Dockerfile and entrypoint.sh for tf serving model api

deployment.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: category-api-serving
  labels:
    app: category-api-serving
spec:
  replicas: 1
  selector:
    matchLabels:
      app: category-api-serving
  strategy:
    rollingUpdate:
      maxSurge: 25%
      maxUnavailable: 25%
    type: RollingUpdate
  template:
    metadata:
      labels:
        app: category-api-serving
    spec:
      containers:
        - image: asia.gcr.io/storage-273502/buzzni-ai-category-api-serving:latest
          lifecycle:
            preStop:
              exec:
                command:
                  - sleep
                  - "5"
          resources:
            requests:
              cpu: 1
          readinessProbe:
            httpGet:
              path: /v1/models/hybrid_classifier
              port: 8501
          name: category-serving-container
      serviceAccountName: gcs-reader

---

apiVersion: autoscaling/v1
kind: HorizontalPodAutoscaler
metadata:
  name: category-api-serving-hpa
  namespace: default
spec:
  maxReplicas: 10
  minReplicas: 1
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: category-api-serving
  targetCPUUtilizationPercentage: 80

deployment_service.yaml

apiVersion: v1
kind: Service
metadata:
  name: category-api-serving-service
  labels:
    app: category-api-serving-service
  namespace: default
spec:
  ports:
    - port: 80
      protocol: TCP
      targetPort: 8501
  selector:
    app: category-api-serving
  sessionAffinity: None
  type: ClusterIP


---

apiVersion: traefik.containo.us/v1alpha1
kind: IngressRoute
metadata:
  name: vpn-web-category-serving-api
  namespace: default
spec:
  entryPoints:
    - vpn-web
  routes:
    - kind: Rule
      match: PathPrefix(`/category-serving/`)
      middlewares:
        - name: api-strip
      services:
        - name: category-api-serving-service
          port: 80

Dockerfile

# https://github.com/tensorflow/serving/blob/master/tensorflow_serving/tools/docker/Dockerfile

ARG TF_SERVING_VERSION=latest
ARG TF_SERVING_BUILD_IMAGE=tensorflow/serving:${TF_SERVING_VERSION}

FROM ${TF_SERVING_BUILD_IMAGE} as build_image
FROM python:3.7

# Install apt package
RUN apt-get update && apt-get install -y --no-install-recommends \
        ca-certificates software-properties-common git curl

RUN apt-get clean && \
    rm -rf /var/lib/apt/lists/*

# Install TF Serving pkg
COPY --from=build_image /usr/bin/tensorflow_model_server /usr/bin/tensorflow_model_server

# Install poetry
RUN pip3 install poetry

# Prepare tf-serving model
WORKDIR /clf_api
COPY poetry.lock .
COPY pyproject.toml .

# Install dependency packages
RUN poetry config virtualenvs.create false
RUN poetry install -n

# Copy other files
COPY . .

# Expose ports
# gRPC
EXPOSE 8500

# REST
EXPOSE 8501

# Make entrypoint
RUN chmod +x scripts/serving_api_entrypoint.sh
ENTRYPOINT ["scripts/serving_api_entrypoint.sh"]

serving_api_entrypoint.sh

#!/bin/bash

export PYTHONPATH=/clf_api
export MODEL_BASE_PATH=/models

mkdir -p ${MODEL_BASE_PATH}
cp ./models.config ${MODEL_BASE_PATH}/

python3.7 buzzni/ai/category_api/serving/prepare_serving.py

tensorflow_model_server --port=8500 --rest_api_port=8501 --model_config_file=${MODEL_BASE_PATH}/models.config "$@"