predictive_log_analyzer
Analyze historical log patterns with machine learning to predict potential failures before critical outages occur, enabling proactive system maintenance.
Instructions
Predict failures using ML analysis of historical log patterns before critical outages occur.
Uses anomaly detection algorithms to correlate log patterns with failure events.
Args:
prediction_window: Time window - "1h", "6h", "24h", "7d" (default: "6h").
confidence_threshold: Min confidence for predictions 0.0-1.0 (default: 0.75).
log_sources: Sources to analyze - pods, services, nodes (default: all).
failure_types: Types to predict - pod_crash, resource_exhaustion, network_issues.
historical_data_range: Historical data period (default: "30d").
model_refresh_interval: Model retrain frequency (default: "24h").
namespaces: Specific namespaces to analyze (default: auto-detect active namespaces).
max_namespaces: Maximum namespaces to scan when auto-detecting (default: 20).
Returns:
Dict: Keys: predictions, model_performance, anomaly_scores, trend_analysis.
Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| prediction_window | No | 6h | |
| confidence_threshold | No | ||
| log_sources | No | ||
| failure_types | No | ||
| historical_data_range | No | 30d | |
| model_refresh_interval | No | 24h | |
| namespaces | No | ||
| max_namespaces | No |