Total Visibility: Harnessing Pixie for a Full Observability Landscape
summary
Pixie uses eBPF and a PEM daemonset with Vizier/Kelvin collectors to grab cluster metrics, full‑body HTTP traces, and sampled CPU profiles into flamegraphs, with out‑of‑the‑box PXL (Python subset) scripts for net, DNS, Redis and more. Claims include 0.3% CPU overhead for sampling and that no telemetry leaves the cluster by default, but node storage keeps only 24‑hour local retention and logging isn’t supported; Grafana integration and CSV export are available for longer storage or dashboards.
AI-written summary · about
chapters (20)
- 0:00Introduction
- 1:07Why Pixie for observability
- 3:12Modern observability challenges
- 8:13Pixie features and advantages
- 9:42Pixie architecture
- 15:21Out-of-the-box scripts and PXL CLI
- 16:44Live demo setup and install
- 18:54Pixie dashboard and service maps
- 20:19Network flow and traffic analysis
- 21:24BPF scripts and DNS flow graphs
- 22:28Cluster and node monitoring
- 23:10Pod-level observability and flame graphs
- 24:15Process memory usage and inspection
- 25:19Profiling with heat maps and eBPF
- 26:22HTTP tracing and microservices
- 27:05Traditional vs eBPF profiling
- 29:40Pixie with Grafana
- 30:46CLI script execution
- 32:34Pixie vs Prometheus tips
- 33:21Q&A
related talks
14mFlexible, Open and Easy Observability for Developers
04mInstant, Auto-telemetry for K8s Observability
30mData Science for Infrastructure: Observe, Understand, Automate
26mTriaging Real-time Security Threats with eBPF-powered Observability
28mBpftrace Meets Pixie: Dynamic Monitoring of K8s Clusters Unleashed
10mLightning Talks: Detecting Data Exfiltration on the Edge with Pixie
26mFlux + Observability: Featuring Prometheus Operator and Pixie
52mMonitoring Kafka without instrumentation using eBPF