core: Log database + reducer name on error

[kim]

In addition to the error message, log the database identity and the
reducer name when a reducer returns an error.

This provides a useful correlation, for example when the error has
upstream effects such as causing elevated error rates in the proxy
connection establishment due to the module rejecting the connection on
purpose.

Expected complexity level and risk

1

Testing

Logging change only.

[bfops]

this seems fine to me, although I think this is one of the logs that we wanted to remove generally since it generally indicates user error, not server error

[kim]

it generally indicates user error

We could omit the error message, maybe?
There's just not much we can do on the proxy side, because the error variants are all private.

[bfops]

it generally indicates user error

We could omit the error message, maybe? There's just not much we can do on the proxy side, because the error variants are all private.

I think the general problem is that error logs should be for actual errors, but a user's mistake isn't really a programmer or runtime error. Our system is usually working perfectly fine when the so-called errors happen.

Maybe we could downgrade it to debug? Either way, we haven't changed it yet, and this extra info seems generally useful.

[kim]

Well ok, maybe something else is going on: we do have a log for when the module rejects the connection, but that doesn't appear in the production logs. In fact, nothing obviously related appears.