Datadog uses gRPC, an open source RPC framework, to enable efficient communication between its distributed services. The company has learned several lessons from running a large mesh of gRPC services in a high-scale Kubernetes environment and identified best practices for using gRPC effectively. These include using Kubernetes headless services and gRPC round-robin to avoid load balance issues, leveraging TLS to handle IP recycling, setting MAX_CONNECTION_AGE to force gRPC to re-resolve from DNS, ensuring proper scale-out detection, setting the keepalive channel option to mitigate the effects of silent connection drops, and properly monitoring services.