Compare commits
2 Commits
d554d7be83
...
cacc686e9a
| Author | SHA1 | Date | |
|---|---|---|---|
|
cacc686e9a | ||
|
|
7bf4666cdb |
@ -220,6 +220,7 @@ HEALTHCHECK --interval=10s --timeout=5s --start-period=15s --retries=3 \
|
||||
ENTRYPOINT ["/app/statisticservice"]
|
||||
|
||||
# ---- Runtime Stage: NotificationService (通知 + uniCloud 推送) ----
|
||||
# health 端点路径是 /health(不是 /healthz),见 backend/pkg/health/health.go:29
|
||||
FROM --platform=linux/amd64 alpine:3.19 AS notificationservice
|
||||
|
||||
RUN apk add --no-cache ca-certificates tzdata
|
||||
@ -230,11 +231,14 @@ COPY --from=builder /tmp/notificationservice /app/notificationservice
|
||||
EXPOSE 20010
|
||||
|
||||
HEALTHCHECK --interval=10s --timeout=5s --start-period=15s --retries=3 \
|
||||
CMD wget --no-verbose --tries=1 --spider http://localhost:21010/healthz || exit 1
|
||||
CMD wget --no-verbose --tries=1 --spider http://localhost:21010/health || exit 1
|
||||
|
||||
ENTRYPOINT ["/app/notificationservice"]
|
||||
|
||||
# ---- Runtime Stage: ModerationService (举报 / 反馈 / 自动隐藏) ----
|
||||
# moderationService main.go 没启 health server(只暴露 Dubbo Triple 协议),
|
||||
# image-level HEALTHCHECK NONE 防 rebuild 后回退到错误的 20011 HEALTHCHECK。
|
||||
# 运行时 compose 用 healthcheck: test: ["NONE"] 显式覆盖,这里 NONE 是双保险。
|
||||
FROM --platform=linux/amd64 alpine:3.19 AS moderationservice
|
||||
|
||||
RUN apk add --no-cache ca-certificates tzdata
|
||||
@ -244,8 +248,7 @@ COPY --from=builder /tmp/moderationservice /app/moderationservice
|
||||
|
||||
EXPOSE 20011
|
||||
|
||||
HEALTHCHECK --interval=10s --timeout=5s --start-period=15s --retries=3 \
|
||||
CMD wget --no-verbose --tries=1 --spider http://localhost:20011 || exit 1
|
||||
HEALTHCHECK NONE
|
||||
|
||||
ENTRYPOINT ["/app/moderationservice"]
|
||||
|
||||
|
||||
@ -116,25 +116,30 @@ SSH_KEY_PATH="$HOME/.ssh/id_rsa" # SSH 密钥路径，默认使用 ~/.ssh/id_r
|
||||
|
||||
# ==================== SSH 别名 ====================
|
||||
# 优先使用 SSH 密钥，如果失败则使用密码
|
||||
# SSH 加 ServerAliveInterval/ServerAliveCountMax 防止长任务时死链(被 NAT/firewall 静默断开)
|
||||
# 又不报 EOF,本地 ssh 进程一直挂着等 server-side stdout 关闭。
|
||||
# 历史教训:2026-07-21 ./deploy.sh restart 起了 3 个容器后 docker-compose 异常退出,
|
||||
# 但 stderr/stdout 没回传,本地 ssh 客户端挂着 25 分钟才被用户手动 kill。
|
||||
# 30s 一次心跳,3 次 (90s) 没回应就判死链,SSH 进程会主动退出。
|
||||
ssh_cmd() {
|
||||
if [ -n "$SERVER_PASSWORD" ]; then
|
||||
sshpass -p "$SERVER_PASSWORD" ssh -o StrictHostKeyChecking=no -p "$SERVER_PORT" "$SERVER_USER@$SERVER_HOST" "$@"
|
||||
sshpass -p "$SERVER_PASSWORD" ssh -o StrictHostKeyChecking=no -o ServerAliveInterval=30 -o ServerAliveCountMax=3 -p "$SERVER_PORT" "$SERVER_USER@$SERVER_HOST" "$@"
|
||||
else
|
||||
ssh -o StrictHostKeyChecking=no -p "$SERVER_PORT" "$SERVER_USER@$SERVER_HOST" "$@"
|
||||
ssh -o StrictHostKeyChecking=no -o ServerAliveInterval=30 -o ServerAliveCountMax=3 -p "$SERVER_PORT" "$SERVER_USER@$SERVER_HOST" "$@"
|
||||
fi
|
||||
}
|
||||
ssh_cmd_batch() {
|
||||
if [ -n "$SERVER_PASSWORD" ]; then
|
||||
sshpass -p "$SERVER_PASSWORD" ssh -o StrictHostKeyChecking=no -p "$SERVER_PORT" "$SERVER_USER@$SERVER_HOST" "$@"
|
||||
sshpass -p "$SERVER_PASSWORD" ssh -o StrictHostKeyChecking=no -o ServerAliveInterval=30 -o ServerAliveCountMax=3 -p "$SERVER_PORT" "$SERVER_USER@$SERVER_HOST" "$@"
|
||||
else
|
||||
ssh -o StrictHostKeyChecking=no -p "$SERVER_PORT" "$SERVER_USER@$SERVER_HOST" "$@"
|
||||
ssh -o StrictHostKeyChecking=no -o ServerAliveInterval=30 -o ServerAliveCountMax=3 -p "$SERVER_PORT" "$SERVER_USER@$SERVER_HOST" "$@"
|
||||
fi
|
||||
}
|
||||
scp_cmd() {
|
||||
if [ -n "$SERVER_PASSWORD" ]; then
|
||||
sshpass -p "$SERVER_PASSWORD" scp -o StrictHostKeyChecking=no -P "$SERVER_PORT" "$@"
|
||||
sshpass -p "$SERVER_PASSWORD" scp -o StrictHostKeyChecking=no -o ServerAliveInterval=30 -o ServerAliveCountMax=3 -P "$SERVER_PORT" "$@"
|
||||
else
|
||||
scp -o StrictHostKeyChecking=no -P "$SERVER_PORT" "$@"
|
||||
scp -o StrictHostKeyChecking=no -o ServerAliveInterval=30 -o ServerAliveCountMax=3 -P "$SERVER_PORT" "$@"
|
||||
fi
|
||||
}
|
||||
|
||||
@ -702,9 +707,24 @@ main() {
|
||||
exit 1
|
||||
fi
|
||||
|
||||
# ⚠️ 不要 `down && up` 两步走:
|
||||
# 1) `up -d --force-recreate` 本身就会 stop+create,先 down 浪费 5-10s
|
||||
# 2) 先 down 会一次性把 postgres/redis 也停了,但这俩根本没 env_file,
|
||||
# 无理由重建,白增 cascade 时间和数据卷 IO 风险
|
||||
# ⚠️ 不要 `up --wait`:会等**所有**服务 healthcheck 变 healthy 才返回。
|
||||
# 但本 compose 里:
|
||||
# - `topfans-oss-cors-init` 是 `restart: "no"` 的一次性任务,跑完 exited(0),
|
||||
# `--wait` 把它当 hang 永远等变 running → 卡死
|
||||
# - `topfans-moderationservice` healthcheck URL 端口写错(20011 → 应 21011)
|
||||
# - `topfans-notificationservice` healthcheck 路径写错(/healthz → 服务没暴露)
|
||||
# 这俩历史遗留 bug 让 `--wait` 永远不退出。
|
||||
# ⚠️ 不要 `sleep 10`:固定 sleep 不准,等不到 healthcheck 也白等。
|
||||
# ✅ 改成 polling gateway healthy:gateway 是 Layer 6 末梢,depends_on 所有
|
||||
# 业务服务 + redis,它的 healthcheck 通过 = 入口可对外服务。
|
||||
# moderation/notification 是 internal healthcheck 不影响 gateway,放过即可。
|
||||
# 给 180s timeout 上限防极端情况。
|
||||
print_step "🔄 重启服务"
|
||||
ssh_cmd "cd ${SERVER_PATH} && docker-compose -f docker-compose.prod.yml down && docker-compose -f docker-compose.prod.yml up -d --force-recreate"
|
||||
sleep 10
|
||||
ssh_cmd "cd ${SERVER_PATH} && docker-compose -f docker-compose.prod.yml up -d --force-recreate && timeout 180 bash -c 'while [ \"\$(docker inspect --format={{.State.Health.Status}} topfans-gateway 2>/dev/null)\" != \"healthy\" ]; do sleep 3; done && echo gateway-healthy'"
|
||||
print_msg "$GREEN" "✅ 服务重启完成"
|
||||
;;
|
||||
|
||||
|
||||
@ -523,7 +523,7 @@ services:
|
||||
expose:
|
||||
- "20010"
|
||||
healthcheck:
|
||||
test: ["CMD-SHELL", "wget --no-verbose --tries=1 --spider http://localhost:21010/healthz || exit 1"]
|
||||
test: ["CMD-SHELL", "wget --no-verbose --tries=1 --spider http://localhost:21010/health || exit 1"]
|
||||
<<: *healthcheck
|
||||
deploy:
|
||||
resources:
|
||||
@ -536,6 +536,12 @@ services:
|
||||
|
||||
# 内容审核服务(举报 / 反馈 / 自动隐藏)
|
||||
# 依赖 user / asset / notification Dubbo 服务,以及 Postgres + Redis
|
||||
# ⚠️ healthcheck: test: ["NONE"]:moderationService/main.go 没启 health server
|
||||
# (只暴露 Dubbo Triple 协议在 20011),加 healthcheck 会永远 unhealthy。
|
||||
# Dockerfile.services 的 HEALTHCHECK 指向 20011 也是错的(指向 dubbo 端口),
|
||||
# 必须用 test: NONE 在 compose 层显式覆盖 image 的 HEALTHCHECK。
|
||||
# gateway depends_on moderationservice: service_started(不依赖 healthy),
|
||||
# 所以不影响入口。健康状态靠 gateway 间接覆盖。
|
||||
moderationservice:
|
||||
image: topfans/moderationservice:latest
|
||||
build:
|
||||
@ -572,8 +578,7 @@ services:
|
||||
expose:
|
||||
- "20011"
|
||||
healthcheck:
|
||||
test: ["CMD-SHELL", "wget --no-verbose --tries=1 --spider http://localhost:20011 || exit 1"]
|
||||
<<: *healthcheck
|
||||
test: ["NONE"]
|
||||
deploy:
|
||||
resources:
|
||||
limits:
|
||||
|
||||
Loading…
Reference in New Issue
Block a user