What is psm-interop?

This project is a test framework that checks whether "proxyless service mesh" applications work correctly on Google Kubernetes Engine. In plain terms, when microservices talk to each other, a "service mesh" manages that traffic, handling things like load balancing and security. The "proxyless" approach means this management happens inside the application itself rather than through a separate intermediary. This testing tool spins up clients and servers in the cloud and verifies they can communicate reliably, either with or without encrypted connections. At a high level, the tool automates the entire process of setting up the cloud infrastructure needed for testing. It creates isolated environments for each test run, configures Google's Traffic Director (the traffic management service), deploys test clients and servers, and then runs various checks. It includes a baseline test suite to confirm core networking features work, plus a security test suite to verify encrypted communication (mTLS/TLS). Developers can run these tests from their local laptop against a remote cloud cluster, and the framework handles the complex wiring automatically. The primary users are engineers working on gRPC (a communication protocol) or teams integrating proxyless service mesh features into their infrastructure. For example, if a team is building a new version of a gRPC library and wants to confirm it still handles traffic routing or mutual TLS correctly, they would use this framework. It helps them catch regressions or infrastructure-related failures before changes reach production. The tool is clearly built for the Google Cloud Platform ecosystem, requiring specific Google Cloud services and permissions to function. The project is marked as experimental and explicitly a work in progress. The roadmap shows the team has focused on making tests more robust and secure, such as running clients and servers in separate isolated environments to better simulate real-world conditions and allowing tests to run in parallel without resource conflicts. The framework also includes helper scripts that let developers manually start and stop individual components, which is useful for debugging specific issues step by step rather than running the entire automated suite.