How to use chrony for two ROS masters on one computer?

edit

Then what do I need to synchronize two ROS masters at one computer? I guess I need to synchronize the sim_time. The chrony won't help with that? What does?

That's something entirely different.

sim_time (I think you mean use_sim_time) is a ROS parameter. If you want all nodes in your two node graphs to use simulation time you'll have to set that parameter on both ROS masters. FKIE Multimaster doesn't sync parameters as far as I know, so you'll have to do that manually (or set it in a .launch file).

@gvdhoorn, I think I have that, Each of the launch file has use_sim_time set to True. When I use this command

$ rosparam get /use_sim_time
true

on both ROS masters I get true.

But inside the simulations each of the sim_times are different. The simulation run at different speed. I think that what gives me trouble and that is what I need to synchronize.

You'll need to make sure there is only a single Clock publisher in your entire network. Otherwise different nodes will use different time bases. That is not going to work.

I've not done this myself, but I assume you'll have to use MFKIE to sync the clock topic from one of the node graphs to the other.

Tbh, I'm not quite sure why someone would want to use two masters and then MFKIE on the same machine, but you've most likely already concluded that it's necessary.

if you used a single master, there would be no problem here afaict.

@gvdhoorn, I would gladly use only one master (although I'm not sure that would solve my problem). I am using two ROS masters, because I need to run two Gazebo simulations. Each of the Gazebo simulations communicate with ROS through a topic called /gazebo, so if two Gazebo's are running, the ROS cannot distinguish between them. And when I was trying to launch each Gazebo in a its own namespace, I couldn't get them to load correctly.

But anyway, I think that the Sim Time is dependent on the step size, RT update rate and the complexity of the simulation, so I would assume not even running two simulations in one ROS master would ensure synchronous time? I don't quite understand this.

Running two instances of Gazebo is going to introduce some complexities, yes. Especially when it comes to comparing time-stamped data between the two.

I'm not aware of any techniques that allow you to sync two instances. That would be something to ask over at Gazebo Answers.

Multiple instances in namespaces should work, but sometimes isn't straightforward.