Enable discovery after launch (perhaps as an action or service call?)

[andrewbest-tri]

Description

When running longer experiments, it may be the case that topics start or stop publishing at different times. When collecting a ros bag, it would be extremely helpful for us to be able to trigger discovery on a running bagger so it can pick up new topics. We've tried pre-selecting the topics that may appear. It is prone to human error and may create issues with QOS. We'd like to be able to do this in a programmatic fashion.

Related Issues

Completion Criteria

Implementation Notes / Suggestions

I suggest using a service call, action call, or possibly adding a flag to the bag verb that would allow a discovery interval.

We could potentially emit new topics the bagger has found or possibly register it for messages related to new topics (assuming those exist).

Testing Notes / Suggestions

[MichaelOrlov]

@andrewbest-tri Starting recording with ros2 bag record -a will not solve your problem?
i.e. running discovery continuously.

[andrewbest-tri]

I am experimenting again with this. In the past I have had missing data with -a, however I also had significant problems using fastRTPS configs. I may have incorrectly diagnosed the problem!

[andrewbest-tri]

Ah, yes this is a misdiagnosed problem. I am able to test this with a small dataset. The larger issue relates to #971 and the fact that i have been running without -a so i dont accidentally grab image topics I don't want.

Running with -a works in that I can see it subscribing to new topics (now that I have gotten rid of my fastrtps issues).
Running without -a and a list of topics (some of which are not yet published) also works. It is confusing because once all topics are detected the bagger will print

 All requested topics are subscribed. Stopping discovery...

which I had been incorrectly understanding as printed when using -a as well. So, this ticket is perhaps a non issue. The real issue will be 971, topic rejection and selection by regex. This would allow me to exclude the image topics I know will inflate my bag size while still catching all other relevant experimental data.

[MichaelOrlov]

@andrewbest-tri Thnak you for the clarification. I am closing this issue then.