Kubernetes-архитектура

Helm-чарт ISP Platform — DRY defaults, 11 приложений, конфигурация по окружениям

Обзор

Единый Helm-чарт (infra/chart/) для staging и production (ArgoCD). В dev приложения запускаются нативно через bun dev / go run.

Ключевые принципы:

DRY defaults — общая конфигурация в defaults:, приложения только переопределяют нужное
range over apps — шаблоны итерируют по apps map, нет дублирования при добавлении приложений
merge в helpers — mustMergeOverwrite(defaults, app) объединяет конфиг автоматически
healthPath / readyPath — каждое приложение может задать свои пути для проб

Приложения (11 deployments)

Приложение	Образ	Порт	Стек	Health Path
docs	isp-platform/docs	3000	Next.js	/
web	isp-platform/web	3000	Next.js	/
customer-core	isp-platform/customer-core	3010	NestJS	/healthz
product-catalog	isp-platform/product-catalog	3011	NestJS	/healthz
billing	isp-platform/billing	3012	NestJS	/healthz
notification	isp-platform/notification	3013	NestJS	/healthz
oms	isp-platform/oms	3014	NestJS	/healthz
network-inventory	isp-platform/network-inventory	3015	NestJS	/healthz
provisioning-go	isp-platform/provisioning-go	8080	Go	/healthz
aaa-go	isp-platform/aaa-go	8081	Go	/healthz
mediation-go	isp-platform/mediation-go	8082	Go	/healthz

Helm values.yaml (структура)

defaults:
  image:
    pullPolicy: IfNotPresent
  replicas: 1
  port: 3000
  healthPath: /api/health # NestJS default
  readyPath: /api/health
  resources:
    requests: { cpu: 50m, memory: 128Mi }
    limits: { cpu: 500m, memory: 512Mi }

apps:
  # --- Frontend ---
  docs:
    image: { repository: isp-platform/docs, tag: latest }
  web:
    image: { repository: isp-platform/web, tag: latest }

  # --- BSS (NestJS) ---
  customer-core:
    image: { repository: isp-platform/customer-core, tag: latest }
    port: 3010
  product-catalog:
    image: { repository: isp-platform/product-catalog, tag: latest }
    port: 3011
  billing:
    image: { repository: isp-platform/billing, tag: latest }
    port: 3012
  notification:
    image: { repository: isp-platform/notification, tag: latest }
    port: 3013
  oms:
    image: { repository: isp-platform/oms, tag: latest }
    port: 3014
  network-inventory:
    image: { repository: isp-platform/network-inventory, tag: latest }
    port: 3015

  # --- OSS (Go) ---
  provisioning-go:
    image: { repository: isp-platform/provisioning-go, tag: latest }
    port: 8080
    healthPath: /healthz
    readyPath: /readyz
  aaa-go:
    image: { repository: isp-platform/aaa-go, tag: latest }
    port: 8081
    healthPath: /healthz
    readyPath: /readyz
  mediation-go:
    image: { repository: isp-platform/mediation-go, tag: latest }
    port: 8082
    healthPath: /healthz
    readyPath: /readyz

Production overrides (values.production.yaml)

apps:
  # Go-сервисы — меньше памяти (distroless, ~10MB binary)
  provisioning-go:
    resources:
      requests: { cpu: 100m, memory: 64Mi }
      limits: { cpu: 500m, memory: 256Mi }
  aaa-go:
    resources:
      requests: { cpu: 100m, memory: 64Mi }
      limits: { cpu: 500m, memory: 256Mi }
  mediation-go:
    resources:
      requests: { cpu: 200m, memory: 128Mi }
      limits: { cpu: 1000m, memory: 512Mi }

  # NestJS-сервисы — стандартные ресурсы
  billing:
    resources:
      requests: { cpu: 200m, memory: 256Mi }
      limits: { cpu: 1000m, memory: 1Gi }

Компоненты

Компонент	Dev	Production
SecurityContext	—	pod: runAsNonRoot, container: drop ALL
Probes	off	startup + readiness + liveness
HPA	off	2–20 реплик, CPU 70%
PDB	off	minAvailable: 1
NetworkPolicy	off	ingress-nginx only, egress DNS + external
TopologySpread	—	maxSkew: 1 по hostname
Ingress	— (apps via `bun dev`)	nginx + cert-manager TLS

Конфигурация по окружениям

	Dev	Staging	Production
Replicas	1	2	HPA 2–20
Resources CPU	50m–2	100m–500m	200m–1
Resources Memory	128Mi–2Gi	256Mi–512Mi	512Mi–1Gi
`NODE_ENV`	development	production	production
TLS	—	staging issuer	prod issuer

Команды

helm template isp infra/chart/                                  # dry-run (default values)
helm template isp infra/chart/ -f infra/chart/values.production.yaml  # production dry-run

helm upgrade --install isp infra/chart \
  -f infra/chart/values.production.yaml \
  -n isp-platform --create-namespace                            # manual install