Files
2026-07-13 13:22:34 +08:00

126 lines
3.7 KiB
YAML

volumes:
pgdata:
storage-data:
services:
postgres:
image: postgres:15
container_name: mlflow-postgres
environment:
POSTGRES_USER: ${POSTGRES_USER}
POSTGRES_PASSWORD: ${POSTGRES_PASSWORD}
POSTGRES_DB: ${POSTGRES_DB}
PGPORT: ${PGPORT}
volumes:
- pgdata:/var/lib/postgresql/data
ports:
- ${PGPORT}:${PGPORT}
healthcheck:
test: ["CMD-SHELL", "pg_isready -U ${POSTGRES_USER} -d ${POSTGRES_DB} -p ${PGPORT}"]
interval: 5s
timeout: 3s
retries: 10
storage:
image: rustfs/rustfs:1.0.0-alpha.83
container_name: storage
environment:
RUSTFS_ADDRESS: :9000
RUSTFS_SERVER_DOMAINS: storage:9000
RUSTFS_REGION: ${AWS_DEFAULT_REGION:-us-east-1}
RUSTFS_ACCESS_KEY: ${AWS_ACCESS_KEY_ID:-s3admin}
RUSTFS_SECRET_KEY: ${AWS_SECRET_ACCESS_KEY:-s3admin}
RUSTFS_CONSOLE_ENABLE: ${RUSTFS_CONSOLE_ENABLE:-true}
ports:
- "9000:9000"
- "9001:9001"
volumes:
- storage-data:/data
restart: unless-stopped
healthcheck:
test: ["CMD-SHELL", 'curl -s http://127.0.0.1:9000/health | grep -q ''"status":"ok"''']
interval: 10s
timeout: 5s
retries: 5
start_period: 10s
create-bucket:
image: amazon/aws-cli:2.33.25
container_name: mlflow-create-bucket
depends_on:
storage:
condition: service_healthy
entrypoint: >
/bin/sh -c "
set -e;
echo 'Waiting for S3 gateway getting ready...';
if aws --endpoint-url=${MLFLOW_S3_ENDPOINT_URL} s3api head-bucket --bucket ${S3_BUCKET} 2>/dev/null; then
echo 'Bucket ${S3_BUCKET} already exists. Skipping creation.';
else
echo 'Creating bucket ${S3_BUCKET}...';
aws --endpoint-url=${MLFLOW_S3_ENDPOINT_URL} s3api create-bucket --bucket ${S3_BUCKET} --region ${AWS_DEFAULT_REGION};
fi
"
environment:
AWS_ACCESS_KEY_ID: ${AWS_ACCESS_KEY_ID}
AWS_SECRET_ACCESS_KEY: ${AWS_SECRET_ACCESS_KEY}
AWS_DEFAULT_REGION: ${AWS_DEFAULT_REGION}
AWS_S3_ADDRESSING_STYLE: path
MLFLOW_S3_ENDPOINT_URL: ${MLFLOW_S3_ENDPOINT_URL}
S3_BUCKET: ${S3_BUCKET}
restart: "no"
mlflow:
image: ghcr.io/mlflow/mlflow:${MLFLOW_VERSION}
container_name: mlflow-server
depends_on:
postgres:
condition: service_healthy
storage:
condition: service_healthy
create-bucket:
condition: service_completed_successfully
environment:
# Backend store URI built from vars
MLFLOW_BACKEND_STORE_URI: ${MLFLOW_BACKEND_STORE_URI}
# S3/RustFS settings
MLFLOW_S3_ENDPOINT_URL: ${MLFLOW_S3_ENDPOINT_URL}
MLFLOW_ARTIFACTS_DESTINATION: ${MLFLOW_ARTIFACTS_DESTINATION}
AWS_ACCESS_KEY_ID: ${AWS_ACCESS_KEY_ID}
AWS_SECRET_ACCESS_KEY: ${AWS_SECRET_ACCESS_KEY}
AWS_DEFAULT_REGION: ${AWS_DEFAULT_REGION}
MLFLOW_S3_IGNORE_TLS: "true"
# Server host/port
MLFLOW_HOST: ${MLFLOW_HOST}
MLFLOW_PORT: ${MLFLOW_PORT}
command:
- /bin/bash
- -c
- |
pip install --no-cache-dir psycopg2-binary boto3
mlflow server \
--backend-store-uri "${MLFLOW_BACKEND_STORE_URI}" \
--artifacts-destination "${MLFLOW_ARTIFACTS_DESTINATION}" \
--serve-artifacts \
--host "${MLFLOW_HOST}" \
--port "${MLFLOW_PORT}"
ports:
- "${MLFLOW_PORT}:${MLFLOW_PORT}"
healthcheck:
test:
[
"CMD",
"python",
"-c",
"import urllib.request; urllib.request.urlopen('http://localhost:${MLFLOW_PORT}/health')",
]
interval: 10s
timeout: 5s
retries: 30
networks:
default:
name: mlflow-network