EuRoC Forum

[proprentnerl]

Hello again,

as I have written in another topic ros cannot handle time_jumps very well, especially tf.
So after solving a task, it is necessary to kill all nodes using tf frames which brings up the problem that there seems to be no sure kill method for all nodes except restarting the roscore (which is not possible in the final evaluation).
I tried to simply relaunch my launch file, which starts all nodes a new but can lead to conditions where a node subscribes to an old node which than gets killed.

Any help would be appreciated!
Lukas

[Peter Lehner]

Hi,

I tried to simply relaunch my launch file, which starts all nodes a new but can lead to conditions where a node subscribes to an old node which than gets killed.

The problem I see is that the old ros nodes are not killed before you start your new nodes. Before relaunching your nodes you should probably check which of your old nodes are still running and kill them manually. You can use the 'rosnode' tool to do so:

Code: Select all

rosnode list

will list all current running nodes.

Code: Select all

rosnode kill YOUR_ROS_NODE

will kill the node with the name 'YOUR_ROS_NODE' and block until its shutdown.

Be careful though not to accidentally kill any of the simulator nodes.

Best,
Peter

[proprentnerl]

rosnode kill is sadly not very reliable...
there is simply no reliable way to reset everything in ros without restarting the roscore