Connection reset by peer when reading exec stdout/stderr

[HofmeisterAn]

Hi 👋,

I'm the maintainer of Testcontainers for .NET and Docker.DotNet. Some users have reached out to me reporting issues with Podman, specifically when running Podman on Linux (not Podman Desktop). I was able to reproduce the problem using Podman 5.5.2 (testcontainers/Docker.DotNet#11 (comment)).

The issue occurs when creating and starting an exec instance, and then reading from stdout or stderr via the hijacked stream. Occasionally, this results in the following exception:

System.IO.IOException: Unable to read data from the transport connection: Connection reset by peer.

So far, I haven't found anything wrong in the Docker.DotNet implementation.

From what I can tell, it seems like when the child process exits, Podman doesn't close the socket gracefully, it just aborts the connection. Is that expected behavior? When I catch the exception and treat it as EOS, everything works fine (cannot imagine this is right).

I couldn't find any official docs about this behavior, and I want to make sure I'm handling it correctly.

[Luap99]

Looking at the code I don't think we explicitly call close on the connection so it may be that we do not correctly handle the case.