CPU leakage

[Netta Doron]

Hi, I've started using this library on our production servers. It's a great help and saves a lot of "passing down" from one function to the next. However, over time I see that it increasingly uses more and more CPU resources, specifically around things related to async-hooks. I took a CPU profile of the server and I see emitHook going from 7% to 30%. Inside emitHook, the processes that grow the most are destroy (1% to 8%), before grows from 2% to ~6%, after (~1% to 6$). Is this expected. Am I using it wrong? Please advise.
Below is my code snippet.

  const _NAMESPACE = 'FOO_LOGGING'; 
  const _LOGGER_CONTEXT = 'logger'; 

 class LoggingContext { 
 constructor(token, instance, level = 'debug') {
      this._loggerBuilder =
         aLogger()
           .withLogentriesToken(token)
           .withLevel(level);

      this._instance = instance;
   }

   runWithContext(aspects, fn, ...args) {
     try {
        const userSession = aspects['session'];
        const requestId = AspectsApi.getRequestId(aspects) || uuid();
        const context = {
            instance: this._instance || 'none',
            requestId,
            userId: userSession && userSession.userGuid
       };
         const logger = this._loggerBuilder.withContext(context);
          const session = createNamespace(_NAMESPACE);
          return session.runAndReturn(() => {
             session.set(_LOGGER_CONTEXT, logger);
             return fn(...args);
         });
         } catch (e) {
            this._loggerBuilder.build().error(e);
         }
         }

    static get() {
       try {
          const session = getNamespace(_NAMESPACE);
          const logger = session.get(_LOGGER_CONTEXT);
          if (logger) {
            return logger.build();
         }
      else {
         console.log('error retrieving logger in context');
      return console;
     }
   } catch (e) {
      if (global.infoLogger) {
        return global.infoLogger;
      } else {
         console.log('error retrieving logger in context:', e);
         return console;
     }
   }  }

[深山猎人]

in my application , when many request coming, cpu will up too , and not down. when remove cls, will be ok! expect answer, 3q

[Rémy Lespagnol]

Same issue... :/

When I have deployed the feature :

I'm on Docker, Node 12 Alpine, NestJS

My server have 3000req/15s

[Scott Latham]

I'm also seeing graphs strikingly similar to what @Themandunord is seeing. Something's definitely leaky - also using Docker, Node 12 Alpine.

So in my case, I was definitely using this wrong.

I had some middleware that was doing something like this:

const { createNamespace } = require('cls-hooked');

module.exports = (req, res, next) => {
    req.headers['x-request-id'] = 'whatever';

    const reqTrace = createNamespace('reqTrace');
    reqTrace.run(() => {
        reqTrace.set('requestId', req.headers['x-request-id']);
        next();
    });
}

We were creating the namespace every time the middle was invoked; moving const reqTrace = createNamespace('reqTrace'); out of the function seems to have resolved this.

Obvious in hindsight, but perhaps there should be a warning or error if you try to create a namespace that already exists?