06fb35b984
端到端延迟里有一半是轮询空等：收发两侧各等一次，1 分钟周期下平均空等 60 秒、最坏 120 秒，这是「一页文件也要 2 分钟」的主因，而真正干活不到 2 秒。 改成 5 秒是安全的：空目录一轮只有两次 Files.list + 过滤，开销可忽略； Spring 的 cron 任务要等上一轮跑完才排下一轮，同一实例内不会重叠。 副作用是损坏包被更频繁重试（失败包当前不删除也不改名，mtime 最老所以 每轮排第一），彻底解决靠方案 5.2 的 .error 改名让位。 部署说明第 10 条原来写「不要改成秒级」，与新默认值矛盾，一并纠正。
190 lines
8.0 KiB
YAML
190 lines
8.0 KiB
YAML
# =============================================================================
|
||
# s-file 信创文件安全传输系统 —— 客户侧单台服务器部署（内网机 / 外网机通用同一份）
|
||
#
|
||
# 一台服务器一套本栈；两台机器靠光闸摆渡目录互通，彼此之间不开任何网络端口。
|
||
# 所有随角色变化的取值都在同目录 .env 里，本文件现场不用改。
|
||
#
|
||
# ./scripts/install.sh iw # 内网机（第一次会把 .env.iw 拷成 .env）
|
||
# ./scripts/install.sh ow # 外网机
|
||
#
|
||
# 手动起也行：cp .env.iw .env && vi .env -> docker load -i images/*.tar -> docker compose up -d
|
||
#
|
||
# 只发布前端端口；后端 / PostgreSQL / Redis / RabbitMQ 仅在容器网络内可达。
|
||
# =============================================================================
|
||
|
||
name: sfile-${SFILE_ROLE:?缺少 SFILE_ROLE（内网机 iw / 外网机 ow）}
|
||
|
||
services:
|
||
web:
|
||
image: ${WEB_IMAGE:?缺少 WEB_IMAGE}
|
||
container_name: sfile-${SFILE_ROLE}-web
|
||
restart: unless-stopped
|
||
ports:
|
||
- "${WEB_HOST:-0.0.0.0}:${WEB_PORT:-80}:80"
|
||
environment:
|
||
TZ: ${TZ:-Asia/Shanghai}
|
||
# nginx 官方镜像的 /etc/nginx/templates/*.template 会按 NGINX_ENVSUBST_FILTER
|
||
# 过滤后 envsubst 生成 conf；限定只替换 SFILE_ 开头的变量，避免动到 $uri/$http_host
|
||
NGINX_ENVSUBST_FILTER: ^SFILE_
|
||
SFILE_SITE_ROOT: ${SFILE_SITE_ROOT:?缺少 SFILE_SITE_ROOT}
|
||
SFILE_APP_UPSTREAM: app
|
||
volumes:
|
||
- ./nginx/default.conf.template:/etc/nginx/templates/default.conf.template:ro
|
||
depends_on:
|
||
app:
|
||
condition: service_healthy
|
||
healthcheck:
|
||
test: ["CMD-SHELL", "curl -sf -o /dev/null http://127.0.0.1:80/ || exit 1"]
|
||
interval: 30s
|
||
timeout: 5s
|
||
retries: 3
|
||
start_period: 20s
|
||
networks: [sfile]
|
||
|
||
app:
|
||
image: ${APP_IMAGE:?缺少 APP_IMAGE}
|
||
container_name: sfile-${SFILE_ROLE}-app
|
||
restart: unless-stopped
|
||
stop_grace_period: 30s
|
||
environment:
|
||
TZ: ${TZ:-Asia/Shanghai}
|
||
# 发号器 workerId 按「容器 IP」取键，取不到会回退 99 而 99 > 31 直接启动失败，
|
||
# 因此这里锁死 app 的容器 IP，并把同一个键写进 JVM 参数，两处永远一致
|
||
JAVA_OPTS: "${APP_JAVA_OPTS:--Xms1g -Xmx3g -Dfile.encoding=UTF-8} -Dcode.snowflake.workid.${APP_NODE_IP:?缺少 APP_NODE_IP}=${APP_WORK_ID:?缺少 APP_WORK_ID}"
|
||
SPRING_CONFIG_ADDITIONAL_LOCATION: file:/config/application.yml
|
||
|
||
SPRING_DATASOURCE_URL: jdbc:postgresql://pg:5432/${POSTGRES_DB:?缺少 POSTGRES_DB}
|
||
SPRING_DATASOURCE_USERNAME: ${POSTGRES_USER:?缺少 POSTGRES_USER}
|
||
SPRING_DATASOURCE_PASSWORD: ${POSTGRES_PASSWORD:?缺少 POSTGRES_PASSWORD}
|
||
|
||
SPRING_REDIS_HOST: redis
|
||
SPRING_REDIS_PORT: 6379
|
||
|
||
SPRING_RABBITMQ_HOST: mq
|
||
SPRING_RABBITMQ_PORT: 5672
|
||
SPRING_RABBITMQ_USERNAME: ${MQ_USER:?缺少 MQ_USER}
|
||
SPRING_RABBITMQ_PASSWORD: ${MQ_PASSWORD:?缺少 MQ_PASSWORD}
|
||
# 两侧交换机/广播 ID 必须不同，防止有人把两套栈并到一台机器时串消息
|
||
RABBITMQ_CUSTOM_TOPIC_EXCHANGE_NAME: s-file-${SFILE_ROLE}-topic
|
||
RABBITMQ_CUSTOM_DELAYED_EXCHANGE_NAME: s-file-${SFILE_ROLE}-delayed
|
||
RABBITMQ_CUSTOM_BROADCAST_ID: s-file-${SFILE_ROLE}-broadcast
|
||
|
||
SECURITY_JWT_SECRET: ${JWT_SECRET:?缺少 JWT_SECRET}
|
||
USER_DEFAULT_PASSWORD: ${USER_DEFAULT_PASSWORD:?缺少 USER_DEFAULT_PASSWORD}
|
||
|
||
# 内网机 1 / 外网机 2，写反了摆渡方向就反了
|
||
FILE_TRANSFER_LOCATION: "${SFILE_TRANSFER_LOCATION:?缺少 SFILE_TRANSFER_LOCATION}"
|
||
FILE_TRANSFER_SEND_DIR: /data/transfer/send
|
||
FILE_TRANSFER_RECEIVE_DIR: /data/transfer/receive
|
||
FILE_TRANSFER_MONITOR_CRON: "${FILE_TRANSFER_MONITOR_CRON:-0/5 * * * * ?}"
|
||
|
||
MESHSAFE_ENABLED: ${MESHSAFE_ENABLED:-true}
|
||
MESHSAFE_BASE_URL: ${MESHSAFE_BASE_URL:-}
|
||
MESHSAFE_CLIENT_ID: ${MESHSAFE_CLIENT_ID:-}
|
||
MESHSAFE_SECRET: ${MESHSAFE_SECRET:-}
|
||
MESHSAFE_CHANNEL: ${MESHSAFE_CHANNEL:-sfile-${SFILE_ROLE}}
|
||
MESHSAFE_FAIL_OPEN: ${MESHSAFE_FAIL_OPEN:-true}
|
||
MESHSAFE_TIMEOUT_SECONDS: ${MESHSAFE_TIMEOUT_SECONDS:-120}
|
||
MESHSAFE_POLL_TIMEOUT_MINUTES: ${MESHSAFE_POLL_TIMEOUT_MINUTES:-60}
|
||
MESHSAFE_POLL_CRON: ${MESHSAFE_POLL_CRON:-0/10 * * * * ?}
|
||
volumes:
|
||
- ./config/application.yml:/config/application.yml:ro
|
||
- ${SFILE_DATA_ROOT:?缺少 SFILE_DATA_ROOT}/upload:/data/upload
|
||
- ${SFILE_DATA_ROOT}/logs:/app/logs
|
||
# 摆渡目录：容器内固定路径，宿主机侧由光闸设备负责单向搬运
|
||
- ${GATE_SEND_DIR:?缺少 GATE_SEND_DIR}:/data/transfer/send
|
||
- ${GATE_RECV_DIR:?缺少 GATE_RECV_DIR}:/data/transfer/receive
|
||
# MeshSafe 与应用同机部署时靠它从容器内访问宿主机端口；
|
||
# Docker < 20.10 不支持 host-gateway，或 MeshSafe 在独立服务器时，删掉这两行即可
|
||
extra_hosts:
|
||
- "host.docker.internal:host-gateway"
|
||
depends_on:
|
||
pg:
|
||
condition: service_healthy
|
||
redis:
|
||
condition: service_healthy
|
||
mq:
|
||
condition: service_healthy
|
||
healthcheck:
|
||
test: ["CMD-SHELL", "curl -s --max-time 5 -o /dev/null http://127.0.0.1:8080/ || exit 1"]
|
||
interval: 30s
|
||
timeout: 8s
|
||
retries: 5
|
||
# 首次建库要跑全量 DDL，冷启动给足时间
|
||
start_period: 180s
|
||
networks:
|
||
sfile:
|
||
ipv4_address: ${APP_NODE_IP}
|
||
|
||
pg:
|
||
image: ${PG_IMAGE:?缺少 PG_IMAGE}
|
||
container_name: sfile-${SFILE_ROLE}-pg
|
||
restart: unless-stopped
|
||
stop_grace_period: 60s
|
||
environment:
|
||
TZ: ${TZ:-Asia/Shanghai}
|
||
POSTGRES_DB: ${POSTGRES_DB:?缺少 POSTGRES_DB}
|
||
POSTGRES_USER: ${POSTGRES_USER:?缺少 POSTGRES_USER}
|
||
POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:?缺少 POSTGRES_PASSWORD}
|
||
# 空库首次启动才执行 initdb.d，已有数据卷时会被忽略
|
||
POSTGRES_INITDB_ARGS: "--encoding=UTF8 --lc-collate=C --lc-ctype=C"
|
||
volumes:
|
||
# postgres 18+ 数据目录挂载点是 /var/lib/postgresql，15- 是 .../data，见 .env 的 PG_DATA_DIR
|
||
- ${SFILE_DATA_ROOT}/pgdata:${PG_DATA_DIR:-/var/lib/postgresql}
|
||
- ./sql/01-init-table.sql:/docker-entrypoint-initdb.d/01-init-table.sql:ro
|
||
- ./sql/02-init-data.sql:/docker-entrypoint-initdb.d/02-init-data.sql:ro
|
||
healthcheck:
|
||
# 必须强制走 TCP(-h 127.0.0.1)：首次 initdb 期间镜像会临时起一个只听 unix socket
|
||
# 的实例，此时用默认 socket 探测也会返回成功，导致 app 抢跑连库失败
|
||
test: ["CMD-SHELL", "pg_isready -h 127.0.0.1 -U ${POSTGRES_USER} -d ${POSTGRES_DB}"]
|
||
interval: 10s
|
||
timeout: 5s
|
||
retries: 10
|
||
start_period: 60s
|
||
networks: [sfile]
|
||
|
||
redis:
|
||
image: ${REDIS_IMAGE:?缺少 REDIS_IMAGE}
|
||
container_name: sfile-${SFILE_ROLE}-redis
|
||
restart: unless-stopped
|
||
command: ["redis-server", "--appendonly", "yes", "--maxmemory", "${REDIS_MAXMEMORY:-512mb}", "--maxmemory-policy", "allkeys-lru"]
|
||
environment:
|
||
TZ: ${TZ:-Asia/Shanghai}
|
||
volumes:
|
||
- ${SFILE_DATA_ROOT}/redis:/data
|
||
healthcheck:
|
||
test: ["CMD", "redis-cli", "ping"]
|
||
interval: 10s
|
||
timeout: 3s
|
||
retries: 5
|
||
networks: [sfile]
|
||
|
||
mq:
|
||
image: ${MQ_IMAGE:?缺少 MQ_IMAGE}
|
||
container_name: sfile-${SFILE_ROLE}-mq
|
||
restart: unless-stopped
|
||
stop_grace_period: 30s
|
||
environment:
|
||
TZ: ${TZ:-Asia/Shanghai}
|
||
RABBITMQ_DEFAULT_USER: ${MQ_USER:?缺少 MQ_USER}
|
||
RABBITMQ_DEFAULT_PASS: ${MQ_PASSWORD:?缺少 MQ_PASSWORD}
|
||
volumes:
|
||
- ${SFILE_DATA_ROOT}/mq:/var/lib/rabbitmq
|
||
healthcheck:
|
||
test: ["CMD-SHELL", "rabbitmq-diagnostics -q ping"]
|
||
interval: 20s
|
||
timeout: 10s
|
||
retries: 10
|
||
start_period: 60s
|
||
networks: [sfile]
|
||
|
||
networks:
|
||
sfile:
|
||
driver: bridge
|
||
ipam:
|
||
config:
|
||
# 锁 app 容器 IP 是为了发号器 workerId 取键；subnet/网关/app IP 三个值必须同网段。
|
||
# 若与客户机器上已有 docker 网段冲突，整段换掉（如 172.31.88.0/24）即可
|
||
- subnet: ${APP_SUBNET:?缺少 APP_SUBNET}
|
||
gateway: ${APP_GATEWAY:?缺少 APP_GATEWAY}
|