Kubernetes Operator (AIOps)

apiVersion: platform.chatcli.io/v1alpha1
kind: Instance
metadata:
  name: chatcli-prod
  namespace: chatcli          # O namespace deve existir antes de criar o Instance
spec:
  replicas: 1
  provider: CLAUDEAI       # OPENAI, CLAUDEAI, GOOGLEAI, XAI, STACKSPOT, OLLAMA, COPILOT
  model: claude-sonnet-4-5

  image:
    repository: ghcr.io/diillson/chatcli
    tag: latest
    pullPolicy: IfNotPresent

  server:
    port: 50051
    tls:
      enabled: true
      secretName: chatcli-tls
    token:
      name: chatcli-auth
      key: token

  watcher:
    enabled: true
    interval: "30s"
    window: "2h"
    maxLogLines: 100
    maxContextChars: 32000
    targets:
      - deployment: api-gateway
        namespace: production
        metricsPort: 9090
        metricsFilter: ["http_requests_*", "http_request_duration_*"]
      - deployment: auth-service
        namespace: production
        metricsPort: 9090
      - deployment: worker
        namespace: batch

  resources:
    requests:
      cpu: 100m
      memory: 128Mi
    limits:
      cpu: 500m
      memory: 512Mi

  persistence:
    enabled: true
    size: 1Gi
    storageClassName: standard

  securityContext:
    runAsNonRoot: true
    runAsUser: 1000
    seccompProfile:
      type: RuntimeDefault

  apiKeys:
    name: chatcli-api-keys

apiVersion: platform.chatcli.io/v1alpha1
kind: Anomaly
metadata:
  name: watcher-highrestartcount-api-gateway-1234567890
  namespace: production
spec:
  signalType: pod_restart    # pod_restart | oom_kill | pod_not_ready | deploy_failing | error_rate | latency_spike
  source: watcher            # watcher | prometheus | manual
  severity: warning          # critical | high | medium | low | warning
  resource:
    kind: Deployment
    name: api-gateway
    namespace: production
  description: "HighRestartCount on api-gateway: container app restarted 8 times"
  detectedAt: "2026-02-16T10:30:00Z"
status:
  correlated: true
  issueRef:
    name: api-gateway-pod-restart-1771276354

apiVersion: platform.chatcli.io/v1alpha1
kind: Issue
metadata:
  name: api-gateway-pod-restart-1771276354
  namespace: production
spec:
  severity: high
  source: watcher
  signalType: pod_restart        # Propagated from Anomaly for tiered Runbook matching
  description: "Correlated incident: pod_restart on api-gateway"
  resource:
    kind: Deployment
    name: api-gateway
    namespace: production
  riskScore: 65
  correlatedAnomalies:
    - name: watcher-highrestartcount-api-gateway-1234567890
    - name: watcher-oomkilled-api-gateway-1234567891
status:
  state: Analyzing          # Detected | Analyzing | Remediating | Resolved | Escalated | Failed
  remediationAttempts: 0
  maxRemediationAttempts: 3
  detectedAt: "2026-02-16T10:30:00Z"
  conditions:
    - type: Analyzing
      status: "True"
      reason: AIInsightCreated

apiVersion: platform.chatcli.io/v1alpha1
kind: AIInsight
metadata:
  name: api-gateway-pod-restart-1771276354-insight
  namespace: production
spec:
  issueRef:
    name: api-gateway-pod-restart-1771276354
  provider: CLAUDEAI
  model: claude-sonnet-4-5
status:
  analysis: "High restart count caused by OOMKilled. Container memory limit (512Mi) is insufficient for the current workload pattern."
  confidence: 0.87
  recommendations:
    - "Increase memory limit to 1Gi"
    - "Investigate possible memory leak in the application"
    - "Monitor GC pressure metrics"
  suggestedActions:
    - name: "Restart deployment"
      action: RestartDeployment
      description: "Restart pods to reclaim leaked memory immediately"
    - name: "Scale up replicas"
      action: ScaleDeployment
      description: "Add more replicas to distribute memory pressure"
      params:
        replicas: "4"
  generatedAt: "2026-02-16T10:31:00Z"

apiVersion: platform.chatcli.io/v1alpha1
kind: RemediationPlan
metadata:
  name: api-gateway-pod-restart-1771276354-plan-1
  namespace: production
spec:
  issueRef:
    name: api-gateway-pod-restart-1771276354
  attempt: 1
  strategy: "Attempt 1 (AI-generated): High restart count caused by OOMKilled"
  actions:
    - type: RestartDeployment
    - type: ScaleDeployment
      params:
        replicas: "4"
  safetyConstraints:
    - "No delete operations"
    - "No destructive changes"
    - "Rollback on failure"
status:
  state: Completed           # Pending | Executing | Completed | Failed | RolledBack
  result: "Deployment restarted and scaled to 4 replicas successfully"
  startedAt: "2026-02-16T10:31:30Z"
  completedAt: "2026-02-16T10:32:15Z"

apiVersion: platform.chatcli.io/v1alpha1
kind: RemediationPlan
metadata:
  name: api-gateway-pod-restart-plan-1
  namespace: production
spec:
  issueRef:
    name: api-gateway-pod-restart-1771276354
  attempt: 1
  strategy: "Agentic AI remediation"
  agenticMode: true
  agenticMaxSteps: 10
  agenticHistory:
    - stepNumber: 1
      aiMessage: "High restart count with OOMKilled. Scaling up to reduce memory pressure."
      action:
        type: ScaleDeployment
        params:
          replicas: "5"
      observation: "SUCCESS: ScaleDeployment executed successfully"
    - stepNumber: 2
      aiMessage: "Pods still restarting. Adjusting memory limits."
      action:
        type: AdjustResources
        params:
          memory_limit: "1Gi"
          memory_request: "512Mi"
      observation: "SUCCESS: AdjustResources executed successfully"
    - stepNumber: 3
      aiMessage: "All pods running stable. Issue resolved."
status:
  state: Completed
  agenticStepCount: 3
  agenticStartedAt: "2026-02-16T10:31:00Z"

apiVersion: platform.chatcli.io/v1alpha1
kind: PostMortem
metadata:
  name: pm-api-gateway-pod-restart-1771276354
  namespace: production
spec:
  issueRef:
    name: api-gateway-pod-restart-1771276354
  resource:
    kind: Deployment
    name: api-gateway
    namespace: production
  severity: high
status:
  state: Open              # Open | InReview | Closed
  summary: "OOMKilled containers caused cascading restarts on api-gateway"
  rootCause: "Memory limit (512Mi) insufficient for current workload pattern"
  impact: "Service degradation for 5 minutes, 30% error rate increase"
  timeline:
    - timestamp: "2026-02-16T10:30:00Z"
      type: detected
      detail: "Issue detected: pod_restart on api-gateway"
    - timestamp: "2026-02-16T10:31:00Z"
      type: action_executed
      detail: "ScaleDeployment to 5 replicas"
    - timestamp: "2026-02-16T10:31:35Z"
      type: action_executed
      detail: "AdjustResources memory_limit=1Gi"
    - timestamp: "2026-02-16T10:32:10Z"
      type: resolved
      detail: "All pods stable, issue resolved"
  lessonsLearned:
    - "Memory limits should account for peak workload patterns"
    - "Set up HPA to auto-scale on memory pressure"
  preventionActions:
    - "Configure HPA with min 3 replicas for api-gateway"
    - "Set memory limit to 1Gi across all environments"
  duration: "2m10s"
  generatedAt: "2026-02-16T10:32:10Z"

Tier 1: SignalType + Severity + ResourceKind (match exato, preferido)
Tier 2: Severity + ResourceKind (fallback quando signal não bate)

1. Runbook manual existente (match tiered)
2. Runbook auto-gerado pela IA (materializado como CR reutilizável)
3. Remediação agêntica por IA (loop observe-decide-act, gera PostMortem + Runbook)
4. Escalonamento (apenas quando agêntico falha após max tentativas)

cd operator

# Build
go build ./...

# Testes (96 funções, 125 com subtests)
go test ./... -v

# Docker (deve ser construído a partir do root do repositório)
docker build -f operator/Dockerfile -t myregistry/chatcli-operator:dev .

# Instalar CRDs no cluster
kubectl apply -f config/crd/bases/

# Deploy o operator
make deploy IMG=myregistry/chatcli-operator:dev