Node crash monitoring

[ralwing]

Hi,

First, I want to commend you on the incredible work you've done for the future of transportation!

Now, to address the main question:
The ROS2 node containers have supported the on_exit action for quite some time, but it was only available in Python launch files. Fortunately, the Jazzy version will also support XML and YAML launch files. However, I haven't seen this feature used in autoware.universe, nor have I found an equivalent for it.

In our case, we monitor the launchers using the supervisor daemon, which restarts the launcher whenever it detects that the launcher has stopped. This allows us to operate our shuttle bus without a remote intervention.

Here are my questions:

1. Does this mean autoware doesn't monitor the node status?
2. Can you detect if a node crashes during the day? If so, what actions do you take in such cases?
3. Is there a future plan to support node crashes? Are You planning to use any ROS2 Node/Launch features for it? e.g. LifeCycleNodes

Thank you very much,
Grzegorz

[ralwing]

There is already a discussion regarding the node monitoring:
https://github.com/orgs/autowarefoundation/discussions/3194
And an open PR:
autowarefoundation/autoware.core#84