Memory leak after upgrading to .net 9 with large PageSize

[valentasm1]

Upgrade project to net9.0 without changing anything. Doing 4th request it consumes ~5gb ram. There is no memory leak with .net8

Just do 4 request to https://localhost:7071/odata/humanentity
Project uses in memory db, but have [EnableQuery(AllowedQueryOptions = AllowedQueryOptions.All, PageSize = 1_000_000_000)] large page size.

Repro repository
https://github.com/valentasm1/Odata9MemoryLeak/tree/main

[corranrogue9]

@valentasm1 can you clarify, are you only changing the .NET version in your .csproj or are you updating other dependencies with the intent of moving to .NET 9?

Also, it's interesting that you are using such a large page size. Paging is actually disabled by default if you're selecting that page size in order to avoid paging. Can you give some more insight about why you've chosen this page size so we might be able to help you work around the issue while we investigate?

[valentasm1]

I only change from <TargetFramework>net8.0</TargetFramework> to <TargetFramework>net9.0</TargetFramework> I understand that such large page size look strange, but still it should not consume 5gb of ram. Maybe it is a sign of memory leaking.
I understand that workaround is to remove page size. Probably more right way than workaround.
Issue persist that it a leaks memory with such small requests. It even dont come back after request ended.

[xuzhg]

@valentasm1 Can I set a remote call with you to understand the details? If avaiable , please share a link So I can join it?

[xuzhg]

I use your repro, run the sample and keep eye on Diagnostic tools on Visual Studio:

And I took four snapshots for the memory:

1. After starting up and no request
2. Send 4 requests
3. Then waiting couple mins
4. Then send another 4 requests.

This is the diff between 1 and 2:

When first request coming, the ASP.NET Core will finish a lot of configurations so the memory usage is a little bit heavy. Can you point me how do you monitor the memory changes?

[valentasm1]

TaskManager.
I tried to run it in release mode. It produce same big memory consumption.