Bunyan GKE Stackdriver Transport

Simple bunyan transport to leverage GKE (Google Kubernetes Engine) logging configuration for Stackdriver

Usage

Just configure this transport for the logger:

const Bunyan = require('bunyan');
const {createStream} = require('bunyan-gke-stackdriver');

const logger = Bunyan.createLogger({
  name: 'MyLogName',
  streams: [createStream()],
});

Optionally, you can configure the log level for the transport and the final output stream (stdout by default).

const Bunyan = require('bunyan');
const {createStream} = require('bunyan-gke-stackdriver');

const logger = Bunyan.createLogger({
  name: 'warnings',
  streams: [createStream(Bunyan.WARN, process.stderr)],
});

The Problem

If your node application is deployed on GKE, and you've enabled logging for it (the default); all your logs will end up in Stackdriver viewer but won't be very readable, and won't be using most of Stackdriver features.

Bunyan uses structured logging, but with a different schema than stackdriver expect to make take advantange of it's features.

Some of the features enabled by using this module:

Show the correct severity (level) in stackdriver
Show the msg in the summary. (Stackdriver expects a message key instead of msg)
Render information about the request.
Track errors in Stackdriver Error Reporting

How does this work?

When using bunyan you can configure transports (streams in reality). If no one is configured, logs go to stdout. In our case, we want them to go to stdout but with a different format. A format that is compatible with what GKE has configured for log recolection.

So, in essence all you need to do is configure this module as the stream.

const Bunyan = require('bunyan');
const {createStream} = require('bunyan-gke-stackdriver');

const logger = Bunyan.createLogger({
  name: 'MyLogName',
  streams: [createStream()],
});

The lifecycle of a log entry in GKE

So, what happens when you do logger.info('hello world')?

First, bunyan creates a log record something like {v: 1, level: 30, msg: 'hello world', ...}. Then, is passed to the configured strem, which if none was configured, is simply sending all to stdout

Second, since you are running on a docker container within a node in GKE; docker will wrap that log entry into something like

{
  "stream": "stdout",
  "time": "2018-08-24T12:41:50.987184687Z",
  "log": "{\"level\":30,\"time\":1535114510986,\"msg\":\"hello world\"...}"
}

Lastly, since you have logging configured for GKE, there a fluentd daemon on each node. fluentd aggregates logs from all your containers, transform them based on it's configuration, and finally exports them. On GKE, fluentd is configured to export log entries to stackdriver. Also, it's configured to unwrap the docker log entry and parse your original entry. It will recognize some fields as part of the stackdriver schema and use them. But mostly it will pass all of them to stackdriver.

The expected log entry

So, to be a good citizen with stackdriver and the fluentd configuration in GKE, we need to make a few changes to the original entry.

Use message instead of msg
When logging an error, use err.stack as the value for message
When logging an http request, use httpRequest to log the details about it
Map level to severity

For more information about the stackdriver log entry schema check:

Log Trasformation

When in bunyan you log like:

logger.info({data: {from: 'me', to: 'you'}}, 'data transfer');

Bunyan will generate

{
  "name": "src-example", // the name configured for the log
  "level": 30, // info is level:30
  "msg": "data transfer", // your message
  "data": {"from": "me", "to": "you"}, // your log data
  "time": "2012-02-06T04:19:35.605Z",
  "src": {
    // only if use src: true while creating the logger
    "file": "/Users/trentm/tm/node-bunyan/examples/src.js",
    "line": 20,
    "func": "Wuzzle.woos"
  },
  "hostname": "banana.local",
  "pid": 123,
  "v": 0
}

To make stackdriver-fluentd compatible it will be transformed into:

{
  "name": "src-example",
  "severity": 200, // map level into stackdriver severity
  "message": "data transfer", // change msg to message
  "data": {"from": "me", "to": "you"}, // your log data
  "time": "2012-02-06T04:19:35.605Z",
  "src": {
    // only if use src: true while creating the logger
    "file": "/Users/trentm/tm/node-bunyan/examples/src.js",
    "line": 20,
    "func": "Wuzzle.woos"
  }
  // erase not important keys v, pid, hostname
}

For Errors

when logging an error with bunyan:

log.info(err); // Special case to log an `Error` instance to the record.
// This adds an "err" field with exception details
// (including the stack) and sets "msg" to the exception
// message.
log.info(err, 'more on this: %s', more);
// ... or you can specify the "msg".

Bunyan by default, will keep your error in err key. We will use the err.stack and set it as the message, since that what's required by stackdriver.

Also, the error log entry for stackdriver requires us to set:

  "serviceContext": {
    "service": string,     // Required.
    "version": string
  },

We will use the name configured as serviceContext.name.

Thanks to this, you can track your error ocurrences in https://cloud.google.com/error-reporting/

Other options

You can bypass fluentd & docker all together and log directly by calling the stackdriver API. I'm not pro using a http transport for log within the applicaton, since every application will be doing it's own buffering and sending, plus you will be loosing context information like the node & pod id.

But, if that's not a proble, simply use https://github.com/googleapis/nodejs-logging-bunyan.

mutru/bunyan-gke-stackdriver