pixie

Pixie is an open-source observability tool for Kubernetes applications. It lets you view the high-level state of a cluster (service maps, cluster resources, application traffic) and drill down into detailed views such as pod state, flame graphs, and individual full-body application requests. Key characteristics described in the README: - Auto-telemetry: Pixie uses eBPF to automatically collect telemetry data including full-body requests, resource and network metrics, and application profiles. - In-cluster edge compute: telemetry is collected, stored and queried locally in the cluster. The README states Pixie uses less than 5% of cluster CPU and in most cases less than 2%. - Scriptability: PxL, a Pythonic query language, can be used across the UI, CLI and client APIs. Documented use cases include: - Network monitoring: traffic flow within the cluster, DNS request flow, individual full-body DNS requests and responses, and a map of TCP drops and retransmits. - Infrastructure health: resource usage by Pod, Node and Namespace, plus CPU flame graphs per Pod and Node. - Service performance: traffic flow between services, latency per service and endpoint, and samples of the slowest requests. - Database query profiling: latency, error and throughput rates for pods and normalized queries, plus individual full-body queries and responses. - Request tracing: full-body requests and responses for supported protocols, and error rate per Service and Pod. - Continuous application profiling to identify performance issues in application code. - Distributed bpftrace deployment: deploy a bpftrace program to all nodes, capture output into a table, and query/visualize it in the UI. - Dynamic Go logging: debug Go binaries in production without recompiling and redeploying. Getting started: installation is documented in the project's install guides. Once installed, users interact with Pixie through a web-based Live UI, a CLI, or an API. The project is community-driven and welcomes contributions. It was contributed by New Relic, Inc. to the Cloud Native Computing Foundation as a Sandbox project in June 2021, and is licensed under Apache License 2.0. Releases are versioned separately for components such as CLI, Cloud, Vizier and Operator.