섀도우로 A/B 테스트를 실시합니다.
- Python 3.8 이상
- Docker
- Kubernetes 클러스터 또는 minikube
이 프로그램은 Kubernetes 클러스터 또는 minikube 가 필요합니다. Kubernetes 클러스터는 독자적으로 구축하거나, 각 클라우드 매니지드 서비스（GCP GKE、AWS EKS、MS Azure AKS 등）를 이용해 주십시오. GCP GKE 클러스터로 가동을 확인했습니다.
- 현재 디렉토리
$ pwd
~/ml-system-in-actions/chapter6_operation_management/shadow_ab_pattern- Docker 이미지 빌드
$ make build_all
# 실행 커맨드
# docker build \
# -t shibui/ml-system-in-actions:shadow_ab_pattern_api_0.0.1 \
# -f Dockerfile \
# .
# docker build \
# -t shibui/ml-system-in-actions:shadow_ab_pattern_loader_0.0.1 \
# -f model_loader/Dockerfile \
# .
# docker build \
# -t shibui/ml-system-in-actions:shadow_ab_pattern_client_0.0.1 \
# -f Dockerfile.client \
# .- Kubernetes 로 서비스 기동
$ make deploy
# 실행 커맨드
# istioctl install -y
# kubectl apply -f manifests/namespace.yml
# kubectl apply -f manifests/
# 가동확인
$ kubectl -n shadow-ab get all
# 출력
# NAME READY STATUS RESTARTS AGE
# pod/client 2/2 Running 0 44s
# pod/iris-rf-7cd8cb9d78-lcsq6 2/2 Running 0 43s
# pod/iris-svc-74dc7654b8-xthmh 2/2 Running 0 44s
# NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
# service/iris ClusterIP 10.4.0.14 <none> 8000/TCP 43s
# NAME READY UP-TO-DATE AVAILABLE AGE
# deployment.apps/iris-rf 3/3 3 3 44s
# deployment.apps/iris-svc 3/3 3 3 44s
# NAME DESIRED CURRENT READY AGE
# replicaset.apps/iris-rf-7cd8cb9d78 3 3 3 44s
# replicaset.apps/iris-svc-74dc7654b8 3 3 3 44s
# NAME REFERENCE TARGETS MINPODS MAXPODS REPLICAS AGE
# horizontalpodautoscaler.autoscaling/iris-rf Deployment/iris-rf <unknown>/70% 3 10 3 43s
# horizontalpodautoscaler.autoscaling/iris-svc Deployment/iris-svc <unknown>/70% 3 10 3 43s- 기동한 API 에 요청
# 클라이언트에 접속
$ kubectl -n shadow-ab exec -it pod/client bash
# 같은 엔드포인트에서 여러번 요청을 보냄
$ curl http://iris.shadow-ab.svc.cluster.local:8000/predict-test/000000
# {"job_id":"000000","prediction":[0.9709315896034241,0.015583082102239132,0.013485366478562355]}
$ curl http://iris.shadow-ab.svc.cluster.local:8000/predict-test/000000
# {"job_id":"000000","prediction":[0.9709315896034241,0.015583082102239132,0.013485366478562355]}
$ curl http://iris.shadow-ab.svc.cluster.local:8000/predict-test/000000
# {"job_id":"000000","prediction":[0.9709315896034241,0.015583082102239132,0.013485366478562355]}
$ curl http://iris.shadow-ab.svc.cluster.local:8000/predict-test/000000
# {"job_id":"000000","prediction":[0.9709315896034241,0.015583082102239132,0.013485366478562355]}
$ curl http://iris.shadow-ab.svc.cluster.local:8000/predict-test/000000
# {"job_id":"000000","prediction":[0.9709315896034241,0.015583082102239132,0.013485366478562355]}
# 클라이언트를 exit 하고, iris-rf 및 iris-svc 로그를 참조합니다. iris-rf、iris-svc 양측에서 추론이 실행되었음을 확인할 수 있습니다.
$ kubectl -n shadow-ab logs pod/iris-rf-7cd8cb9d78-lcsq6 iris-rf
# 출력
# [2021-02-06 09:51:35] [INFO] [14] [src.app.routers.routers] [_predict_test] [37] execute: [000000]
# [2021-02-06 09:51:35] [INFO] [14] [src.ml.prediction] [predict] [50] predict proba [0.99999994 0. 0. ]
# [2021-02-06 09:51:35] [INFO] [14] [src.app.routers.routers] [wrapper] [33] [iris_rf.onnx] [/predict-test] [000000] [1.0628700256347656 ms] [None] [[0.9999999403953552, 0.0, 0.0]]
# [2021-02-06 09:51:35] [INFO] [14] [uvicorn.access] [send] [458] 10.0.2.36:0 - "GET /predict-test/000000 HTTP/1.1" 200
$ kubectl -n shadow-ab logs pod/iris-svc-74dc7654b8-xthmh iris-svc
# 출력
# [2021-02-06 09:51:35] [INFO] [8] [src.app.routers.routers] [_predict_test] [37] execute: [000000]
# [2021-02-06 09:51:35] [INFO] [8] [src.ml.prediction] [predict] [50] predict proba [0.97093159 0.01558308 0.01348537]
# [2021-02-06 09:51:35] [INFO] [8] [src.app.routers.routers] [wrapper] [33] [iris_svc.onnx] [/predict-test] [000000] [0.8084774017333984 ms] [None] [[0.9709315896034241, 0.015583082102239132, 0.013485366478562355]]
# [2021-02-06 09:51:35] [INFO] [8] [uvicorn.access] [send] [458] 127.0.0.1:48148 - "GET /predict-test/000000 HTTP/1.1" 200- Kubernetes 에서 서비스 삭제
$ kubectl delete ns shadow-ab
$ istioctl x uninstall --purge -y
$ kubectl delete ns istio-system