fraud-detection-demo/docker-compose.yaml at main · PureStorage-OpenConnect/fraud-detection-demo · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
services:

  # ── Pod 1: Data Generation ──────────────────────────────────────────────
  data-gather:
    build: ./pods/data-gather
    volumes:
      - raw_data:/data/raw
      - stress_config:/data/stress
    environment:
      - OUTPUT_PATH=/data/raw
      - STRESS_CONFIG_PATH=/data/stress/stress.conf
      - NUM_WORKERS=${NUM_WORKERS:-8}
      - CHUNK_SIZE=${CHUNK_SIZE:-100000}
      - RUN_MODE=${RUN_MODE:-once}
      - TARGET_ROWS=${TARGET_ROWS:-1000000}
      - FRAUD_RATE=${FRAUD_RATE:-0.005}
      - KAGGLE_SEED_PATH=${KAGGLE_SEED_PATH:-}
    profiles: ["pipeline"]

  # ── Pod 2: Feature Engineering ──────────────────────────────────────────
  data-prep:
    build: ./pods/data-prep
    volumes:
      - raw_data:/data/raw:ro
      - features_data:/data/features
    environment:
      - INPUT_PATH=/data/raw
      - OUTPUT_PATH=/data/features
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]
    profiles: ["pipeline"]
    depends_on:
      data-gather:
        condition: service_completed_successfully

  # ── Pod 3: Model Training ────────────────────────────────────────────────
  model-build:
    build: ./pods/model-build
    volumes:
      - features_data:/data/features:ro
      - model_repo:/data/models
    environment:
      - INPUT_PATH=/data/features
      - MODEL_REPO=/data/models
      - MAX_SAMPLES=${MAX_SAMPLES:-500000}
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]
    profiles: ["pipeline"]
    depends_on:
      data-prep:
        condition: service_completed_successfully

  # ── Pod 4: Triton Inference ──────────────────────────────────────────────
  inference:
    build: ./pods/inference
    ports:
      - "8000:8000"   # HTTP
      - "8001:8001"   # gRPC
      - "8002:8002"   # Metrics
    volumes:
      - model_repo:/data/models:ro
    environment:
      - MODEL_REPO=/data/models
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]
    healthcheck:
      test: ["CMD", "curl", "-sf", "http://localhost:8000/v2/health/ready"]
      interval: 10s
      timeout: 5s
      retries: 12
      start_period: 60s

  # ── Pod 5: Backend + Dashboard ───────────────────────────────────────────
  backend:
    build: ./pods/backend
    ports:
      - "8080:8080"
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock   # docker compose control
      - raw_data:/data/raw:ro
      - features_data:/data/features:ro
      - model_repo:/data/models:ro
      - stress_config:/data/stress
      - ./docker-compose.yaml:/app/docker-compose.yaml:ro
    environment:
      - COMPOSE_FILE=/app/docker-compose.yaml
      - COMPOSE_PROJECT=${COMPOSE_PROJECT:-fraud-det-v31}
      - PROMETHEUS_URL=http://prometheus:9090
      - STRESS_CONFIG_PATH=/data/stress/stress.conf
      - RAW_DATA_PATH=/data/raw
      - FEATURES_DATA_PATH=/data/features
      - MODEL_REPO_PATH=/data/models
    depends_on:
      - prometheus

  # ── Monitoring: Prometheus ───────────────────────────────────────────────
  prometheus:
    image: prom/prometheus:v2.51.0
    ports:
      - "9090:9090"
    volumes:
      - ./monitoring/prometheus.yml:/etc/prometheus/prometheus.yml:ro

  # ── Monitoring: DCGM Exporter (GPU metrics) ─────────────────────────────
  dcgm-exporter:
    image: nvcr.io/nvidia/k8s/dcgm-exporter:3.3.5-3.4.0-ubuntu22.04
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
    ports:
      - "9400:9400"
    cap_add:
      - SYS_ADMIN
    environment:
      - DCGM_EXPORTER_COLLECT_INTERVAL_MS=200

# ── Volumes ──────────────────────────────────────────────────────────────────
volumes:
  raw_data:       # FlashBlade NFS mount in production; named Docker volume for local dev
  features_data:
  model_repo:
  stress_config:

# ── Network ──────────────────────────────────────────────────────────────────
networks:
  default:
    name: fraud-net