Configuration_Guide

Configuration Guide

This guide walks through godon's configuration structure using examples from the godon repository.

Configuration Overview

A godon configuration defines an optimization run — what to tune, how to apply changes, what to observe, and what constraints to enforce.

┌─────────────────────────────────────────────────────────────┐
│                    Configuration                             │
│                                                              │
│  ├── Search Space     (what parameters to tune)             │
│  ├── Objectives       (what to optimize)                    │
│  ├── Effectuator      (how to apply changes)                │
│  ├── Reconnaissance   (how to observe results)              │
│  └── Guardrails       (safety constraints)                  │
│                                                              │
└─────────────────────────────────────────────────────────────┘

Example Configurations

Example configurations are available in the godon examples directory.

These examples cover common scenarios:

Example	Description
Database tuning	PostgreSQL, MySQL configuration optimization
Application tuning	Web server, cache configuration
Kubernetes tuning	Resource limits, replica counts

Configuration Sections

Search Space

Define parameters and their ranges:

parameters:
  pool_size:
    type: integer
    bounds: [5, 50]

  timeout_ms:
    type: continuous
    bounds: [100.0, 5000.0]

  algorithm:
    type: categorical
    choices: [lru, lfu, fifo]

Objectives

Define what to optimize:

objectives:
  - name: latency_p99
    direction: minimize

  - name: throughput
    direction: maximize

Effectuator

Define how to apply configurations:

effectuator:
  type: ssh
  host: 10.0.0.50
  user: admin
  apply:
    command: |
      sed -i 's/pool_size=.*/pool_size={{ pool_size }}/' /etc/app/config.ini
      systemctl restart app

Reconnaissance

Define how to observe results:

reconnaissance:
  type: prometheus
  endpoint: http://prometheus:9090
  metrics:
    - name: latency_p99
      query: histogram_quantile(0.99, rate(http_request_duration_seconds_bucket[1m]))

Guardrails

Define safety limits:

guardrails:
  - name: max_latency
    metric: latency_p99
    condition: "< 100ms"
    on_violation: fail_trial

Full Example

See the examples directory for complete, runnable configurations.