Akvorado release 2.0: collects network flows with IPFIX and sFlow
12 hours ago
- #network-monitoring
- #performance-optimization
- #software-release
- Akvorado 2.0 introduces a major architectural change by splitting the inlet service into two parts: inlet and outlet, improving performance and simplifying the code.
- The new outlet service handles flow processing tasks like reading from Kafka and pushing to ClickHouse, with dynamic scaling of workers based on batch size and latency.
- Performance improvements include a reported 35% lower CPU usage and resolution of long-standing BMP component issues.
- Enhanced observability with Prometheus metrics, Kafka consumer lag metrics, and Docker Compose setup for Prometheus, Loki, and Grafana (though Grafana dashboards are currently broken).
- Significant documentation improvements, including updated architecture details, troubleshooting sections, and developer-focused content, with an LLM proofread.
- CI enhancements include parallel test execution, Docker image builds after test success, and end-to-end tests for Docker Compose setups.
- Docker improvements: multi-platform builds, optimized images for different CPU architectures, and replacement of deprecated Bitnami images with Apache Kafka.
- Go toolchain updates require Go 1.25, with automatic newer toolchain downloads, and adoption of modern packages like google/go-cmp and franz-go for Kafka.
- JavaScript dependency reduction from 653 to 225 by replacing npm-run-all, patch-package, and eslint with more efficient alternatives like Pnpm and Oxlint.
- Future plans include adding Grafana dashboards, integrating OVH’s Grafana plugin, and moving some ClickHouse tasks back to the outlet service for flexibility.