[Bug]: [java] Memory leak in long working tests

[piotrczarnak]

Version

1.52.0

Steps to reproduce

You need docker to run this project
https://github.com/piotrczarnak/bug-1816_spring
Steps to reproduce:

1. Download repo
   2 .Go to project with terminal and use docker build -t [imagename] .
2. docker run --init --ipc=host -t [tagoimagefrom2ndpoint] mvn compile exec:java -D exec.mainClass="org.example.App"
3. Wait ~4h and you will see memory consumption growing.

To check memory consumption per process use stem -t (already installed)

Expected behavior

Memory consumption stays on the same level.

Actual behavior

I have constantly running tests in docker container. Each hour I see memory consumption is growing even though I close each test page, browser and playwright itself. Test stared with 456 MB and after 8h it reached 4GB

Additional context

Tests are run in docker container which is created by myself. Tests are run each 1000 ms.

I noticed that there are many not closed processes. After using stem -t

Environment

OS: jvm-ubuntu:21
Language: Java

[Skn0tt]

It's hard for me to find the time to leave it running for 4 hours, but I checked your repro to see if we're leaking browser processes. I couldn't find evidence of it, the browsers are gracefully closing fine. If we were leaking browser processes, there'd be much more (1 per second!) and memory would fill up much quicker. The number of processes you're seeing on htop is most likely just the normal chromium subprocess structure.

I could definitely see us having a memory leak somewhere, but doing the math, it's around 4gb/8/60/60=139kb per iteration. Maybe you can employ some Java memory profiling tools to figure out where it is. If you can provide us with pointers, and it's easy enough to fix, we can take a look. Otherwise, this is low priority for us.

[unlikelyzero]

linking to #1841