[Bug]: Possible Memory leak?

[peterLam2]

🔎 Search Terms

Memory Leak, memory, performance, OOM

The problem

is 3.x.x Possible Memory leak?

What version of Winston presents the issue?

3.17.0

What version of Node are you using?

v20.19.0

If this worked in a previous version of Winston, which was it?

v2.4.7

Minimum Working Example

run script with node --max-old-space-size=200 

    logger = winston.createLogger(
      {
        transports: [
          new winston.transports.Console()
        ]
      }
    );

      for (let i = 0; i < 100000; i ++){
          logger.info('testing');
      }

Additional information

tested in winston 3.1.0, the situation is the same. and downgrade to v2.4.7 it work fine.

[DABH]

We are very open to any help investigating this issue. It looks like @eric-wei1990 @peterLam2 have created some good MWEs here and in #2465 - perhaps you can investigate a bit deeper and see where any potential leaks may be coming from?

[peterLam2]

@DABH
in 3.0.0-rc0 branch, removing
// var self = this;
// setImmediate(function () {
// self.emit('logged', info);
// });
will solve the memory leaking problem.

winston/lib/winston/transports/console.js

Line 53 in 1b9cef8

setImmediate(() => this.emit('logged', info));

but in 3.17.0, removing this line only make less memory leak it look like there are other leaking point i haven't found out.

[yash-dayama]

Hi @DABH & @peterLam2

I've investigated this issue and created a PR (#2564) that addresses the memory leak. Here's what I found and fixed:

1. The root cause was indeed the setImmediate wrapper in the Console transport's log method, which creates unnecessary pending operations under high load.

2. My fix:

   • Removes the setImmediate wrapper since console writes are already async
   • Adds comprehensive stress tests to verify memory usage remains stable
   • Maintains full backwards compatibility

3. Testing results:

   • Tested with the original MWE (100k messages)
   • Added memory usage tracking and event listener cleanup verification
   • Confirmed stable memory usage with garbage collection enabled

The PR includes detailed tests and documentation. Would appreciate your review when you have a chance.
Let me know if you need any additional information or changes!

[vilvic]

Hi, is this going to be fixed in 3.x? We've experience this issue in NFT and production and have regressed back to 2.4.7 from 3.17.0. We're seeing a significant memory leak which eventually crashes. On 2.4.7 we can see the memory is very stable. On 3.17.0 we run out of RAM under high load with the GC unable to recover.