bug: Memory leak using watchEvent

[adriencohen]

Is there an existing issue for this?

• I have searched the existing issues

Package Version

0.3.36

Current Behavior

Using the watchEvent function is leading to a growing memory consumption which can only be stopped using the unwatch function returned by watchEvent. Memory consumption will then stagnate.

Expected Behavior

Memory consumption shouln't be evergrowing

Steps To Reproduce

A quick example running the following code will produce the graph attached after:

import { createPublicClient } from 'viem';
import { mainnet } from 'viem/chains';

const client = createPublicClient({
  chain: mainnet,
  transport: http(),
});

setTimeout(() => {
  const unwatch = client.watchEvent(
    { 
      onLogs: () => {},
    }
  ) 
  console.log('watch')


  setTimeout(() => {
    unwatch()
    console.log('Unwatched')
  }, 300000)

}, 300000)


setInterval(() => {
    console.log('Memory used:', process.memoryUsage().heapUsed / 1024 / 1024);
}, 10000)

Using http transport (node v18.12.1):

Using ws transport (node v18.12.1):

Link to Minimal Reproducible Example (StackBlitz, CodeSandbox, GitHub repo etc.)

No response

Anything else?

No response

[fubhy]

I am seeing similar results in other observers too. So likely not related to watchEvent itself but underlying utils.

[jxom]

Think I found the culprit. It's in our withTimeout util. Need to find out what's causing it though.

[jxom]

can you try see if this still happens on npm i [email protected]?

the memory leak seems to be gone completely when i remove abort controllers, seems like we need to clean them up: 6be7922#diff-8b7a876b21ac77521fcb4a60160e0baceb18c2415682748d306b0d2fb445c480R35

[fubhy]

I don't think it's about cleaning up the AbortController and instead more about cleaning up the lingering fetch?

[jxom]

What do you mean? When I remove the AbortController completely, I no longer see a memory leak.

[jxom]

Might be an issue with how built-in fetch handles signals: nodejs/node#46525 (comment)

[fubhy]

True. But when I run a script with short timeouts the memory still goes up so I suspect there's sth. with fetch?

[fubhy]

Might be an issue with how built-in fetch handles signals: nodejs/node#46525 (comment)

Good find. Maybe that's what I'm seeing?

[fubhy]

can you try see if this still happens on npm i [email protected]?

the memory leak seems to be gone completely when i remove abort controllers, seems like we need to clean them up: 6be7922#diff-8b7a876b21ac77521fcb4a60160e0baceb18c2415682748d306b0d2fb445c480R35

Can confirm that works to also clean up the lingering fetchs.

[jxom]

Looks like this was an issue with previous versions of Node, and was fixed downstream here: nodejs/undici#2049. Also related: #385 (reply in thread)

I tried again on Node v20, and it seems to work fine.

Solution: Update Node to >=20.1.0. No change needed on viem side.

[github-actions]

This issue has been locked since it has been closed for more than 14 days.

If you found a concrete bug or regression related to it, please open a new bug report with a reproduction against the latest Viem version. If you have any questions or comments you can create a new discussion thread.