Pixie — eBPF-Based Auto-Instrumentation for Kubernetes Observability

What it is

Pixie is a CNCF sandbox project that instruments Kubernetes workloads using eBPF probes attached at the kernel boundary. There are no agents to import, no SDKs to adopt, and no sampling. Pixie captures full-fidelity HTTP, gRPC, MySQL, Postgres, Redis, DNS, and TLS traffic automatically from every pod on every node.

Pixie is designed for platform engineering teams running Kubernetes who need instant observability without modifying application code. It runs as a DaemonSet with data staying inside the cluster.

How it saves time or tokens

Traditional observability requires adding instrumentation libraries to every service, configuring exporters, and managing sampling rates. Pixie eliminates all of that. Deploy it once, and it captures everything at the kernel level. There is no code change, no redeployment, and no per-service configuration. Debugging sessions that previously required log correlation across services become single PxL script queries.

How to use

1. Install the Pixie CLI.
2. Deploy Pixie to your Kubernetes cluster using your kubeconfig.
3. Run built-in scripts to view live traffic.

# Install the CLI
bash -c "$(curl -fsSL https://withpixie.ai/install.sh)"

# Deploy to your cluster
px deploy

# View live HTTP traffic by service
px run px/http_data

# Open the live UI
px live http_data

Example

Querying HTTP request latency by service using PxL (Pixie's query language):

import px

df = px.DataFrame(table='http_events', start_time='-5m')
df.service = df.ctx['service']
df.latency_ms = df.resp_latency_ns / 1e6
df = df.groupby('service').agg(
    p50=('latency_ms', px.quantiles, 0.5),
    p99=('latency_ms', px.quantiles, 0.99),
    count=('latency_ms', px.count),
)
px.display(df)

This returns p50 and p99 latency for every service in the cluster without any prior instrumentation.

Related on TokRepo

• AI Tools for Monitoring — Observability and monitoring tools for modern infrastructure
• AI Tools for DevOps — DevOps and platform engineering tools

Common pitfalls

• Pixie requires Linux kernel 4.14+ with eBPF support. Some managed Kubernetes providers restrict eBPF capabilities.
• Data stays in-cluster by default, which means limited retention (typically 24 hours). For long-term storage, export to an external backend.
• PxL scripts have a learning curve. Start with the built-in scripts before writing custom queries.