log full stack traces on error

[dcrissman]

No description provided.

[dcrissman]

A stacktrace would be very useful for a bug I am tracking down.

@bserdar, @jewzaam, @paterczm - This is ready for review.

[jewzaam]

@dcrissman what is the impact of throwing a runtime exception? The concern I have is that it will get back to the client. We do not want to have a request fail because a hook fails, this is a bad user experience. It will be confusing. Does the hook processing ignore errors thrown by hook implementations?

[dcrissman]

The concern I have is that it will get back to the client.

The Mediator wraps all CRUD operations in a try/catch and handles accordingly. That said I should probably use Error.get instead.

We do not want to have a request fail because a hook fails, this is a bad user experience.

I would say it depends on the hook. In the case of the esb hook, if an event was missed I would think that would be pretty bad.

[jewzaam]

Clients don't care about integration implications. They just want their
data request to occur. If the hook is firing it means their data has been
persisted. Telling them a hook failed has no value to the client. It
should be logged and ideally monitored. I think of it in terms of a
relational database. If you have a trigger on a table, the client knows
nothing about it. If the trigger gets into an invalid / broken state
operations against that table fail. From an operations point of view, the
quickest fix is to disable the trigger if it can't be quickly fixed. We
have an opportunity with lightblue hooks to not impact the client when
something in a hook is broken. Ergo, don't throw unchecked exceptions.

On Wed, Sep 30, 2015 at 8:05 AM, Dennis Crissman notifications@github.com
wrote:

The concern I have is that it will get back to the client.

The Mediator wraps all CRUD operations in a try/catch and handles
accordingly. That said I should probably use Error.get instead.

We do not want to have a request fail because a hook fails, this is a bad
user experience.

I would say it depends on the hook. In the case of the esb hook, if an
event was missed I would think that would be pretty bad.

—
Reply to this email directly or view it on GitHub
#37 (comment)
.

[bserdar]

A more descriptive error message would be useful here. Same for the next one,

[bserdar]

Maybe we can use a different logger, and send the error logs to a separate logfile?

[dcrissman]

@bserdar, @jewzaam - Requested changes made.