Open Virtual Switch (OVS) Exporter

A Prometheus exporter for Open Virtual Switch (OVS) that provides comprehensive metrics from OVS and OVN components, including advanced DPDK/PMD performance metrics.

Note: This is an enhanced fork of the original greenpau/ovs_exporter with significant improvements including:

Updated to Go 1.24 with modernized dependencies

Comprehensive PMD/DPDK performance metrics

Prometheus naming convention compliance

System ID retrieval from OVS database (no config file required)

Flow cache performance metrics

Detailed drop statistics

Security updates and vulnerability fixes

Features

Comprehensive Metrics: Exports 100+ metrics from OVS components
DPDK/PMD Support: Advanced performance metrics for DPDK deployments
Auto-Discovery: Automatically retrieves system ID from OVS database
Multi-Tenant Ready: Supports external IDs for tenant association
Production Ready: Battle-tested in large-scale deployments
Low Overhead: Efficient polling with configurable intervals
Prometheus Compliant: Follows official naming conventions

Installation

Binary Installation

Download the latest release for your architecture:

# Linux AMD64
wget https://github.com/Liquescent-Development/ovs_exporter/releases/download/v2.8.0/ovs-exporter-2.8.0.linux-amd64.tar.gz
tar xvzf ovs-exporter-2.8.0.linux-amd64.tar.gz

# Linux ARM64
wget https://github.com/Liquescent-Development/ovs_exporter/releases/download/v2.8.0/ovs-exporter-2.8.0.linux-arm64.tar.gz
tar xvzf ovs-exporter-2.8.0.linux-arm64.tar.gz

Install as a systemd service:

cd ovs-exporter-*
sudo ./install.sh
systemctl status ovs-exporter

The install script will:

Automatically detect system architecture (amd64/arm64)
Build the binary if not present (requires Go and Make)
Create an ovs_exporter user and group
Install the binary to /usr/sbin/ovs-exporter
Set up systemd service with proper capabilities
Configure permissions for OVS socket access
Optionally configure remote syslog forwarding

Remote Syslog Configuration

The install script supports forwarding logs to a remote syslog server:

# Forward logs via UDP (default)
sudo ./install.sh --syslog-server 192.168.1.100

# Forward logs via TCP with custom port
sudo ./install.sh --syslog-server syslog.example.com --syslog-port 6514 --syslog-protocol tcp

# View all options
sudo ./install.sh --help

Remote syslog options:

--syslog-server HOST - Remote syslog server hostname or IP address
--syslog-port PORT - Remote syslog server port (default: 514)
--syslog-protocol PROTO - Protocol to use: udp or tcp (default: udp)

Verify the installation:

curl -s localhost:9475/metrics | grep ovs_up
# HELP ovs_up Is OVN stack up (1) or is it down (0).
# TYPE ovs_up gauge
ovs_up 1

Building from Source

Requirements:

Go 1.24 or later
Make
Open vSwitch installed (for testing)

git clone https://github.com/Liquescent-Development/ovs_exporter.git
cd ovs_exporter

# Build for current architecture (auto-detected)
make

# Build for specific architecture
make BUILD_OS="linux" BUILD_ARCH="arm64"

# Run quick test (requires sudo)
make qtest

# Install using the install script (recommended)
sudo ./install.sh

# Or install manually via make
sudo make deploy

Docker

docker run -d \
  --name ovs-exporter \
  --net host \
  --pid host \
  -v /var/run/openvswitch:/var/run/openvswitch:ro \
  -v /var/log/openvswitch:/var/log/openvswitch:ro \
  -v /etc/openvswitch:/etc/openvswitch:ro \
  ghcr.io/liquescent-development/ovs-exporter:latest

Configuration

Command Line Flags

ovs-exporter --help

Key flags:

Flag	Default	Description
`-web.listen-address`	`:9475`	Address to listen on for metrics
`-web.telemetry-path`	`/metrics`	Path for metrics endpoint
`-ovs.poll-interval`	`15`	Seconds between metric collections
`-ovs.poll-timeout`	`5`	Timeout for OVS operations
`-log.level`	`info`	Log level (debug, info, warn, error)
`-database.vswitch.socket.remote`	`unix:/var/run/openvswitch/db.sock`	OVS database socket
`-database.vswitch.file.system.id.path`	`/etc/openvswitch/system-id.conf`	System ID file (fallback only)

System ID Configuration

The exporter automatically retrieves the system ID in the following order:

From OVS database: ovs-vsctl get Open_vSwitch . external-ids:system-id
From file: /etc/openvswitch/system-id.conf (fallback)
Uses "unknown" if both fail (non-fatal)

For newer OVS versions, no system-id.conf file is needed!

Systemd Configuration

Edit /etc/sysconfig/ovs-exporter to set options:

OPTIONS="-ovs.poll-interval 10 -log.level debug"

Metrics

See METRICS.md for complete documentation of all metrics.

Metric Categories

System Metrics: Overall health, version info, total requests, failed requests
Process Metrics: Component PIDs, file sizes, log events
Datapath Metrics: Flows, lookups, masks, interfaces
Interface Metrics: Traffic statistics, errors, configuration
PMD/DPDK Metrics: CPU utilization, packet processing, cache performance
vHost Metrics: Queue statistics, TX retries, contentions
Drop Statistics: Detailed drop reasons and counters

Key Metrics Examples

# System health
ovs_up{system_id="uuid"} 1
ovs_info{system_id="uuid", hostname="host1", ovs_version="2.17.0"} 1
ovs_requests_total{system_id="uuid"} 12345
ovs_failed_requests_total{system_id="uuid"} 2

# Interface traffic (with Prometheus naming conventions)
ovs_interface_rx_packets_total{uuid="123", name="eth0"}
ovs_interface_rx_bytes{uuid="123", name="eth0"}
ovs_interface_rx_dropped_total{uuid="123", name="eth0"}

# PMD performance (DPDK)
ovs_pmd_cpu_utilization_ratio{pmd_id="0", numa_id="0"}
ovs_pmd_cycles_per_packet{pmd_id="0", numa_id="0"}
ovs_flow_cache_emc_hit_ratio{pmd_id="0", numa_id="0"}

# External IDs (for multi-tenant)
ovs_interface_external_ids{uuid="123", key="tenant-id", value="customer-456"} 1
ovs_interface_external_ids{uuid="123", key="vm-uuid", value="vm-789"} 1

Example Queries

Basic Monitoring

# Check OVS health across all nodes
up{job="ovs-exporter"}

# Interface packet rate
rate(ovs_interface_rx_packets_total[5m])

# Interface errors
rate(ovs_interface_rx_errors_total[5m])

# Datapath flow count
ovs_dp_flows

# PMD CPU utilization (DPDK)
ovs_pmd_cpu_utilization_ratio * 100

Multi-Tenant Monitoring

OVS supports external IDs on interfaces for metadata. Use Prometheus joins to aggregate by tenant:

Setting External IDs in OVS

# Add tenant ID to an interface
ovs-vsctl set Interface vnet0 external-ids:tenant-id="customer-123"
ovs-vsctl set Interface vnet0 external-ids:project="web-app"
ovs-vsctl set Interface vnet0 external-ids:environment="production"

Querying by Tenant

# Total bandwidth by tenant
sum by (value) (
  rate(ovs_interface_rx_bytes[5m])
  * on(uuid) group_left(value)
  ovs_interface_external_ids{key="tenant-id"}
)

# Packet rate by tenant
sum by (value) (
  rate(ovs_interface_rx_packets_total[5m])
  * on(uuid) group_left(value)
  ovs_interface_external_ids{key="tenant-id"}
)

# Errors by tenant and environment
sum by (tenant, env) (
  rate(ovs_interface_rx_errors_total[5m])
  * on(uuid) group_left(tenant)
    label_replace(ovs_interface_external_ids{key="tenant-id"}, "tenant", "$1", "value", "(.*)")
  * on(uuid) group_left(env)
    label_replace(ovs_interface_external_ids{key="environment"}, "env", "$1", "value", "(.*)")
)

Using Recording Rules for Performance

For frequently-used tenant queries, create recording rules in Prometheus:

# prometheus-rules.yml
groups:
  - name: tenant_metrics
    interval: 30s
    rules:
      - record: tenant:ovs_interface_rx_bytes:rate5m
        expr: |
          sum by (value) (
            rate(ovs_interface_rx_bytes[5m])
            * on(uuid) group_left(value)
            ovs_interface_external_ids{key="tenant-id"}
          )

      - record: tenant:ovs_interface_rx_packets:rate5m
        expr: |
          sum by (value) (
            rate(ovs_interface_rx_packets_total[5m])
            * on(uuid) group_left(value)
            ovs_interface_external_ids{key="tenant-id"}
          )

Then query the pre-computed metrics:

tenant:ovs_interface_rx_bytes:rate5m

Advanced Queries

# Top 5 interfaces by traffic
topk(5, rate(ovs_interface_rx_bytes[5m]))

# Interfaces with high drop rate
rate(ovs_interface_rx_dropped_total[5m]) > 100

# PMD threads with low cache hit rate (DPDK)
ovs_flow_cache_emc_hit_ratio < 0.9

# vHost queue saturation
ovs_pmd_avg_vhost_queue_length / ovs_pmd_max_vhost_queue_length > 0.8

# Datapath lookup efficiency
ovs_dp_lookups_hit_total / (ovs_dp_lookups_hit_total + ovs_dp_lookups_missed_total)

# Identify specific drop reasons
topk(10, increase(ovs_datapath_drops_total[5m]))

Grafana Dashboards

Recommended Panels

System Overview
- OVS Up/Down status
- Component versions
- Failed requests rate
Interface Metrics
- Traffic rates (in/out)
- Error rates
- Drop rates
- Top interfaces by traffic
PMD Performance (DPDK only)
- CPU utilization heatmap
- Cycles per packet
- Cache hit rates
- Queue depths
Multi-Tenant View
- Bandwidth per tenant
- Top tenants by traffic
- Error rates per tenant

Import Dashboard

A sample Grafana dashboard is available at grafana/dashboard.json.

Troubleshooting

Exporter Issues

No metrics available

# Check if OVS is running
systemctl status openvswitch

# Check socket permissions
ls -l /var/run/openvswitch/db.sock

# Check exporter logs
journalctl -u ovs-exporter -f

# Test manual connection
ovs-vsctl show

System ID issues

# Check if system-id is in database (newer OVS)
ovs-vsctl get Open_vSwitch . external-ids:system-id

# Check fallback file (older OVS)
cat /etc/openvswitch/system-id.conf

# Set system-id manually if needed
ovs-vsctl set Open_vSwitch . external-ids:system-id=$(uuidgen)

PMD metrics missing

# Verify DPDK is enabled
ovs-vsctl get Open_vSwitch . other_config:dpdk-init

# Check PMD threads
ovs-appctl dpif-netdev/pmd-stats-show

Performance Issues

High cardinality concerns

The exporter uses external_ids as separate metrics, not labels
Use Prometheus joins for tenant aggregation
Consider recording rules for frequently-used queries

Slow queries

Optimize PromQL queries (avoid regex where possible)
Use recording rules for complex joins
Consider increasing polling interval if needed

Development

Project Structure

ovs_exporter/
├── cmd/ovs_exporter/      # Main application
├── pkg/ovs_exporter/       # Core exporter logic
├── assets/systemd/         # Systemd service files
├── Makefile               # Build automation
├── METRICS.md             # Complete metrics documentation
└── CLAUDE.md              # AI assistant guidelines

Building

# Standard build
make

# Cross-compilation
make BUILD_OS=linux BUILD_ARCH=arm64

# Run tests
make test

# Create distribution
make dist

Testing

# Unit tests
go test ./...

# Integration test (requires OVS)
make qtest

# Coverage report
make coverage

Release Process

# 1. Update VERSION file
echo "2.6.0" > VERSION

# 2. Build release artifacts
make dist

# 3. Create git tag
git tag -a v2.6.0 -m "Release v2.6.0"
git push origin v2.6.0

# 4. Upload to GitHub releases
# Use GitHub UI or gh CLI

Contributing

Contributions are welcome! Please:

Fork the repository
Create a feature branch
Add tests for new functionality
Ensure metrics follow Prometheus naming conventions
Update METRICS.md for new metrics
Submit a pull request

License

Apache License 2.0 - See LICENSE file

Credits

Original author: Paul Greenberg
Enhanced fork: Liquescent Development
Inspired by: Red Hat's OVS observability blog

Support

For issues and questions:

GitHub Issues: https://github.com/Liquescent-Development/ovs_exporter/issues
Metrics Documentation: METRICS.md

Name		Name	Last commit message	Last commit date
Latest commit History 79 Commits
.github/workflows		.github/workflows
assets		assets
cmd/ovs_exporter		cmd/ovs_exporter
contrib/kolla-ansible		contrib/kolla-ansible
dashboards		dashboards
pkg/ovs_exporter		pkg/ovs_exporter
.gitignore		.gitignore
.goreleaser.yml		.goreleaser.yml
Dockerfile.build-from-source		Dockerfile.build-from-source
Dockerfile.example		Dockerfile.example
Dockerfile.multiarch		Dockerfile.multiarch
Dockerfile.smart		Dockerfile.smart
LICENSE		LICENSE
MAINTAINERS.md		MAINTAINERS.md
METRICS.md		METRICS.md
MIGRATION.md		MIGRATION.md
Makefile		Makefile
README.md		README.md
VERSION		VERSION
go.mod		go.mod
go.sum		go.sum
install.sh		install.sh

Folders and files

Latest commit

History

Repository files navigation

Open Virtual Switch (OVS) Exporter

Table of Contents

Features

Installation

Binary Installation

Remote Syslog Configuration

Building from Source

Docker

Configuration

Command Line Flags

System ID Configuration

Systemd Configuration

Metrics

Metric Categories

Key Metrics Examples

Example Queries

Basic Monitoring

Multi-Tenant Monitoring

Setting External IDs in OVS

Querying by Tenant

Using Recording Rules for Performance

Advanced Queries

Grafana Dashboards

Recommended Panels

Import Dashboard

Troubleshooting

Exporter Issues

No metrics available

System ID issues

PMD metrics missing

Performance Issues

High cardinality concerns

Slow queries

Development

Project Structure

Building

Testing

Release Process

Contributing

License

Credits

Support

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases 5

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages