# Bookstore — Part 12 ch.06 "Model serving and inference": the recommender
# served as a PLAIN Deployment (built-in objects only). This is the
# kind-runnable serving path — `docker build`, `kind load`, `kubectl apply`,
# done. No KServe required.
#
# The CRD-backed equivalent (KServe `InferenceService`) is
# `recommender-inferenceservice.yaml` in this directory; it ADDS scale-to-zero,
# transformer/explainer wiring, and model versioning. Both consume the SAME
# image (`bookstore/recommender-serve:dev`) and the SAME `model.joblib`
# artifact from `../train/recommender-train-job.yaml`.
#
# BUILT-IN OBJECTS ONLY (apps/v1 Deployment + v1 Service in the sibling
# `recommender-service.yaml`). `kubectl apply --dry-run=client/server` PASSES
# cleanly anywhere.
#
# PSA — `bookstore-ml` is `enforce: restricted` (Part 12 ch.01). Pod carries
# the full restricted shape (runAsNonRoot, non-root UID 65532, drop ALL caps,
# seccomp RuntimeDefault, allowPrivilegeEscalation:false, readOnlyRootFilesystem
# + an emptyDir at /tmp). Image bakes uid 65532 too.
#
# MODEL — mounted from the SAME PVC the train Job writes to
# (`recommender-model`). On kind that's the simplest path. In production the
# model is more commonly fetched from a model registry / object store at
# startup or baked into the image — see the chapter's "Production notes".
#
# Apply (after `recommender-train-job.yaml` has completed and produced
# `model.joblib` on the PVC):
#   kubectl apply -f examples/bookstore/ml/serve/recommender-deployment.yaml
#   kubectl apply -f examples/bookstore/ml/serve/recommender-service.yaml
#   kubectl rollout status deploy/recommender -n bookstore-ml --timeout=120s
#   kubectl port-forward -n bookstore-ml svc/recommender 8080:8080 &
#   curl -s -X POST localhost:8080/v1/models/recommender:predict \
#     -H 'content-type: application/json' \
#     -d '{"instances":[{"book_id":1,"k":3}]}'
apiVersion: apps/v1
kind: Deployment
metadata:
  name: recommender
  namespace: bookstore-ml
  labels:
    app.kubernetes.io/part-of: bookstore-ml
    app.kubernetes.io/component: recommender-serve
    ml.bookstore/path: deployment
spec:
  replicas: 1                          # HPA can scale this; see chapter
  revisionHistoryLimit: 3
  strategy:
    type: RollingUpdate
    rollingUpdate:
      maxUnavailable: 0
      maxSurge: 1
  selector:
    matchLabels:
      app.kubernetes.io/part-of: bookstore-ml
      app.kubernetes.io/component: recommender-serve
  template:
    metadata:
      labels:
        app.kubernetes.io/part-of: bookstore-ml
        app.kubernetes.io/component: recommender-serve
    spec:
      automountServiceAccountToken: false
      securityContext:                 # pod-level — restricted
        runAsNonRoot: true
        runAsUser: 65532
        runAsGroup: 65532
        fsGroup: 65532                 # PVC mount readable by uid 65532
        seccompProfile:
          type: RuntimeDefault
      containers:
        - name: recommender
          image: bookstore/recommender-serve:dev
          imagePullPolicy: IfNotPresent
          ports:
            - containerPort: 8080
              name: http
          env:
            - name: MODEL_DIR
              value: /workspace/model
            - name: MODEL_NAME
              value: recommender
            - name: DEFAULT_K
              value: "5"
          resources:
            requests:
              cpu: "100m"
              memory: 256Mi
            limits:
              cpu: "1"
              memory: 512Mi
          readinessProbe:
            httpGet: { path: /ready, port: 8080 }
            initialDelaySeconds: 2
            periodSeconds: 5
          livenessProbe:
            httpGet: { path: /healthz, port: 8080 }
            initialDelaySeconds: 10
            periodSeconds: 20
          securityContext:             # container-level — restricted
            allowPrivilegeEscalation: false
            readOnlyRootFilesystem: true
            capabilities:
              drop: ["ALL"]
          volumeMounts:
            - name: model
              mountPath: /workspace/model
              readOnly: true
            - name: scratch
              mountPath: /tmp
      volumes:
        - name: model                  # the same PVC the train Job wrote
          persistentVolumeClaim:
            claimName: recommender-model
            readOnly: true             # serve never writes to the model
        - name: scratch
          emptyDir:
            sizeLimit: 64Mi