talon/docker-compose.yml at main · milvus-io/talon · 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
# Talon local development cluster.
#
# Default (memory backend, single coordinator + worker):
#   docker compose up
#   # UI + API: http://127.0.0.1:8000/ui
#
# HA (etcd backend, 3 coordinators + worker):
#   docker compose --profile ha up
#   # UI + API (any coordinator): http://127.0.0.1:8000/ui
#
# Images default to the published GHCR builds. To build locally instead:
#   docker compose build          # uses the deploy/docker/*.Dockerfile
#   docker compose up --build
#
# The worker requires an object-store backend to be *configured* to start; the
# Azure account/SAS below are placeholders so the control plane, API, and UI
# work for exploration. Reading real objects needs real credentials.

name: talon

services:
  # --- default: single-node memory backend ---------------------------------
  coordinator:
    image: ghcr.io/milvus-io/talon-coordinator:latest
    build:
      context: .
      dockerfile: deploy/docker/coordinator.Dockerfile
    command: ["--cluster-id", "demo", "--node-id", "coord-0"]
    environment:
      TALON_COORDINATOR_STATE_BACKEND: memory
    ports:
      - "7000:7000" # control plane
      - "8000:8000" # admin API + UI
    healthcheck:
      test: ["CMD", "curl", "-fsS", "http://127.0.0.1:8000/healthz"]
      interval: 10s
      timeout: 3s
      retries: 5

  worker:
    image: ghcr.io/milvus-io/talon-worker:latest
    build:
      context: .
      dockerfile: deploy/docker/worker.Dockerfile
    command:
      - "--coordinator"
      - "coordinator:7000"
      - "--cluster-id"
      - "demo"
      - "--node-id"
      - "worker-0"
    environment:
      # Placeholder object-store credentials so the worker starts; the control
      # plane / UI work, but reading real objects needs real credentials.
      TALON_WORKER_AZURE_ACCOUNT: demo
      TALON_WORKER_AZURE_SAS: "sv=demo&sig=demo"
      # Routable address other containers use to reach this worker's RPC port;
      # the default (the listen address, 0.0.0.0:7001) is an unreachable
      # wildcard bind, so the worker refuses to start without this.
      TALON_WORKER_ADVERTISE_ADDR: "worker:7001"
    ports:
      - "8001:8001" # worker admin (metrics/health/status)
    depends_on:
      coordinator:
        condition: service_healthy

  # --- HA profile: etcd backend, 3 coordinators ----------------------------
  etcd:
    image: gcr.io/etcd-development/etcd:v3.5.16
    profiles: ["ha"]
    command:
      - etcd
      - --data-dir=/etcd-data
      - --listen-client-urls=http://0.0.0.0:2379
      - --advertise-client-urls=http://etcd:2379
    healthcheck:
      test: ["CMD", "etcdctl", "--endpoints=http://127.0.0.1:2379", "endpoint", "health"]
      interval: 5s
      timeout: 3s
      retries: 10

  # Three explicit coordinators (distinct node ids) share one etcd, active-active.
  # Only coord-ha-0 publishes its UI; the others are reachable on the compose
  # network. This mirrors the HA topology shown in the management console.
  coord-ha-0: &coord-ha
    image: ghcr.io/milvus-io/talon-coordinator:latest
    build:
      context: .
      dockerfile: deploy/docker/coordinator.Dockerfile
    profiles: ["ha"]
    command: ["--node-id", "coord-ha-0"]
    environment: &coord-ha-env
      TALON_COORDINATOR_STATE_BACKEND: etcd
      TALON_COORDINATOR_HA_ENABLED: "true"
      TALON_COORDINATOR_REPLICAS: "3"
      TALON_COORDINATOR_CLUSTER_ID: demo-ha
      TALON_COORDINATOR_ETCD_ENDPOINTS: etcd:2379
    ports:
      - "8000:8000" # admin API + UI (published from this coordinator)
    depends_on:
      etcd:
        condition: service_healthy
    healthcheck:
      test: ["CMD", "curl", "-fsS", "http://127.0.0.1:8000/healthz"]
      interval: 10s
      timeout: 3s
      retries: 5

  coord-ha-1:
    <<: *coord-ha
    profiles: ["ha"]
    command: ["--node-id", "coord-ha-1"]
    environment: *coord-ha-env
    ports: []

  coord-ha-2:
    <<: *coord-ha
    profiles: ["ha"]
    command: ["--node-id", "coord-ha-2"]
    environment: *coord-ha-env
    ports: []

  worker-ha:
    image: ghcr.io/milvus-io/talon-worker:latest
    build:
      context: .
      dockerfile: deploy/docker/worker.Dockerfile
    profiles: ["ha"]
    command:
      - "--coordinator"
      - "coord-ha-0:7000"
      - "--cluster-id"
      - "demo-ha"
      - "--node-id"
      - "worker-0"
    environment:
      TALON_WORKER_AZURE_ACCOUNT: demo
      TALON_WORKER_AZURE_SAS: "sv=demo&sig=demo"
      # See the default worker's TALON_WORKER_ADVERTISE_ADDR comment above.
      TALON_WORKER_ADVERTISE_ADDR: "worker-ha:7001"
    ports:
      - "8001:8001"
    depends_on:
      coord-ha-0:
        condition: service_healthy