Skip to content

A Kubernetes-native network observability tool that uses eBPF to auto-instrument network traffic and export it as Flow Traces via OpenTelemetry, providing deep visibility into cluster communications.

License

Apache-2.0, GPL-2.0 licenses found

Licenses found

Apache-2.0
LICENSE-APACHE
GPL-2.0
LICENSE-GPL2
Notifications You must be signed in to change notification settings

elastiflow/mermin

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Mermin

Mermin is a powerful, Kubernetes-native network observability tool that uses eBPF to efficiently capture network traffic and export it as Flow Traces via the OpenTelemetry Protocol (OTLP). It provides deep visibility into your cluster's network communications with zero application changes required.


Why Mermin?

The Problem

Your APM traces show application behavior. Your network monitoring shows IP-level statistics. But there's a gap: when a trace shows a slow network span, you have no way to correlate that with actual network flow data. When network teams see congestion, they can't map it back to specific services or pods.

The MELT stack (Metrics, Events, Logs, Traces) is missing network flow data—connection-level information that bridges application performance with network reality.

What Mermin Does

Mermin captures network traffic using eBPF and exports it as Flow Traces—network flows represented as OpenTelemetry spans. This brings network visibility into the OTel ecosystem using a standard signal type.

The "Sweet Spot": Why Flow Data?

Observability involves trade-offs between granularity and overhead. Flow data sits between two extremes:

  • Not Raw PCAP: Full packet capture is expensive to store and query. Mermin aggregates packets into flows—you get connection-level detail without payload overhead.
  • Not Just Counters: Metrics tell you bandwidth usage but miss connection context—timing, retransmissions, directionality.

Flow data provides granular, connection-level detail that's lightweight enough to run always-on in production.

Key Capabilities

  • Auto-Instrumentation for Your Network Stack: Just as eBPF-based APM tools auto-instrument application code, Mermin auto-instruments your network layer. Deploy once per node, get visibility into all traffic—no per-service configuration required.
  • Kubernetes-Native Enrichment: Flows include Pod, Service, and Deployment metadata. You see frontend-serviceredis-cache, not 10.42.0.510.42.0.8.
  • Zero Code Changes: eBPF captures traffic transparently—no sidecars, no application modifications, no service mesh required.
  • Standards-Based Export: Native OTLP output integrates with your existing observability stack (Tempo, Jaeger, Elastic, etc.).
  • Production-Ready: Low-overhead kernel-level capture designed for always-on operation.

Quick Start

The fastest way to get started with Mermin is to deploy it to a local Kubernetes cluster using our quickstart guide:

Follow the Complete Quickstart Guide

Or deploy directly with Helm:

helm repo add elastiflow https://elastiflow.github.io/mermin
helm install mermin elastiflow/mermin --namespace mermin --create-namespace

Once deployed, Mermin runs as a DaemonSet with one pod per node, automatically capturing network traffic and exporting Flow Traces to your configured OTLP endpoint.

How It Compares

Feature Mermin eBPF APM Agents Traditional NetFlow/IPFIX Service Mesh (Istio/Linkerd) Packet Capture Tools
Kubernetes Context ✅ Native ✅ Native ❌ None ✅ Native ❌ None
Application Changes ✅ Zero ✅ Zero ✅ Zero ❌ Sidecar injection ✅ Zero
Network Data Type ✅ Flow Records ❌ Counters only ✅ Flow Records ⚠️ Request/response ✅ Full packets
Connection Context ✅ Full details ❌ Aggregated metrics ✅ Full details ⚠️ L7 only ✅ Full packets
Performance Overhead ✅ Minimal (eBPF) ✅ Minimal (eBPF) ✅ Low ⚠️ Moderate (sidecars) ❌ High (full capture)
Standards-Based Export ✅ OTLP Traces ⚠️ OTLP Metrics ⚠️ NetFlow/IPFIX ⚠️ Prometheus/vendor-specific ❌ PCAP files
Bidirectional Flows ✅ Yes ❌ Separate counters ✅ Yes ⚠️ Limited ❌ Packet-level only
Deployment Complexity ✅ Simple DaemonSet ✅ Simple DaemonSet ✅ Simple ⚠️ Complex ✅ Simple

Key Differentiators:

  • vs eBPF APM Agents: Exports flow records (with timing, flags, directionality) as traces, not aggregated counter metrics
  • vs NetFlow/IPFIX: Adds Kubernetes context and uses modern OTLP standard
  • vs Service Meshes: No application changes, lower overhead, but L3/L4 only (not L7)
  • vs Packet Capture: Aggregated flows instead of raw packets, with metadata enrichment

Architecture at a Glance

Mermin operates as a DaemonSet in Kubernetes (or as a privileged container on bare metal), with one instance running on each node:

  1. Packet Capture: eBPF programs attached to network interfaces capture packets at the kernel level.
  2. Flow Aggregation: Packets are aggregated into bidirectional network flows with connection state tracking.
  3. Metadata Enrichment: Flows are decorated with Kubernetes metadata (pods, services, deployments, labels).
  4. Flow Traces Export: Flows are converted to OpenTelemetry trace spans and exported via OTLP
  5. Observability Backend: Flow Traces are stored, analyzed, and visualized in your platform (Elastic, Grafana Tempo, Jaeger, etc.)

For a deeper dive into Mermin's architecture, see our Architecture Overview.

Documentation

Comprehensive documentation is available in the docs/ directory:

Getting Started

Deployment

Configuration

Observability Backends

Troubleshooting

Development & Contributing

Contributing

We welcome contributions from the community! Whether you're fixing bugs, adding features, improving documentation, or sharing feedback, your contributions help make Mermin better for everyone.

Ways to Contribute

  • 🐛 Report bugs via GitHub Issues.
  • 💡 Request features or share ideas in GitHub Discussions.
  • 📝 Improve documentation - PRs for doc improvements are always welcome.
  • 🔧 Submit code - See our Development Workflow to get started.
  • 💬 Help others - Answer questions in Discussions or Slack.

Getting Started as a Contributor

  1. Read the Development Workflow - Learn how to build, test, and develop Mermin.
  2. Check out Good First Issues - Find beginner-friendly tasks.
  3. Join the conversation - Connect with us on Slack or GitHub Discussions.

All contributors are expected to follow our code of conduct and maintain a welcoming, inclusive environment.

Community & Support

Get Help:

Stay Updated

  • ⭐ Star the repo on GitHub to stay notified of new releases.
  • 📢 Follow us for announcements and updates.
  • 🔔 Watch the repository for issue and discussion notifications.

Enterprise Support

Artifacts

Docker images are available in the GitHub Container Registry:

Standard Production Image

docker pull ghcr.io/elastiflow/mermin:latest

Debug Image (includes shell and debugging tools)

docker pull ghcr.io/elastiflow/mermin:latest-debug

The debug image is built using the gcr.io/distroless/cc-debian12:debug base image and provides additional debugging tools compared to the standard image.

License

With the exception of eBPF code, Mermin is distributed under the terms of the Apache License (version 2.0).

Any contribution intentionally submitted for inclusion in this crate by you, shall be licensed Apache-2.0, without any additional terms or conditions.

eBPF

All eBPF code is distributed under the terms of the GNU General Public License, Version 2.

Any contribution intentionally submitted for inclusion in this project by you, shall be dual licensed GPL-2, without any additional terms or conditions.

About

A Kubernetes-native network observability tool that uses eBPF to auto-instrument network traffic and export it as Flow Traces via OpenTelemetry, providing deep visibility into cluster communications.

Topics

Resources

License

Apache-2.0, GPL-2.0 licenses found

Licenses found

Apache-2.0
LICENSE-APACHE
GPL-2.0
LICENSE-GPL2

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Packages

 
 
 

Contributors 13