Add optional startup timeout while waiting for the initial federate connections#352
Add optional startup timeout while waiting for the initial federate connections#352NicoBiernat wants to merge 2 commits into
Conversation
| } | ||
| // This will release the critical section and allow other tasks to run. | ||
| self->env->wait_until(self->env, self->env->get_physical_time(self->env) + wait_before_retry); | ||
| self->env->wait_until(self->env, lf_time_add(self->env->get_physical_time(self->env), wait_before_retry)); |
|
Thanks for contributing! Before we accept this, we need to work through the exact semantics of your proposal. Startup and shutdown is quite subtle in distributed reactor programs. The current implementation in reactor-uc requires that all federates are running and in range of their neighbors to successfully start. We were discussing the possibility of having some "optional" federates that were not required for the startup procedure. First, could you explain a little more about your use case? What exactly should happen if the STARTUP_TIMEOUT expires? You wish to reset the MCU in question? Then what? What about the other federates? Why is a reset necessary? |
| #define TRANSIENT_WAIT_TIME MSEC(250) | ||
| #endif | ||
|
|
||
| #ifndef STARTUP_TIMEOUT |
There was a problem hiding this comment.
So you are suggesting solving this via a compile time flag instead of an annotation?
There was a problem hiding this comment.
Well, I just didn't know how to add a new annotation to lfc and get its value in reactor-uc 😄
But an annotation would probably be a cleaner solution.
I have the requirement that the program (federate) should never "hang" / infinitely loop. Some background on why: I want to benchmark federated reactor-uc on ~140 RP2040 + W5500 (Ethernet). This hardware is part of the light art installation Project Lighthouse (https://www.youtube.com/watch?v=VU26qkZiyl8) in our office building at Kiel University, so updating the firmware manually for every benchmark is quite infeasible. I know, this is kind of a niche use case, so I could just use my fork, if you don't need or want the feature. |
3 participants
Lingua Franca supports a global timeout property, after which the application is terminated.
However in federated LF applications, all federates have to initially connect to their "neighbors" and decide on a start tag. Only after that, the timeout starts.
But what if not all federates are available at the start?
Then the federates will wait (possibly forever) until the missing federate finally answers.
I really need to make sure that the application terminates, even in case of an error when some federates might be unavailable. In my case the termination ensures a reset of the MCU into the bootloader for multiple remote devices.
I added a
STARTUP_TIMEOUTwith a default value ofFOREVER, so the behavior is the same if you don't specify it.Also, I changed the
wait_untilcall to uself_time_addinstead of simple addition, which handles NEVER and FOREVER correctly.