ML_Models_Production_Guide/docker-compose.local-infra.yml at main · Alba-Intelligence/ML_Models_Production_Guide · 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
# docker-compose.local-infra.yml
# Local emulation: compute-plane infrastructure
# Emulates Kubernetes (K3s) and Slurm scheduling.
# Pair with docker-compose.aws-emulator.yml for AWS APIs and docker-compose.dev.yml for the data plane.
#
# Usage:
#   docker compose -f docker-compose.aws-emulator.yml -f docker-compose.dev.yml -f docker-compose.local-infra.yml up
#   or separately:
#   docker compose -f docker-compose.local-infra.yml up
#
# DeploymentProfile: local_emulation

services:

  # ── Kubernetes in Docker (K3s) ─────────────────────────────────────────────
  k3s:
    image: rancher/k3s:v1.29.4-k3s1
    container_name: ml_deploy_k3s
    privileged: true
    command: server --disable traefik --disable metrics-server
    ports:
      - "6443:6443"         # Kubernetes API
      - "8080:8080"
    environment:
      - K3S_KUBECONFIG_OUTPUT=/output/kubeconfig.yaml
      - K3S_KUBECONFIG_MODE=666
    volumes:
      - k3s_data:/var/lib/rancher/k3s
      - k3s_output:/output
    networks:
      - ml_local
    healthcheck:
      test: ["CMD", "k3s", "kubectl", "get", "nodes"]
      interval: 30s
      timeout: 15s
      retries: 10
      start_period: 60s

  # ── Slurm cluster (controller + worker) ───────────────────────────────────
  slurmctld:
    image: nathanhess/slurm:latest
    container_name: ml_deploy_slurmctld
    hostname: slurmctld
    command: ["slurm_controller"]
    ports:
      - "6817:6817"         # slurmctld port
    volumes:
      - slurm_shared:/shared
      - slurm_state:/var/spool/slurmctld
    networks:
      ml_local:
        aliases:
          - slurmctld
    healthcheck:
      test: ["CMD", "scontrol", "ping"]
      interval: 20s
      timeout: 10s
      retries: 6
      start_period: 30s

  slurmd:
    image: nathanhess/slurm:latest
    container_name: ml_deploy_slurmd
    hostname: slurmd
    command: ["slurm_worker"]
    depends_on:
      slurmctld:
        condition: service_healthy
    volumes:
      - slurm_shared:/shared
    networks:
      ml_local:
        aliases:
          - slurmd
    environment:
      - SLURM_CONTROLLER_HOST=slurmctld

volumes:
  k3s_data:
  k3s_output:
  slurm_shared:
  slurm_state:

networks:
  ml_local:
    name: ml_local
    driver: bridge