2 daemon stop

Without this step, ROS 2 CLI commands (e.g. ros2 node list) may not work properly since they would query ROS graph information from the ROS 2 daemon that may have been started with different a RMW.

Run the talker

# terminal 2
export RMW_IMPLEMENTATION=rmw_zenoh_cpp
ros2 run demo_nodes_cpp talker

Run the listener

# terminal 2
export RMW_IMPLEMENTATION=rmw_zenoh_cpp
ros2 run demo_nodes_cpp listener

The listener node should start receiving messages over the /chatter topic.

Configuration

By default, Zenoh sessions created by rmw_zenoh will attempt to connect to a Zenoh router to receive discovery information. To understand more about Zenoh routers and Zenoh sessions, see Zenoh documentation.

Checking for a Zenoh router.

The ZENOH_ROUTER_CHECK_ATTEMPTS environment variable can be used to configure if and how a Zenoh session checks for the presence of a Zenoh router. The behavior is explained in the table below.

ZENOH_ROUTER_CHECK_ATTEMPTS	Session behavior
unset or 0	Indefinitely waits for connection to a Zenoh router.
< 0	Skips Zenoh router check.
> 0	Attempts to connect to a Zenoh router in ZENOH_ROUTER_CHECK_ATTEMPTS attempts with 1 second wait between checks.

Session and Router configs

rmw_zenoh relies on separate configurations files to configure the Zenoh router and Zenoh session respectively. For more information on the topology of Zenoh adopted in rmw_zenoh, please see Design. Default configuration files are used by rmw_zenoh however certain environment variables may be set to provide absolute paths to custom configuration files. The table below summarizes the default files and the environment variables for the Zenoh router and Zenoh session. For a complete list of configurable parameters, see zenoh/DEFAULT_CONFIG.json5.

	Default config	Envar for custom config
Router	DEFAULT_RMW_ZENOH_ROUTER_CONFIG.json5	ZENOH_ROUTER_CONFIG_URI
Session	DEFAULT_RMW_ZENOH_SESSION_CONFIG.json5	ZENOH_SESSION_CONFIG_URI

For example, to set the path to a custom Zenoh router configuration file,

export ZENOH_ROUTER_CONFIG_URI=$HOME/MY_ZENOH_ROUTER_CONFIG.json5

Connecting multiple hosts

By default, all discovery traffic is local per host, where the host is the PC running a Zenoh router. To bridge communications across two hosts, the Zenoh router configuration for one the hosts must be updated to connect to the other Zenoh router at startup. This is done by specifying an endpoint in host's Zenoh router configuration file to as seen below. In this example, the Zenoh router will connect to the Zenoh router running on a second host with IP address 192.168.1.1 and port 7447.

{
  connect: {
    endpoints: ["tcp/192.168.1.1:7447"],
  },
}

Note: To connect multiple hosts, include the endpoints of all Zenoh routers in the network.

Logging

The core of Zenoh is implemented in Rust and uses a logging library that can be configured via a RUST_LOG environment variable. This variable can be configured independently for each Node and the Zenoh router. For instance:

• RUST_LOG=zenoh=info activates information logs about Zenoh initialization and the endpoints it's listening on.
• RUST_LOG=zenoh=info,zenoh_transport=debug adds some debug logs about the connectivity events in Zenoh.
• RUST_LOG=zenoh=info,zenoh::net::routing::queries=trace adds some trace logs for each query (i.e. calls to services and actions).
• RUST_LOG=zenoh=debug activates all the debug logs.

For more information on the RUST_LOG syntax, see https://docs.rs/env_logger/latest/env_logger/#enabling-logging.

Known Issues

Crash when program terminates

When a program terminates, global and static objects are destructed in the reverse order of their construction. The Thread Local Storage is one such entity which the tokio runtime in Zenoh uses. If the Zenoh session is closed after this entity is cleared, it causes a panic like seen below.

thread '<unnamed>' panicked at /rustc/aedd173a2c086e558c2b66d3743b344f977621a7/library/std/src/thread/local.rs:262:26:
cannot access a Thread Local Storage value during or after destruction: AccessError

This can happen with rmw_zenoh if the ROS 2 Context is not shutdown explicitly before the program terminates. In this scenario, the Context will be shutdown inside the Context's destructor which then closes the Zenoh session. Since the ordering of global/static objects is not defined, this often leads to the above panic.

The recommendation is to ensure the Context is shutdown before a program terminates. One way to ensure this is to call rclcpp::shutdown() when the program exits. Note that composable nodes should never call rclcpp::shutdown(), as the composable node container will automatically do this.

For more details, see #170.