@driffel/universal-google-logger

Universal Google Logger

⏩ Use the same library to log across GCE VMs, Firebase Functions, etc.

⏩ Makes sharing common code across environments easier.

Usage with default logger instance:

/**
 * Import method directly from default logger instance.
 *
 * Available methods:
 * alert, critical, debug, emergency, error, info, notice, warning
 *
 * Can also import default logger via the `logger` export.
 */
import {info} from "@driffel/universal-google-logger";

info("Text payload")
info({someField: "JSON payload"})

Usage with custom instance:

import {UniversalGoogleLogger} from "@driffel/universal-google-logger";


const customLogger = new UniversalGoogleLogger("Custom logger name");

customLogger.info("Text payload")
customLogger.info({someField: "JSON payload"})

Usage with trace-aware Express middleware:

import {attachContextWithTraceState} from "@driffel/universal-google-logger";

const app = express().use(attachContextWithTraceState({rethrowExceptions: false}));

Advisory:

This library is not fully tested for all possible environment permutations. It is an MVP (minimum viable product) that has been tested under standard GCE VMs, as well as Firebase Functions (v2) configured under a Cloud Run-based deployment. It should work in various other enviroments, but feedback as well as PR's are welcome (also feel free to inform if any configs, etc. are inaccurate; this is the culmination of a "best effort" in bringing together multiple info sources of varying age).

Problem Statement:

When logging in Firebase/Google Cloud, you have three options, but none of them fulfill all requirements of propagating the correct metadata and log severity levels across environments. This makes developing for shared code across environments difficult when it comes to logging.

• console.log / warn / error / etc...
  Propagates entries to Logs Explorer and sets the correct metadata fields for discovery using default filters from a function or VM, but does not properly embed the correct severity based on the console method called.

• @google-cloud/logging library
  Generally considered the more modern and preferred approach to logging. Works well with handling metadata and severity level in GCE VM environments, but does not populate most of the metadata when in Firebase Function environments, making log discovery difficult.

• firebase-functions/logger library
  Works well with all required metadata and appropriate severity levels, but only works for Firebase Functions.

Solution:

This library wraps the functionality of @google-cloud/logging library and fills the gaps in metadata wherever available from google-provided environent variables or, if needed, the metadata server. It also propagates the correct severity levels, since it leverages google's logging library directly.

There are also helper functions for including the trace metadata field which enables the "Show entries for this trace" feature that groups entries in Logs Explorer. This is done by leveraging AsyncLocalStorage to retain trace state through the call stack.

Environment variable configs (all optional):

• LOGGER_NAME: Used to set a name for the default logger instance. Default value: "node-app"
• PROJECT_STRING_ID: Used to populate the projectId parameter of the Logging service in @google-cloud/logging library. This parameter is often provided through built-in env vars, or as a last resort from the metadata server (if using a GCE VM environment).
• FUNCTION_REGION: Used to populate the region/location field in Firebase Function logs. Default value: "us-central1"
• LOG_TO_CONSOLE: Set this to "true" to route logs to console.log instead of using google logging library. Useful for localhost/dev logging.

Note regarding process.env values:

To ensure deterministic behavior, all environment variables are loaded only once, upon first import, to prevent potentially inconsistent behavior or configs due to something like a delayed dotenv config call.

Note regarding metadata server request:

A single synchronous API call, if necessary, is made in the very beginning upon file import. This is to load the project ID in the event that it's not provided by an expected environment variable (PROJECT_STRING_ID, GCLOUD_PROJECT, or GOOGLE_CLOUD_PROJECT).

Although blocking requests are typically frowned upon, this was considered the lesser of two evils compared to the increased complexity of guaranteeing the required field is loaded asynchronously before it's needed. It should not cause any perfomance issues, as it only blocks one time, in the very beginning upon file import. The call can be entirely avoided as well, by manually populating PROJECT_STRING_ID.