[RFC] lava_scheduler_app: Better handle device configuration errors
This is prototype, RFC changes to address issues reported in #430 . Fairly speaking, it may be possible to reject that issue as happening due to system misconfiguration. But that misconfiguration comes from the official LAVA Docker setup, with the misconfiguration being known issue for half a year: lava/pkg/docker-compose#4 (closed) , and over this time, it was shipped like that, and it's known that many parties adopt that configuration, so they may be affected.
I also find it very frustrating to debug issues like that, when their reporting is suppressed, so was anxious to try how hard it would be to improve the situation, so coded this prototype, which IMHO solves the core concern (no detailed reporting of errors, and as you try to unwind the situation, you see that explicit steps were taken to suppress propagation of errors, on multiple levels).
That said, this is indeed just a prototype, a "seed" for the core team to consider, and work from, as it definitely needs more thorough handling. There's of course a risk of starting to throw exceptions from where they were previously suppressed, as, if not handled properly, that may lead to fatal system errors. Again, I personally consider it fully worth to handle exception properly.
Individual commits also has detailed descriptions of various specific points.