206 lines
8.7 KiB
Bash
206 lines
8.7 KiB
Bash
|
#!/usr/bin/env bash
|
|||
|
|
# =============================================================================
|
|||
|
|
# s-file 客户侧一键安装（在每台服务器上各执行一次）
|
|||
|
|
#
|
|||
|
|
# cd /opt/sfile && ./scripts/install.sh iw # 内网机
|
|||
|
|
# cd /opt/sfile && ./scripts/install.sh ow # 外网机
|
|||
|
|
#
|
|||
|
|
# 首次执行会按角色把交付包里的 .env.iw / .env.ow 拷成 .env；之后改 .env 即可，
|
|||
|
|
# 想回到交付默认值就删掉 .env 重跑。
|
|||
|
|
#
|
|||
|
|
# 做五件事：按角色选配置 -> 环境自检 -> 载入镜像 tar -> 建数据目录 -> 起栈并等到健康
|
|||
|
|
# 幂等：重复执行只重建有差异的容器，不动已初始化好的数据库
|
|||
|
|
# =============================================================================
|
|||
|
|
set -euo pipefail
|
|||
|
|
|
|||
|
|
HERE="$(cd "$(dirname "$0")/.." && pwd)"
|
|||
|
|
cd "$HERE"
|
|||
|
|
|
|||
|
|
ROLE_ARG="${1:-}"
|
|||
|
|
|
|||
|
|
envget() { grep -E "^[[:space:]]*$1=" .env | tail -1 | cut -d= -f2- | tr -d '"' | sed 's/[[:space:]]*$//'; }
|
|||
|
|
|
|||
|
|
step() { printf '\n\033[1;36m==> %s\033[0m\n' "$1"; }
|
|||
|
|
die() { printf '\033[1;31m!!! %s\033[0m\n' "$1" >&2; exit 1; }
|
|||
|
|
|
|||
|
|
case "$ROLE_ARG" in
|
|||
|
|
""|iw|ow) ;;
|
|||
|
|
*) die "未知角色「$ROLE_ARG」，只支持 iw（内网机）或 ow（外网机）" ;;
|
|||
|
|
esac
|
|||
|
|
|
|||
|
|
# ---------- 0. 配置就位 ----------
|
|||
|
|
if [ ! -f .env ]; then
|
|||
|
|
case "$ROLE_ARG" in
|
|||
|
|
iw|ow)
|
|||
|
|
[ -f ".env.$ROLE_ARG" ] || die "缺少 .env.$ROLE_ARG（交付包不完整）"
|
|||
|
|
step "按角色初始化配置：cp .env.$ROLE_ARG .env"
|
|||
|
|
cp ".env.$ROLE_ARG" .env
|
|||
|
|
;;
|
|||
|
|
"") die "用法：./scripts/install.sh iw|ow（iw=内网机，ow=外网机）" ;;
|
|||
|
|
esac
|
|||
|
|
elif [ -n "$ROLE_ARG" ]; then
|
|||
|
|
# 已有 .env 时不让角色参数悄悄覆盖现场改动
|
|||
|
|
CUR_ROLE="$(envget SFILE_ROLE)"
|
|||
|
|
[ "$CUR_ROLE" = "$ROLE_ARG" ] \
|
|||
|
|
|| die ".env 已存在且角色是 $CUR_ROLE，与参数 $ROLE_ARG 不符；确认要换角色请先 rm .env"
|
|||
|
|
fi
|
|||
|
|
|
|||
|
|
# ---------- 0. 前置自检 ----------
|
|||
|
|
step "检查 docker 与 compose"
|
|||
|
|
command -v docker >/dev/null 2>&1 || die "未安装 docker，请先装 Docker Engine（信创系统用厂商源，版本号 >= 20.10）"
|
|||
|
|
docker info >/dev/null 2>&1 || die "docker 守护进程不可用，检查服务是否启动、当前用户是否在 docker 组"
|
|||
|
|
if docker compose version >/dev/null 2>&1; then
|
|||
|
|
DC="docker compose"
|
|||
|
|
elif command -v docker-compose >/dev/null 2>&1; then
|
|||
|
|
DC="docker-compose"
|
|||
|
|
else
|
|||
|
|
die "缺少 compose：安装 docker compose v2 插件，或用 docker-compose 命令"
|
|||
|
|
fi
|
|||
|
|
echo " 使用：$DC"
|
|||
|
|
DOCKER_VER="$(docker version --format '{{.Server.Version}}')"
|
|||
|
|
echo " docker server：$DOCKER_VER"
|
|||
|
|
|
|||
|
|
[ -f .env ] || die "缺少 .env —— 先执行：cp .env.example .env && vi .env"
|
|||
|
|
|
|||
|
|
if grep -q "CHANGE_ME" .env; then
|
|||
|
|
# MeshSafe 的 secret 只有真的启用检测时才必须填，否则客户没接 MeshSafe 也装不了机
|
|||
|
|
if [ "$(envget MESHSAFE_ENABLED)" = "true" ] && grep -qE "^MESHSAFE_SECRET=.*CHANGE_ME" .env; then
|
|||
|
|
die "启用了 MeshSafe 但 MESHSAFE_SECRET 还是占位值：按 MeshSafe 交付单改成白名单里的 secret，或把 MESHSAFE_ENABLED 改成 false"
|
|||
|
|
fi
|
|||
|
|
rest="$(grep -n CHANGE_ME .env | grep -v '^.*MESHSAFE_SECRET=' | cut -d: -f2 | tr '\n' ' ')"
|
|||
|
|
[ -z "$rest" ] || die ".env 里还有 CHANGE_ME 未替换：$rest"
|
|||
|
|
fi
|
|||
|
|
|
|||
|
|
step "检查交付物"
|
|||
|
|
[ -d images ] || die "缺少 images/ 目录（镜像 tar 应与本脚本同级目录的 images/ 下）"
|
|||
|
|
tar_count=$(find images -maxdepth 1 -name '*.tar' | wc -l)
|
|||
|
|
[ "$tar_count" -eq 5 ] || die "images/ 下应有 5 个 .tar（app/web/pg/redis/mq），实际 $tar_count 个：$(ls images)"
|
|||
|
|
for f in sql/01-init-table.sql sql/02-init-data.sql; do
|
|||
|
|
[ -s "$f" ] || die "缺少 $f，参见 sql/README.md"
|
|||
|
|
done
|
|||
|
|
echo " 镜像 tar：5 个，SQL 脚本就位"
|
|||
|
|
|
|||
|
|
# ---------- 1. 校验交付盘 + 镜像架构 ----------
|
|||
|
|
# 650M 的 tar 经 U 盘/内网 scp 拷坏是最常见的现场故障，坏包 load 到一半才报错更难查
|
|||
|
|
if [ -f images/SHA256SUMS.txt ] && [ "${SKIP_CHECKSUM:-0}" != "1" ]; then
|
|||
|
|
step "校验镜像 tar 完整性（sha256，约半分钟）"
|
|||
|
|
( cd images && sha256sum -c SHA256SUMS.txt ) \
|
|||
|
|
|| die "有 tar 与 SHA256SUMS.txt 不一致：交付盘拷坏了，重新拷一份；确认无碍可 SKIP_CHECKSUM=1 跳过"
|
|||
|
|
fi
|
|||
|
|
|
|||
|
|
step "核对镜像架构"
|
|||
|
|
HOST_ARCH="$(uname -m)"
|
|||
|
|
case "$HOST_ARCH" in
|
|||
|
|
x86_64) EXPECT="amd64" ;;
|
|||
|
|
aarch64|arm64) EXPECT="arm64" ;;
|
|||
|
|
*) EXPECT="unknown" ;;
|
|||
|
|
esac
|
|||
|
|
echo " 本机：$HOST_ARCH (期望镜像架构 ${EXPECT})"
|
|||
|
|
|
|||
|
|
# 架构在镜像的 config blob 里，不在 manifest.json —— manifest.json 只给 Config 的路径
|
|||
|
|
tar_arch() {
|
|||
|
|
local t="$1" cfg
|
|||
|
|
cfg="$(tar -xOf "$t" manifest.json 2>/dev/null | tr -d '\n' \
|
|||
|
|
| { grep -oE '"Config":"[^"]+"' || true; } | awk -F'"' 'NR==1{print $4}')"
|
|||
|
|
[ -n "$cfg" ] || { echo ""; return 0; }
|
|||
|
|
tar -xOf "$t" "$cfg" 2>/dev/null | tr -d '\n' \
|
|||
|
|
| { grep -oE '"architecture":"[a-z0-9-]+"' || true; } | awk -F'"' 'NR==1{print $4}'
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
if [ "$EXPECT" != "unknown" ]; then
|
|||
|
|
# app 和 pg 各查一个：一个是自己构建的，一个是基础镜像，架构错了必是出盘时选错平台
|
|||
|
|
for t in images/sfile-app.tar images/postgres.tar; do
|
|||
|
|
[ -f "$t" ] || continue
|
|||
|
|
# 不能用 `| head -1`：pipefail 下 head 提前关管道会让上游 grep 收到 SIGPIPE(141)，脚本静默中断
|
|||
|
|
IMG_ARCH="$(tar_arch "$t")"
|
|||
|
|
if [ -n "${IMG_ARCH:-}" ] && [ "$IMG_ARCH" != "$EXPECT" ]; then
|
|||
|
|
die "$(basename "$t") 是 $IMG_ARCH，本机是 $EXPECT —— 别硬装，回来重出对应架构的包"
|
|||
|
|
fi
|
|||
|
|
printf ' %-18s %s\n' "$(basename "$t")" "${IMG_ARCH:-未识别，继续}"
|
|||
|
|
done
|
|||
|
|
fi
|
|||
|
|
|
|||
|
|
# ---------- 2. 载入镜像 ----------
|
|||
|
|
step "docker load 载入离线镜像"
|
|||
|
|
for t in images/*.tar; do
|
|||
|
|
echo " <- $t"
|
|||
|
|
docker load -i "$t"
|
|||
|
|
done
|
|||
|
|
|
|||
|
|
# compose 用的是 .env 里的 tag，缺一个就会去联网拉，离线机上直接卡死
|
|||
|
|
for img in "$(envget APP_IMAGE)" "$(envget WEB_IMAGE)" "$(envget PG_IMAGE)" \
|
|||
|
|
"$(envget REDIS_IMAGE)" "$(envget MQ_IMAGE)"; do
|
|||
|
|
docker image inspect "$img" >/dev/null 2>&1 \
|
|||
|
|
|| die "载入后仍看不到镜像 $img —— tar 与 .env 的 tag 不一致，别继续，回来看 images/ 和 .env"
|
|||
|
|
echo " ✓ $img"
|
|||
|
|
done
|
|||
|
|
|
|||
|
|
# ---------- 3. 目录 ----------
|
|||
|
|
step "创建数据与摆渡目录"
|
|||
|
|
DATA_ROOT="$(envget SFILE_DATA_ROOT)"
|
|||
|
|
GATE_SEND="$(envget GATE_SEND_DIR)"
|
|||
|
|
GATE_RECV="$(envget GATE_RECV_DIR)"
|
|||
|
|
[ -n "$DATA_ROOT" ] || die ".env 里 SFILE_DATA_ROOT 为空"
|
|||
|
|
|
|||
|
|
mkdir -p "$DATA_ROOT"/{upload,logs,pgdata,redis,mq}
|
|||
|
|
mkdir -p "$GATE_SEND" "$GATE_RECV"
|
|||
|
|
|
|||
|
|
# 数据盘挂载点没起来时直接建目录会落到根分区，先确认一下
|
|||
|
|
if ! mountpoint -q "$DATA_ROOT" 2>/dev/null; then
|
|||
|
|
printf ' \033[1;33m提示：%s 不是独立挂载点，确认容量够（上传+库+日志）\033[0m\n' "$DATA_ROOT"
|
|||
|
|
fi
|
|||
|
|
|
|||
|
|
# pgdata 提前给到容器内 postgres 的 uid=999，避免镜像入口脚本 chown 失败；
|
|||
|
|
# redis/mq 的官方入口脚本自己会改属主，这里只保证目录存在且可写
|
|||
|
|
chown -R 999:999 "$DATA_ROOT/pgdata" 2>/dev/null || true
|
|||
|
|
chmod -R 750 "$DATA_ROOT"
|
|||
|
|
|
|||
|
|
# SELinux（麒麟/统信多为 enforcing）：无 :z 标记的绑定挂载需要放开标签，否则库起不来
|
|||
|
|
if command -v chcon >/dev/null 2>&1; then
|
|||
|
|
chcon -Rt svirt_sandbox_file_t "$DATA_ROOT" "$GATE_SEND" "$GATE_RECV" 2>/dev/null \
|
|||
|
|
|| echo " chcon 失败（未启用 SELinux 可忽略）"
|
|||
|
|
fi
|
|||
|
|
echo " $DATA_ROOT / $GATE_SEND / $GATE_RECV 就位"
|
|||
|
|
|
|||
|
|
# ---------- 4. 起栈 ----------
|
|||
|
|
step "启动服务（首次建库约 1-3 分钟）"
|
|||
|
|
$DC up -d
|
|||
|
|
|
|||
|
|
step "等待健康检查"
|
|||
|
|
ROLE="$(envget SFILE_ROLE)"
|
|||
|
|
for svc in pg redis mq app web; do
|
|||
|
|
name="sfile-${ROLE}-${svc}"
|
|||
|
|
printf ' %-24s ' "$name"
|
|||
|
|
for _ in $(seq 1 60); do
|
|||
|
|
st="$(docker inspect --format '{{if .State.Health}}{{.State.Health.Status}}{{else}}{{.State.Status}}{{end}}' "$name" 2>/dev/null || echo missing)"
|
|||
|
|
case "$st" in
|
|||
|
|
healthy|running) printf '\033[32m%s\033[0m\n' "$st"; break ;;
|
|||
|
|
missing) sleep 2 ;;
|
|||
|
|
*) sleep 5 ;;
|
|||
|
|
esac
|
|||
|
|
done
|
|||
|
|
[ "${st:-}" = "healthy" ] || [ "${st:-}" = "running" ] || {
|
|||
|
|
printf '\033[31m超时（最后状态：%s）\033[0m\n' "${st:-unknown}"
|
|||
|
|
$DC ps
|
|||
|
|
echo " 看日志：docker logs --tail 200 ${name}"
|
|||
|
|
exit 1
|
|||
|
|
}
|
|||
|
|
done
|
|||
|
|
|
|||
|
|
# ---------- 5. 出口信息 ----------
|
|||
|
|
IP="$(hostname -I 2>/dev/null | tr ' ' '\n' | { grep -E '^[0-9]+\.' || true; } | awk 'NR==1{print}')"
|
|||
|
|
PORT="$(envget WEB_PORT)"
|
|||
|
|
step "完成"
|
|||
|
|
cat <<EOF
|
|||
|
|
访问地址： http://${IP:-<本机IP>}:${PORT:-80}
|
|||
|
|
角色： ${ROLE}（传输方向 location=$(envget SFILE_TRANSFER_LOCATION)，站点 $(envget SFILE_SITE_ROOT)）
|
|||
|
|
摆渡： 发出 $(envget GATE_SEND_DIR) / 收到 $(envget GATE_RECV_DIR)
|
|||
|
|
MeshSafe： $(envget MESHSAFE_ENABLED) -> $(envget MESHSAFE_BASE_URL)
|
|||
|
|
|
|||
|
|
常用命令：
|
|||
|
|
$DC ps 看状态
|
|||
|
|
docker logs -f sfile-${ROLE}-app 看后端日志
|
|||
|
|
$DC restart app 改完 .env/config 后重启后端
|
|||
|
|
EOF
|
|||
|
|
echo " 首次登录后立刻改掉 admin 的默认口令。"
|