How to detect error on startup?

[aseemk]

Hi @brianc,

Forky is great — we've been using it for two years now. Thank you!

One question: a couple of times now, we've accidentally shipped a bug where we hit a runtime error on startup. Forky sees the process crash and restarts it. But this just repeats ad infinitum.

Because the master Forky process itself doesn't fail, we don't detect any error immediately. The good news is that we do eventually detect a failure after a brief time threshold. (Heroku sees that we haven't bound to the network port after a minute ==> R10 error.)

Just curious: is there any way Forky could detect an error on startup, whether exactly or heuristically? E.g. an error <N seconds from spawn, or >N errors in <M seconds, or similar?

Thanks!

[brianc]

Yah we need to implement some kinda exponential back-off on startup. I just dusted this off & updated it for newer versions of node...so I might get to this soon.!