How atharvasss/ai-k8s-incident-response works
This repository demonstrates a Kubernetes incident-response workflow: operators observe a simulated failure, investigate metrics and centralized logs, use an AI analyzer for remediation guidance, and verify recovery. The README documents the observability and recovery stages; the sampled analyzer reads incident data and calls a local Ollama endpoint. The collector script and Kubernetes manifests are not sampled, so their detailed behavior and connections cannot be verified here.
Main entry points: app.yaml Kubernetes Cluster · elastic-values.yaml Elasticsearch · ai_analyzer.py AI Analyzer · incident.txt Simulated Failure
Components 13
Kubernetes Workload
- Kubernetes Cluster
k8s/app.yaml - Demo Application
k8s/app.yaml - Log Generator
k8s/log-generator.yaml - Simulated Failure
incident-data/incident.txt
Observability
- Prometheus / Grafana
Log Investigation
- Elasticsearch
elastic-values.yaml - Kibana
kibana-values.yaml
AI Analysis
- Incident Collector
scripts/collect_incident.sh - Incident Evidence
incident-data/incident.txt - AI Analyzer
scripts/ai_analyzer.py - Local Ollama
Recovery
- Recovery Verification
k8s/app.yaml
Other components
- Operator
Connections 11
- Operator → Simulated Failure: triggers
README.md - Simulated Failure → Demo Application: causes failure
README.md - Demo Application → Kubernetes Cluster: runs in
README.md - Kubernetes Cluster → Prometheus / Grafana: observed with
README.md - Demo Application → Elasticsearch: logs centralized
README.md - Kubernetes Cluster → Elasticsearch: logs centralized
README.md - Elasticsearch → Kibana: visualized in
README.md - Operator → Incident Collector: collects evidence
README.md - AI Analyzer → Incident Evidence: reads
scripts/ai_analyzer.py - AI Analyzer → Local Ollama: requests analysis
scripts/ai_analyzer.py - Operator → Recovery Verification: checks recovery
README.md