Release v2.3.1

What's Changed

New Features

• Add total requests counter metric (ovs_requests_total) for accurate success rate calculation
  • Tracks all requests made to the OVS stack
  • Enables accurate calculation of request success rates
  • Previously only failed requests were tracked, making true success rate impossible to determine

Improvements

• Updated Grafana dashboards to use accurate success rate formula: (1 - (failed_requests / total_requests)) * 100
• Enhanced metrics documentation with new metric and example queries
• Added IncrementRequestCounter() method to track total requests across all OVS operations

Metrics

New metric added:

• ovs_requests_total - Counter for total number of requests to OVN stack

Example Usage

# Calculate request success rate
(1 - (rate(ovs_failed_requests_total[5m]) / rate(ovs_requests_total[5m]))) * 100

Full Changelog

v2.3.0...v2.3.1

Release v2.3.0

What's Changed

Major Improvements

• Prometheus Naming Compliance: All metrics now follow official Prometheus naming conventions

  • Added _total suffix to all counter metrics
  • Added unit suffixes (_bytes, _seconds, _kilobits, etc.)
  • Changed percentage metrics to ratios (0-1 scale)
  • Improved naming clarity throughout

• System ID Auto-Discovery: The exporter now retrieves system-id from the OVS database first

  • No more need for /etc/openvswitch/system-id.conf on newer OVS versions
  • Automatically falls back to file for older versions
  • Uses ovs-vsctl get Open_vSwitch . external-ids:system-id

• Documentation Overhaul

  • New comprehensive METRICS.md documenting all 100+ metrics
  • Completely rewritten README with correct installation instructions
  • Added extensive multi-tenant monitoring examples
  • Included troubleshooting guide

Multi-Tenant Monitoring

The README now includes detailed examples of using Prometheus joins with external_ids for tenant-based aggregation, avoiding cardinality issues while enabling powerful multi-tenant queries.

Breaking Changes

⚠️ Many metric names have changed to comply with Prometheus naming conventions. Please review METRICS.md for the complete list of current metric names.

Full Changelog

• refactor: update metrics to conform to Prometheus naming best practices
• feat: retrieve system-id from OVS database before falling back to file
• docs: consolidate metrics documentation into comprehensive METRICS.md
• feat: add comprehensive flow cache performance metrics
• feat: implement comprehensive PMD metrics collection

Installation

Download the appropriate release for your architecture:

Linux AMD64

```bash
wget https://github.com/Liquescent-Development/ovs_exporter/releases/download/v2.3.0/ovs-exporter-2.3.0.linux-amd64.tar.gz
tar xvzf ovs-exporter-2.3.0.linux-amd64.tar.gz
cd ovs-exporter-*
sudo ./install.sh
```

Linux ARM64

```bash
wget https://github.com/Liquescent-Development/ovs_exporter/releases/download/v2.3.0/ovs-exporter-2.3.0.linux-arm64.tar.gz
tar xvzf ovs-exporter-2.3.0.linux-arm64.tar.gz
cd ovs-exporter-*
sudo ./install.sh
```

Verification

```bash
systemctl status ovs-exporter
curl -s localhost:9475/metrics | grep ovs_up
```

v2.2.0 - Complete PMD and Flow Cache Metrics

🎉 Complete Feature Implementation

This release completes the implementation of ALL metrics from the Red Hat OVS observability blog post.

✅ PMD Performance Metrics

• CPU Metrics: Per-PMD thread CPU utilization, busy/idle cycles
• Packet Processing: Packets per iteration, total packets processed
• RX/TX Batch Statistics: Complete batch size tracking with averages and maximums
• Iteration Tracking: Busy vs idle iterations, sleep iterations
• Cache Hit Rates: Exact match, masked hits, misses per PMD thread

✅ Flow Cache Performance

• EMC (Exact Match Cache): Hit rate, total hits, insertions
• SMC (Signature Match Cache): Hit rate, total hits
• Megaflow Cache: Hit rate, hits, misses
• Total Flow Lookups: Comprehensive lookup tracking

✅ Detailed Drop Statistics

• All 24 specific drop counter categories
• Individual metrics via ovs_datapath_drops_total with drop_reason label
• Includes critical counters like drop_action_of_pipeline

🐛 Bug Fixes

• Fixed missing PMD metrics in Prometheus Describe() method
• Added proper metric registration for all new metrics

📊 New Metrics Summary

• 9 new flow cache performance metrics
• 24 individual drop counter metrics
• Enhanced PMD metrics with complete coverage
• All metrics properly exposed to Prometheus

📚 Documentation

• Updated PMD_METRICS.md with all new metrics
• Added example Prometheus queries for flow cache monitoring
• Comprehensive metric descriptions and labels

🔧 Compatibility

• Go 1.24 support
• Multi-architecture binaries (AMD64 and ARM64)
• Compatible with DPDK and non-DPDK OVS deployments

v2.1.0 - Enhanced PMD Metrics

What's New

Enhanced PMD/DPDK Metrics

• Comprehensive PMD performance metrics collection
• CPU utilization tracking for PMD threads
• RX/TX batch statistics with averages and maximums
• Individual drop counter metrics (24 specific drop reasons)
• vHost queue metrics and TX statistics
• Flow lookup hit/miss statistics
• Suspicious iteration detection
• Fallback mechanisms for non-DPDK deployments

Features

• All metrics from Red Hat's OVS observability blog post
• Automatic detection of DPDK vs non-DPDK deployments
• Enhanced parsing for complete metric extraction
• Detailed drop categorization via coverage metrics

Documentation

• Comprehensive PMD metrics documentation in PMD_METRICS.md
• Example queries and Grafana dashboard suggestions
• Troubleshooting guide for PMD metric collection

Compatibility

• Go 1.24 support
• Updated Prometheus client libraries
• Multi-architecture support (AMD64 and ARM64)

v2.0.0 - Major Upgrade with PMD Metrics

Major Improvements

• Go 1.24 - Modernized from Go 1.20
• PMD/DPDK Metrics - Comprehensive performance metrics for DPDK deployments
• Security Updates - All dependencies updated with vulnerability fixes
• Enhanced Coverage - All metrics from Red Hat's OVS observability blog

New Metrics

• PMD performance metrics (cycles, packets, batches)
• vHost statistics (TX retries, contention, IRQs)
• Upcall metrics
• See PMD_METRICS.md for full documentation

Breaking Changes

• Module path changed to github.com/Liquescent-Development/ovs_exporter
• Minimum Go version is now 1.24

Migration

See MIGRATION.md for upgrade instructions from v1.x