This page groups host-level tuning guidance relevant to production Spooky deployments.

Host Priorities

• sufficient UDP buffer sizing
• sufficient file descriptor limits
• stable CPU scheduling under multi-worker load
• predictable network path MTU
• minimal interference from unrelated noisy workloads

Linux Tuning Areas

Important areas to validate:

• receive and send socket buffer sizes
• device backlog and packet budget
• file descriptor ceilings
• capability model for privileged ports
• conntrack impact, if present in the environment

Built-In Project Guidance

The repository already includes:

• production guidance in Production Deployment
• a Linux sysctl helper in scripts/sysctl-linux-network-tuning.sh

Use those as a baseline, then tune with real traffic and host telemetry.

Practical Advice

• do not treat aggressive sysctl values as universally correct
• validate tuning with the same traffic pattern you expect in production
• keep cert, config, and log path permissions minimal
• isolate the process from unrelated noisy co-located workloads where possible
• verify privileged-port bind strategy before rollout