New Relic appender

[polymeris]

No description provided.

[polymeris]

Noticed :?err is not assigned unless the exception is passed as first parameter, e.g. (timbre/error "Error doing thing" exception) doesn't work. Is that correct/intended?

[ptaoussanis]

Is that correct/intended?

@polymeris Hi Camilo, yes - that's correct.

[ptaoussanis]

Hi there. Am not at all familiar with NewRelic. Just want to check that it's intentional to call "noticeError" even when there's no error?

[polymeris]

It's intentional. noticeError can take a string or an exception as first parameter.
Semantically, is arguable wether it makes sense, but to me there is an error if you call a timbre logging function, and it should be reported, even if there is no associated exception.

[polymeris]

NewRelic API for reference: http://newrelic.github.io/java-agent-api/javadoc/com/newrelic/api/agent/NewRelic.html

[ptaoussanis]

Thanks for the clarification.

So this NewRelic API doesn't have a notion of logging anything besides an error? The idea here is that this appender intentionally only logs errors?

(NewRelic/noticeError ?err params)

Is there a reason you don't want more of the data content for this call?

All you're be logging in this case is the exception and level. Not the namespace, logging message, args, etc.

Any reason you wouldn't want, for example, to add (output-fn data) to the params map here?

[polymeris]

So this NewRelic API doesn't have a notion of logging anything besides an error?

No other messages, as far as I can tell. It logs transactions (web, db, and others), but that's a different story.

All you're be logging in this case is the exception and level. Not the namespace, logging message, args, etc.

The stack trace (and thus the namespace) is logged.

Sure we could log more stuff. TBH, I haven't even figured out how to get New Relic to show those "params" you pass to noticeError. It mostly cares about: exception class, exception message, and the containing transaction (a function which you can designate as such, usually).

There is also the notion of "attributes". I might explore those in the future as a way to add more info to error logging, but for the time being I only wanted the basics.

[polymeris]

No problem. It's true that NR is not a logging facility.

That said, for me, it's just practical that at the very least error and fatal should go trough New Relic (and that's what I actually do in my project, not sure why I specified :warn as the default). Sure I could call noticeError directly, but if I already have the timbre hook, and the cases where I log an error (i.e. when the transaction or job failed) are exactly the same I would report to NR... why not use that?

[ptaoussanis]

the cases where I log an error (i.e. when the transaction or job failed) are exactly the same I would report to NR... why not use that?

Because you'd be complecting things that are semantically very different:

1. Reporting an API-specific error to a particular API.
2. General-purpose, multi-level logging to an arbitrary set of appenders.

So there's two questions here. First: is this the best approach to take in your specific case? I'd argue not, when you can just do this with a function that sends the notification and triggers a related Timbre logging call if you like. That way you're not tying certain logging levels into a specific semantic meaning. What if you call Timbre's error function somewhere that has nothing to do with this NewRelic library?

Second question: is this going to be widely useful to enough other developers to warrant inclusion in Timbre itself, ongoing maintenance, etc.? Do you see many other users wanting this particular behaviour?

Does that make sense / seem reasonable?

[polymeris]

Regarding the first question, I understand your point, but I don't see how I could not want all errors to be recorded by New Relic. Or conversly, if there were things I did not want sent to New Relic, I'd argue they are not errors, and I would be abusing timbre/error by logging them with it. (I'll need to think about this point, tho)
Slack isn't a logging app either, but I still want my errors to be sent there, too.

Regarding the second question... I don't know. If you are not convinced, maybe we can just close this PR. We can always reopen it later if the need for this appender becomes evident. Thank you for your patience.

[ptaoussanis]

but I don't see how I could not want all errors to be recorded by New Relic.

Okay, so it actually makes sense to send all application errors from any source to New Relic? It's a little difficult from https://newrelic.com/ to figure out what this software actually does. Looked to me like some kind of metrics library but I must be misunderstanding?

What happens with the info that you report with NewRelic/noticeError?

To be absolutely clear:

1. It makes sense to send every application-level error to NewRelic even if the error has nothing to do with NewRelic itself?
2. It makes sense to send only error-level information; there's no analog for other logging levels?

Thank you for your patience.

No problem, and likewise thank you.

[polymeris]

It makes sense to send every application-level error to NewRelic even if the error has nothing to do with NewRelic itself?

Yes, New Relic is a monitoring and analytics tool. They provide error analytics, among other things. So one can look at pretty graphs about which hosts suffer how many errors of this or that class at this or that time of the day, during this or that transaction/request/job.

It makes sense to send only error-level information; there's no analog for other logging levels?

As far as I can tell, there are no other log levels (I am no expert, we only started using this thing).

[ptaoussanis]

Okay, thanks for clearing the confusion. Will merge!