NodeUp: Hardening for Error Propagation and Post-Mortem Debugging

[ScottONeal]

After reading through this issue: nodejs/post-mortem#16
and trying to establish my own battle hardened methodology for handling uncaught errors at run time and debugging post-mortem. It would be nice to have a conversation on industry best practices for designing your application to provide the maximal amount of information for debugging. I think this should also include industry best practices for what to properly do for uncaughtException and unhandledRejection. Or, do I just always run with: --abort-on-uncaught-exception.

What are the best tools for understanding core files and how I can we bring this process out from the Node.js Ivory Tower?

Maybe this topic is a bit too technical for a podcast, but it would be nice to have a dialog about the state of production debugging and exception handling. What are the best practices and what is overkill.

[dshaw]

@ScottONeal Great topic. I have some producty answers to this topic. Beyond that work, the teams at Netflix, Joyent and Uber are doing some really interesting work in this space. Notably, these are all orgs that have either never gone Promises or have had to eliminate them to maintain observability. It would be nice to mix in a more positive Promises story too if it exists.

In the interim, folks definitely need to check out @StinkyDofu's To Err is Human talk: https://enterprisejs.io/events-sfbay002/ . So good!

[ScottONeal]

Thanks for the resource @dshaw! I have also noticed the traceability issues with promises and have tried to remove them, when possible.

Seems like Promises have introduced themselves almost everywhere in the ecosystem, while it's a nice abstraction, the costs/complexity are becoming a bit expensive when debugging!

[chrisdickinson]

I'd be happy to represent the promises use-case if this comes up!