[OCP] 3scale Operator 리소스 조정(관리)

ARO(Azure Redhat Openshift)에 3scale을 올려 사용하고있는데,

개발서버 쪽에 자원이 부족하다는 알림에 급히 3scale resource 조정을 확인해보았다.

 

3scale 기본사양

 먼저 3scale 각 pod들의 cpu/memory 제한을 살펴보자.

Component	CPU Requests	CPU Limits	Memory Requests	Memory Limits
system-app's system-master	50m	1000m	600Mi	800Mi
system-app's system-provider	50m	1000m	600Mi	800Mi
system-app's system-developer	50m	1000m	600Mi	800Mi
system-sidekiq	100m	1000m	500Mi	2Gi
system-sphinx	80m	1000m	250Mi	512Mi
system-redis	150m	500m	256Mi	32Gi
system-mysql	250m	No limit	512Mi	2Gi
system-postgresql	250m	No limit	512Mi	2Gi
backend-listener	500m	1000m	550Mi	700Mi
backend-worker	150m	1000m	50Mi	300Mi
backend-cron	50m	150m	40Mi	80Mi
backend-redis	1000m	2000m	1024Mi	32Gi
apicast-production	500m	1000m	64Mi	128Mi
apicast-staging	50m	100m	64Mi	128Mi
zync	150m	1	250M	512Mi
zync-que	250m	1	250M	512Mi
zync-database	50m	250m	250M	2G

cpu limit로 보면, 최대 13 core 까지 사용할 수 있는 상태이다.

ARO는 worker node VM사이즈에 따라 cpu, memory, pv 수량 제한들이 명확히 있는데,

현재 사용하는 worker node의 vm 제품은

 Standard_E4s_v3으로 vCPU 4core, Memory 32GB, pv는 8개까지 붙일 수 있다.

 

worker node*3대로 사용할 수 있는 최대 한도는 vcpu 12core, Memory 96GB, pv 24개....

이로써 일단 request cpu를 10m 으로 조정하였다.

 

3scale 리소스 조정

cmd버전

그냥 oc 명령어로 수정해도됨.

oc edit apimanager <생성한 manager 이름> -n 네임스페이스

$ oc edit apimanager apimanager-sample -n 3scale

웹콘솔버전

리소스 수정을 위해 3scale Operator > APIManager > YAML 로 이동한다.

 

YAML 파일 수정

apiVersion: apps.3scale.net/v1alpha1
kind: APIManager
metadata:
  annotations:
    apps.3scale.net/apimanager-threescale-version: '2.11'
    apps.3scale.net/threescale-operator-version: 0.8.1
spec:
  imageStreamTagImportInsecure: false
  resourceRequirementsEnabled: true
  system:
    appSpec:
      developerContainerResources:
        limits:
          cpu: 1
          memory: 800M
        requests:
          cpu: 10m
          memory: 600M
      masterContainerResources:
        limits:
          cpu: 1
          memory: 800M
        requests:
          cpu: 10m
          memory: 600M
      providerContainerResources:
        limits:
          cpu: 1
          memory: 800M
        requests:
          cpu: 10m
          memory: 600M
      replicas: 1
    database:
      mysql:
        resources:
          limits:
            cpu: 250m
            memory: 2Gi
          requests:
            cpu: 10m
            memory: 512M
    fileStorage:
      persistentVolumeClaim:
        storageClassName: azure-file
    memcachedResources:
      limits:
        cpu: 250m
        memory: 96M
      requests:
        cpu: 10m
        memory: 64M
    redisResources:
      limits:
        cpu: 500m
        memory: 32Gi
      requests:
        cpu: 10m
        memory: 256Mi
    sidekiqSpec:
      replicas: 1
      resources:
        limits:
          cpu: '1'
          memory: 2Gi
        requests:
          cpu: 10m
          memory: 500M
    sphinxSpec:
      resources:
        limits:
          cpu: '1'
          memory: 512M
        requests:
          cpu: 10m
          memory: 250M
  appLabel: 3scale-api-management
  zync:
    appSpec:
      replicas: 1
      resources:
        limits:
          cpu: '1'
          memory: 512M
        requests:
          cpu: 10m
          memory: 250M
    databaseResources:
      limits:
        cpu: 250m
        memory: 2Gi
      requests:
        cpu: 10m
        memory: 250M
    queSpec:
      replicas: 1
      resources:
        limits:
          cpu: '1'
          memory: 512M
        requests:
          cpu: 10m
          memory: 250M
  backend:
    cronSpec:
      replicas: 1
      resources:
        limits:
          cpu: 500m
          memory: 500M
        requests:
          cpu: 10m
          memory: 100M
    listenerSpec:
      replicas: 1
      resources:
        limits:
          cpu: 1
          memory: 700M
        requests:
          cpu: 10m
          memory: 550M
    redisResources:
      limits:
        cpu: '2'
        memory: 32Gi
      requests:
        cpu: 10m
        memory: 1Gi
    workerSpec:
      replicas: 1
      resources:
        limits:
          cpu: '1'
          memory: 300M
        requests:
          cpu: 10m
          memory: 50M
  tenantName: 3scale
  apicast:
    managementAPI: status
    openSSLVerify: false
    productionSpec:
      replicas: 1
      resources:
        limits:
          cpu: '1'
          memory: 128M
        requests:
          cpu: 10m
          memory: 64M
    registryURL: 'http://apicast-staging:8090/policies'
    responseCodes: true
    stagingSpec:
      replicas: 1
      resources:
        limits:
          cpu: 100m
          memory: 128M
        requests:
          cpu: 10m
          memory: 64M
  wildcardDomain: apps.i3jhq2a5.australiaeast.aroapp.io

 

반영결과 확인

$ oc get po -o custom-columns="Name:metadata.name,CPU-Requet:spec.containers[*].resources.requests.cpu,CPU-limit:spec.containers[*].resources.limits.cpu,MEM-Request:spec.containers[*].resources.requests.memory,MEM-limit:spec.containers[*].resources.limits.memory" -n 3scale

Name	CPU-Requet	CPU-limit	MEM-Request	MEM-limit
apicast-production-6-fbc42	10m	1	64M	128M
apicast-staging-5-wtcnj	10m	100m	64M	128M
backend-cron-5-4ccgp	10m	500m	100M	500M
backend-listener-5-72vm7	10m	1	550M	700M
backend-redis-7-n28pq	10m	2	1Gi	32Gi
backend-worker-5-cxqdz	10m	1	50M	300M
system-app-8-hw9qh (master)	10m	1	600M	800M
system-app-8-hw9qh (provider)	10m	1	600M	800M
system-app-8-hw9qh (developer)	10m	1	600M	800M
system-memcache-5-m88fh	10m	250m	64M	96M
system-mysql-5-wp8m6	10m	250m	512M	2Gi
system-redis-5-dx66s	10m	500m	256Mi	32Gi
system-sidekiq-5-zkg9l	10m	1	500M	2Gi
system-sphinx-5-srp8t	10m	1	250M	512M
threescale-operator-controller-manager-v2-74dd5566f5-g2dqp	100m	100m	100Mi	100Mi
zync-5-987jb	10m	1	250M	512M
zync-database-5-92kj5	10m	250m	250M	2Gi
zync-que-5-j58hl	10m	1	250M	512M

 

redhat 사이트의 docs를 통해 직접 테스트하였습니다.