Logging documentation?

[evdevdev]

Hi there,

I've been working with @graphql-yoga/node and I cannot find any documentation around logging. There are details about how to log when connected to another server (i.e. Fastify). However, I don't see anything for @graphql-yoga/node itself. Am I missing something? I also tried pouring over old issues, and I don't see anything there.

I did find this dicussion for helping with dev logging. However, I am looking for some more robust guidance, specifically around how to prep for production logging.

Since I'm creating a discussion anyone, perhaps it'd be helpful if I included my next task: Is there a way to set up logs such that Prisma logging and yoga logging all go through the same pipes?

Thanks in advance for any guidance! I'm loving this library!

[ardatan]

Hi!
We are aware of that missing part in our documentation.
This is part of our roadmap;
#1195

Today in v2, logging parameter accepts any logger implementation that supports different levels as method like debug, error, warn and info.
Also Yoga provides a very basic logger implementation by default which you can use together with other libraries;
https://github.com/dotansimha/graphql-yoga/blob/master/packages/common/src/logger.ts

We'd love to help you if you can explain more what you want to achieve :)

Thanks!

[evdevdev]

Thanks for following up! This is helpful.

My dream default state would be something that could elegantly pipe into DataDog (or some such log tool) that I could use to build performance dashboards, i.e. "how's my server doing? Is it going fast or slow? Where are the bottle necks? etc."

I think a good example of this comes from the Rails ecosystem in the package called lograge. Here's an example log statement from them:

method=GET path=/jobs/833552.json format=json controller=JobsController  action=show status=200 duration=58.33 view=40.43 db=15.26

The key attributes there are that you can easily parse that in order to graph things like duration. You can also do things like grouping duration with your controller in order to identify performance bottlenecks. I could see a graphql doing something similar with a breakdown by query name (for example).

Another decent example comes from Redwood. It's pretty minimal, but it does have times and queries out of the box.

[dthyresson]

Another decent example comes from Redwood. It's pretty minimal, but it does have times and queries out of the box.

Hi @earnold - I implemented logging in RedwoodJS and fundamentally it is the Pino logger and then uses an envelope/yoga plugin to do provide some useful helpers (like timing etc). You can find that in useRedwoodLogger and then the use of Pino in the logger.

It's Pino that can send the logs to Datadog or other places via its transports/destinations and there are packages for popular ones.

That said, logging may not be all you want here -- but more telemetry or event-based output like the Sentry or telemetry plugins in envelop https://www.envelop.dev/plugins/use-open-telemetry ... which you may consider instead/

These work in the lifecycle of the request rather than explicit log statements (with levels etc).