Memory leak in Image Optimization

[Pythe1337N]

What version of Next.js are you using?

10.0.9

What version of Node.js are you using?

14.15.5

What browser are you using?

Chrome

What operating system are you using?

alpine (docker image)

How are you deploying your application?

express server

Describe the Bug

Upgrading from 10.0.7 to 10.0.8 or 10.0.9 results in a server side memory leak when using the component.
Consumes over 1GB of memory after only 10-20 image optimizations.

Expected Behavior

Memory usage should be normal.

To Reproduce

Import Image from 'next/image' and use it as the the documentation says https://nextjs.org/docs/api-reference/next/image with an allowed external image source.
In this case layout="fixed" was used.

[shuding]

Could you please try installing the latest canary (next@canary)? It should be fixed there, see also #22925.

[Pythe1337N]

Tried with 10.0.10-canary.4 and I'm still having issues. Been running in a 512 MB ram docker image up to version 10.0.7. Increased it to 1024 MB to test this, but it's still using all of the memory available.

[fabb]

we are also having memory leaks and pod crashes since updating from next.js 10.0.7 to 10.0.9 yesterday. we updated a few other dependencies as well, so it's not yet certain that the next.js updated caused the issue. note that we do not use the image optimization feature.

[Pythe1337N]

I did a rollback on all deps other than for next and still had the same problem. The memory leak might be part of the image optimization, but that is what is running slow. To fetch an image from the _next/image endpoint is really slow while everything else seems to run as usual. Will try to only fetch images to try to pinpoint the leak.

[fabb]

i did a rollback of just next.js to 10.0.7 and it fixed the issue. so it's definitely caused by next.js. as i said, we are not using the image optimization feature at all.

[shuding]

We've fixed some know issues regarding memory usage in canary already. It would be great if you can test next@canary vs latest stable release and share more information, so we can identify the cause quickly!

[ahmetuysal]

I think the problem persists in 10.0.10-canary.5. Memory utilization is steadily increasing in our website deployed to AWS Beanstalk. It was stable until I updated next to 10.0.10-canary.5 from 10.0.7.

[tiagocarmo]

I had the same memory problem here.
We did some tests on an application that we developed, when version 10.0.7 occurs, the memory increases consumption.
With version 10.0.6 it is ok to use.

[stefanprobst]

as i understand it, these memory issues are caused by the switch from sharp to squoosh in #22253. the reason for the change given in that pr is that:

the native sharp dependency (which doesn't work on some Linux variants, nor M1 Mac)

i wonder if this is still a valid concern with the soon-to-be-released sharp v0.28 (see here). i'm also curious about a response to the sharp maintainer's comment here.

[gu-stav]

Alrighty, then here again :)

v10.0.10-canary.7 doesn't fix the issue in my case. It's still:

10.0.10-canary.7: 1960MB
10.0.7: 230MB

[mjenzen]

canary 7 didnt fix the issue for me either. I am showing a massive memory leak on my Windows 10 dev machine at 6,5GB/16GB and 3.5GB/8GB on my Azure Linux App Service. The resources are never recovered until the dev process or app service are stopped