cls-request-context
This library is based on cls-rtracer and makes it easier to add custom CLS properties
for the request context.
How to use it - Step 1
Install:
npm install --save cls-request-context cls-hooked
Note: cls-hooked
has to be installed explicitly, as it's a peer dependency for this library. See this issue for more details.
Note for TypeScript users: typings are included.
How to use it - Step 2 (Common instructions)
Use the middleware (or plugin) provided by the library before the first middleware that needs to have access to request ids. Note that some middlewares, e.g. express-session, body-parser, or express-jwt, may cause CLS context (i.e. Async Hooks execution path) to get lost. To avoid such issues, you should use any third party middleware that does not need access to request ids before you use this middleware. See issue #20 as an example.
How to use it - Step 2 (Express users)
Use the middleware provided by the library:
const express = require('express')
const rTracer = require('cls-request-context')
const app = express()
app.use(rTracer.expressMiddleware())
Obtain request id in middlewares on the incoming request:
app.get('/api/v1/entity/{id}', (req, res, next) => {
entityService.find(req.params.id)
.then((entity) => {
const requestId = rTracer.id()
console.log(`requestId: ${requestId}`)
res.json(entity)
})
.catch(next)
})
You can access the same request id from code that does not have access to the Express' req
object.
async function find (entityId) {
const requestId = rTracer.id()
}
How to use it - Step 2 (Fastify users)
Use the middleware provided by the library:
const fastify = require('fastify')()
const rTracer = require('cls-request-context')
fastify.use(rTracer.fastifyMiddleware())
Obtain request id in middlewares on the incoming request:
app.get('/test', async (request, reply) => {
const entity = await entityService.find(request.params.id)
const requestId = rTracer.id()
console.log(`requestId: ${requestId}`)
reply.send(entity)
})
You can access the same request id from code that does not have access to the Fastify's request
object.
async function find (entityId) {
const requestId = rTracer.id()
}
How to use it - Step 2 (Koa users)
Use the middleware provided by the library:
const Koa = require('koa')
const rTracer = require('cls-request-context')
const app = new Koa()
app.use(rTracer.koaMiddleware())
Obtain request id in middlewares on the incoming request:
app.use(async (ctx) => {
const entity = await entityService.find(req.params.id)
const requestId = rTracer.id()
console.log(`requestId: ${requestId}`)
ctx.body = entity
})
You can access the same request id from code that does not have access to the Koa's ctx
object.
async function find (entityId) {
const requestId = rTracer.id()
}
Koa v1 support
For Koa v1 use the koaV1Middleware(options)
function.
How to use it - Step 2 (Hapi users)
Use the plugin provided by the library:
const Hapi = require('@hapi/hapi')
const rTracer = require('cls-request-context')
const init = async () => {
const server = Hapi.server({
port: 3000,
host: 'localhost'
})
await server.register({
plugin: rtracer.hapiPlugin
})
}
init()
Obtain request id in route handlers on the incoming request:
server.route({
method: 'GET',
path: '/test',
handler: async (request, h) => {
const entity = await entityService.find(request.params.id)
const requestId = rTracer.id()
console.log(`requestId: ${requestId}`)
return entity
}
})
You can access the same request id from code that does not have access to the Hapi's request
object.
async function find (entityId) {
const requestId = rTracer.id()
}
Integration with loggers
The main use case for this library is request id generation and logging automation. You can integrate with any logger library in a single place and get request ids in logs across your application.
Without having request id, as a correlation value, in your logs, you will not be able to determine which log entries belong to code that handles the same request. You could generate request ids manually and store them in the Express' req
object (or Fastify's request
, or Koa's ctx
), but then you will have to explicitly pass the object into all other modules on the route. And that's when cls-request-context
comes to the rescue!
Here is how you can integrate cls-rtracer
with winston, one of most popular logging libraries.
const { createLogger, format, transports } = require('winston')
const { combine, timestamp, printf } = format
const rTracerFormat = printf((info) => {
const rid = rTracer.id()
return rid
? `${info.timestamp} [request-id:${rid}]: ${info.message}`
: `${info.timestamp}: ${info.message}`
})
const logger = createLogger({
format: combine(
timestamp(),
rTracerFormat
),
transports: [new transports.Console()]
})
Complete samples for Express, Fastify and Koa are available in /samples/
directory.
Middleware configuration
These are the available config options for the middleware functions. All config entries are optional.
{
useHeader: false,
headerName: 'X-Request-Id'
}
Troubleshooting
To avoid weird behavior:
-
Make sure you require cls-request-context
as the first dependency in your app. Some popular packages may use async which breaks CLS.
-
Make sure you use any third party middleware (or plugin) that does not need access to request ids before you use cls-request-context
. See this section.
Note: there is a small chance that you are using one of rare libraries that do not play nice with Async Hooks API, which is internally used by the cls-hooked
library. So, if you face the issue when CLS context (and thus, the request id) is lost at some point of async calls chain, please submit GitHub issue with a detailed description.
Note for Node 10 users:
- Node 10.0.x-10.3.x is not supported. That's because V8 version 6.6 introduced a bug that breaks async_hooks during async/await. Node 10.4.x uses V8 v6.7 where the bug is fixed. See: https://github.com/nodejs/node/issues/20274.
Performance impact
Note that this library has a certain performance impact on your application due to CLS (or more precisely, Async Hooks API) usage. So, you need to decide if the benefit of being able to trace requests in logs without any boilerplate is more valuable for you than the disadvantage of performance impact.
The author of this library did some basic performance testing and got about 10–15% RPS (request per second) degradation when cls-request-context
is used. See this post for more details.
License
Licensed under MIT.