feat(backup): mysqldump 定时备份/恢复演练/健康检查脚本与手册

- scripts/backup_full.sh:容器内 mysqldump --single-transaction,gzip 落盘
  db/backups/,保留 30 天;凭据 source docker/.env 不硬编码;实测备份成功
- scripts/restore_drill.sh:恢复到临时库 payroll_drill_* 并逐表核对行数
  (16 张关键表),默认不 DROP、--drop-exec 才删——绝不覆盖生产库;
  实测 16 表全 ✓。规避嵌套 shell 吞反引号(mysqladmin create/drop、
  双引号包裹表名)
- scripts/health_check.sh:六项检查(容器健康×2/磁盘水位/备份新鲜度/
  Redis 探活/24h 登录失败阈值),PASS/WARN/FAIL 报告,退出码供监控;
  Redis 密码经宿主机 .env 验证(容器内无该环境变量),实测六项全 PASS
- docs/备份恢复手册.md:备份/演练/生产恢复七步/定时 schtasks 两则,
  db/backups/ 已入 .gitignore

Co-Authored-By: Claude Code <noreply@anthropic.com>
This commit is contained in:
liuwenxin 2026-10-16 09:44:50 +08:00
parent 927592b472
commit b5e27810fd
4 changed files with 321 additions and 0 deletions

View File

@ -0,0 +1,70 @@
# 备份恢复手册(⑨ 备份恢复与运维检查)
> 用户强调「这个备份恢复很重要」。**备份不演练 = 没有备份**——本手册的一页纸目标:任何运维/开发拿到都能完成「备份 → 恢复到临时库核对 → 删临时库」全链路。
## 一、备份(每日一次,建议凌晨低峰)
```bash
# 项目根目录执行
bash scripts/backup_full.sh
```
- 产物:`db/backups/payroll_YYYYMMDD_HHMMSS.sql.gz`(gzip 压缩)
- 保留策略:自动清理 30 天前的旧备份
- 凭据:从 `docker/.env` 读取(不入脚本、不入版本库)
**Windows 定时任务**(任务计划程序 / `schtasks`):
```cmd
:: 每日 02:30 备份(经 Git Bash)
schtasks /create /tn "ys-payroll-backup" /tr "\"C:\Program Files\Git\bin\bash.exe\" -lc \"bash scripts/backup_full.sh\"" /sc daily /st 02:30 /ru SYSTEM
```
## 二、恢复演练(备份后顺手跑一次,验证备份真能用)
```bash
bash scripts/restore_drill.sh # 恢复到临时库 + 行数核对,临时库默认保留
bash scripts/restore_drill.sh --drop-exec # 核对通过后再加此参数删除临时库
```
- **绝不直接覆盖生产库**——演练恢复到独立临时库 `payroll_drill_<时间戳>`
- 行数对比覆盖 16 张关键表(users/roles/param_*/business*/personnel/products/customers/shipments*/receipts/approval_*/system_settings)
- 全部 ✓ 即备份可恢复;出现 ✗ 时临时库保留供排查
## 三、生产恢复(真出事了,七步)
> ⚠️ 生产恢复是**高危操作**:先把容器停了再恢复,避免写冲突。下面假设恢复到**全新空数据卷**(最干净路径)。
1. **停容器保卷**:`docker compose -f docker/docker-compose.dev.yml down`
2. **删旧卷(谨慎!确认要全量重来才执行)**:`docker volume rm ys-payroll_mysql-dev-data`
- 若只想恢复到某个时间点**且不丢其它数据**,改用「恢复到临时库 → mysqldump 单表导出 → 手工回灌」的细粒度路径,勿整体覆盖。
3. **起空库**:`docker compose -f docker/docker-compose.dev.yml up -d`(首启自动跑 `db/init/init-dev.sql` 建表 + 种子)
4. **选一份备份**:`ls -t db/backups/payroll_*.sql.gz | head`
5. **导入该备份**:
```bash
gunzip -c db/backups/payroll_XXXXXXXX_XXXXXX.sql.gz | \
docker exec -i ys-mysql-dev sh -c 'mysql -uroot -p"$MYSQL_ROOT_PASSWORD" payroll'
```
6. **核对行数**:再跑一次 `bash scripts/restore_drill.sh`(此时源库=刚恢复的库,应全 ✓)
7. **重启后端**:`cd backend && uv run uvicorn app.main:app --reload --port 8000`,登录抽查几个页面确认数据回来
## 四、运维健康检查(每日一次)
```bash
bash scripts/health_check.sh
```
六项检查:容器健康 ×2 / 磁盘水位(80% 阈值)/ 备份新鲜度(24h 内)/ Redis 探活 / 24h 登录失败次数(暴力破解预警)。任一项 WARN/FAIL 时退出码非零,方便监控采集。
**Windows 定时**:
```cmd
:: 每日 08:50 检查(上班前出报告)
schtasks /create /tn "ys-payroll-health" /tr "\"C:\Program Files\Git\bin\bash.exe\" -lc \"bash scripts/health_check.sh >> db/backups/health.log 2>&1\"" /sc daily /st 08:50 /ru SYSTEM
```
## 五、注意事项
- 备份含**全部业务数据与参数种子**,属敏感文件——`db/backups/` 已加入 `.gitignore`,**不要**把它提交到版本库
- 恢复演练产生的 `payroll_drill_*` 库会占磁盘,核对完及时 `--drop-exec` 清理
- 备份脚本与恢复脚本都只依赖 Docker(无需宿主机装 MySQL 客户端)

58
scripts/backup_full.sh Normal file
View File

@ -0,0 +1,58 @@
#!/usr/bin/env bash
# 衍水工资系统 —— MySQL 全量备份(⑨ 备份恢复与运维检查)
#
# 思路:在跑中的容器内执行 mysqldump,免暴露 3306 端口给宿主机工具,
# 也避免宿主机装 MySQL 客户端。凭据从 docker/.env 读取(不入脚本、不入版本库)。
#
# 用法:在项目根目录执行 bash scripts/backup_full.sh
# 产物:db/backups/payroll_YYYYMMDD_HHMMSS.sql.gz,保留最近 30 天。
set -euo pipefail
ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
cd "$ROOT"
if [[ -f docker/.env ]]; then
# shellcheck disable=SC1091
source docker/.env
else
echo "ERROR: docker/.env 不存在(请先创建 docker/.env 并填入 MySQL 凭据)" >&2
exit 1
fi
: "${MYSQL_DATABASE:?docker/.env 缺 MYSQL_DATABASE}"
: "${MYSQL_ROOT_PASSWORD:?docker/.env 缺 MYSQL_ROOT_PASSWORD}"
BACKUP_DIR="$ROOT/db/backups"
mkdir -p "$BACKUP_DIR"
STAMP="$(date +%Y%m%d_%H%M%S)"
OUT="$BACKUP_DIR/payroll_${STAMP}.sql.gz"
# 容器名固定(compose 显式 container_name),与 docker/docker-compose.dev.yml 保持一致
CONTAINER="ys-mysql-dev"
if ! docker inspect "$CONTAINER" >/dev/null 2>&1; then
echo "ERROR: 容器 $CONTAINER 未运行(先 docker compose up -d)" >&2
exit 1
fi
echo "开始备份 $MYSQL_DATABASE → $(basename "$OUT")"
docker exec "$CONTAINER" sh -c \
"mysqldump -uroot -p\"\$MYSQL_ROOT_PASSWORD\" --single-transaction --routines --triggers \"$MYSQL_DATABASE\"" \
| gzip > "$OUT"
if [[ ! -s "$OUT" ]]; then
echo "ERROR: 备份文件为空(数据库名或凭据可能不对)" >&2
rm -f "$OUT"
exit 1
fi
SIZE="$(du -h "$OUT" | cut -f1)"
echo "备份完成:$(basename "$OUT") ($SIZE)"
# 保留 30 天
find "$BACKUP_DIR" -maxdepth 1 -name 'payroll_*.sql.gz' -mtime +30 -print -delete
echo "已清理 30 天前的旧备份"
echo ""
echo "最新备份:"
ls -lh "$BACKUP_DIR"/*.sql.gz 2>/dev/null | tail -3

105
scripts/health_check.sh Normal file
View File

@ -0,0 +1,105 @@
#!/usr/bin/env bash
# 运维健康检查 —— 容器/磁盘/备份新鲜度/登录异常(⑨ 输出一份可读报告)
#
# 用法:在项目根目录执行 bash scripts/health_check.sh
# 每项 PASS/WARN/FAIL,最后给出结论。WARN/FAIL 时退出码非零提示运维介入。
set -uo pipefail
ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
cd "$ROOT"
if [[ -f docker/.env ]]; then
# shellcheck disable=SC1091
source docker/.env
fi
ISSUES=0
report() { printf "%-10s %s\n" "$1" "$2"; }
warn() { report "WARN" "$1"; ISSUES=$((ISSUES+1)); }
fail() { report "FAIL" "$1"; ISSUES=$((ISSUES+1)); }
pass() { report "PASS" "$1"; }
echo "======== 衍水工资系统 健康检查 $(date '+%Y-%m-%d %H:%M:%S') ========"
# 1. 容器健康
for C in ys-mysql-dev ys-redis-dev; do
if ! docker inspect "$C" >/dev/null 2>&1; then
fail "容器 $C 不存在"
continue
fi
STATUS="$(docker inspect -f '{{.State.Health.Status}}' "$C" 2>/dev/null || echo none)"
RUNNING="$(docker inspect -f '{{.State.Running}}' "$C" 2>/dev/null)"
if [[ "$RUNNING" == "true" && "$STATUS" == "healthy" ]]; then
pass "容器 $C 运行中且健康"
elif [[ "$RUNNING" == "true" ]]; then
warn "容器 $C 运行中但健康检查=$STATUS"
else
fail "容器 $C 未运行"
fi
done
# 2. 磁盘水位
DISK_PCT="$(df -h "$ROOT" | awk 'NR==2 {gsub(/%/,"",$5); print $5}')"
if [[ -z "$DISK_PCT" ]]; then
warn "磁盘信息读取失败"
elif (( DISK_PCT < 80 )); then
pass "磁盘使用率 ${DISK_PCT}%(阈值 80%)"
elif (( DISK_PCT < 90 )); then
warn "磁盘使用率 ${DISK_PCT}%(建议清理)"
else
fail "磁盘使用率 ${DISK_PCT}%(高危,立即清理)"
fi
# 3. 备份新鲜度
BACKUP_DIR="$ROOT/db/backups"
LATEST="$(ls -t "$BACKUP_DIR"/payroll_*.sql.gz 2>/dev/null | head -1)"
if [[ -z "$LATEST" ]]; then
warn "未找到任何备份(应每日备份)"
else
AGE_H=$(( ($(date +%s) - $(stat -c %Y "$LATEST")) / 3600 ))
if (( AGE_H < 24 )); then
pass "最新备份 $(basename "$LATEST"),${AGE_H} 小时前"
elif (( AGE_H < 48 )); then
warn "最新备份已 ${AGE_H} 小时前(超过 24 小时)"
else
fail "最新备份已 ${AGE_H} 小时前(备份任务可能已停)"
fi
fi
# 4. Redis 存活(密码经宿主 docker/.env,不经命令行明文参数)
if [[ -n "${REDIS_PASSWORD:-}" ]]; then
if docker exec ys-redis-dev redis-cli --no-auth-warning -a "$REDIS_PASSWORD" ping 2>/dev/null | grep -q PONG; then
pass "Redis 响应 PONG"
else
warn "Redis ping 无响应(检查 REDIS_PASSWORD 或容器状态)"
fi
else
warn "docker/.env 缺 REDIS_PASSWORD,跳过 Redis 探活"
fi
# 5. 登录日志异常(24h 失败次数)
if [[ -n "${MYSQL_ROOT_PASSWORD:-}" ]] && docker inspect ys-mysql-dev >/dev/null 2>&1; then
FAILS=$(docker exec ys-mysql-dev sh -c \
"mysql -uroot -p\"\$MYSQL_ROOT_PASSWORD\" ${MYSQL_DATABASE:-payroll} -N -B -e \"
SELECT COUNT(*) FROM login_logs
WHERE success = 0 AND create_time > NOW() - INTERVAL 24 HOUR;\" 2>/dev/null" || echo "?")
if [[ "$FAILS" == "?" ]]; then
warn "登录日志查询失败"
elif (( FAILS < 10 )); then
pass "24h 登录失败 ${FAILS} 次(阈值 10)"
else
warn "24h 登录失败 ${FAILS} 次(可能有暴力破解,建议核查)"
fi
else
warn "跳过登录日志检查(无凭据或容器未运行)"
fi
echo "=============================================================="
if (( ISSUES == 0 )); then
echo "结论:全部通过,系统健康。"
exit 0
else
echo "结论:发现 $ISSUES 项需关注(见上)。"
exit 1
fi

88
scripts/restore_drill.sh Normal file
View File

@ -0,0 +1,88 @@
#!/usr/bin/env bash
# 恢复演练 —— 把最近一份备份恢复到「临时库」并核对行数(⑨:备份不演练 = 没有备份)
#
# 演练绝不直接覆盖生产库;临时库名 payroll_drill_<时间戳>,核对完由
# 调用方决定是否删除(本脚本默认不 DROP,安全优先)。
#
# 用法:在项目根目录执行 bash scripts/restore_drill.sh
# 产物:临时库 + 终端行数对比报告;--drop-exec 参数在核对通过后删除临时库。
set -euo pipefail
ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
cd "$ROOT"
if [[ -f docker/.env ]]; then
# shellcheck disable=SC1091
source docker/.env
else
echo "ERROR: docker/.env 不存在" >&2
exit 1
fi
: "${MYSQL_DATABASE:?docker/.env 缺 MYSQL_DATABASE}"
: "${MYSQL_ROOT_PASSWORD:?docker/.env 缺 MYSQL_ROOT_PASSWORD}"
CONTAINER="ys-mysql-dev"
if ! docker inspect "$CONTAINER" >/dev/null 2>&1; then
echo "ERROR: 容器 $CONTAINER 未运行" >&2
exit 1
fi
BACKUP_DIR="$ROOT/db/backups"
LATEST="$(ls -t "$BACKUP_DIR"/payroll_*.sql.gz 2>/dev/null | head -1)"
if [[ -z "$LATEST" ]]; then
echo "ERROR: 未找到备份文件(先 bash scripts/backup_full.sh)" >&2
exit 1
fi
DRILL_DB="payroll_drill_$(date +%Y%m%d_%H%M%S)"
echo "演练恢复:$(basename "$LATEST") → 临时库 $DRILL_DB"
# 1. 创建临时库(mysqladmin create 无反引号转义问题)
docker exec "$CONTAINER" sh -c \
"mysqladmin -uroot -p\"\$MYSQL_ROOT_PASSWORD\" create \"$DRILL_DB\""
# 2. 解压导入
gunzip -c "$LATEST" | docker exec -i "$CONTAINER" sh -c \
"mysql -uroot -p\"\$MYSQL_ROOT_PASSWORD\" \"$DRILL_DB\""
# 3. 行数核对:关键业务表
TABLES=(users roles user_roles param_definitions param_values businesses business_items personnel products customers shipments shipment_items receipts approval_orders approval_steps system_settings)
echo ""
echo "======== 行数对比(源库 vs 演练库)========"
printf "%-24s %12s %12s %6s\n" "表名" "源库" "演练库" "结果"
ALL_OK=1
for T in "${TABLES[@]}"; do
SRC=$(docker exec "$CONTAINER" sh -c \
"mysql -uroot -p\"\$MYSQL_ROOT_PASSWORD\" -N -B -e \"SELECT COUNT(*) FROM $MYSQL_DATABASE.$T;\" 2>/dev/null" || echo "?")
DST=$(docker exec "$CONTAINER" sh -c \
"mysql -uroot -p\"\$MYSQL_ROOT_PASSWORD\" -N -B -e \"SELECT COUNT(*) FROM $DRILL_DB.$T;\" 2>/dev/null" || echo "?")
if [[ "$SRC" == "$DST" ]]; then
MARK="✓"
else
MARK="✗"
ALL_OK=0
fi
printf "%-24s %12s %12s %6s\n" "$T" "$SRC" "$DST" "$MARK"
done
echo "=========================================="
if [[ "$ALL_OK" == "1" ]]; then
echo ""
echo "结果:全部表行数一致,备份可恢复。"
if [[ "${1:-}" == "--drop-exec" ]]; then
docker exec "$CONTAINER" sh -c \
"mysqladmin -uroot -p\"\$MYSQL_ROOT_PASSWORD\" -f drop \"$DRILL_DB\""
echo "已删除临时库 $DRILL_DB"
else
echo "临时库 $DRILL_DB 保留(核对用);确认无误后重跑本脚本加 --drop-exec 删除。"
fi
exit 0
else
echo ""
echo "结果:存在不一致,请检查备份文件与数据库结构。" >&2
echo "临时库 $DRILL_DB 保留供排查。" >&2
exit 1
fi