fix: check-alert v3 修复钉钉推送断档 P0 假阳性

根因:sqlite_exec 吞掉 stderr(2>/dev/null),查询失败返回空串,
0 默认为 0 导致误报

修复:业务层取数显式调用 sqlite3 + 检查退出码 + 正则校验,
查询失败跳过检测并 warn 日志
This commit is contained in:
oa-monitor committed 2026-11-04 19:41:18 +08:00
1 parent 7ab75364b1
commit 20fa48fb34
17 files changed
+1747 -26

No files matched your search

+120
View File
@@ -0,0 +1,120 @@
#!/usr/bin/env bash
# ============================================================================
# OA Monitor · 告警检测(每 1 分钟)
# 遍历所有指标,超过阈值触发钉钉告警
# 告警冷却由 common.sh 的 alert_cooldown 表处理(10 分钟同类型不重复)
# ============================================================================
set -u
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
source "${SCRIPT_DIR}/../lib/common.sh"
NOW=$(date +%s)
log_info "[ALERT] ========== 告警检测开始 =========="
# ============================================================================
# 1. DB 层检测
# ============================================================================
# Threads_running 过高
TR=$(sqlite_exec "SELECT threads_running FROM db_metrics ORDER BY ts DESC LIMIT 1;")
if [[ -n "${TR}" ]] && [[ ${TR} -gt ${ALERT_THREADS_RUNNING_HIGH} ]]; then
BODY="**Threads_running: ${TR}**(阈值 ${ALERT_THREADS_RUNNING_HIGH})\n\n活跃连接:$(sqlite_exec 'SELECT threads_connected FROM db_metrics ORDER BY ts DESC LIMIT 1;')\n活跃空闲:$(sqlite_exec 'SELECT threads_idle FROM db_metrics ORDER BY ts DESC LIMIT 1;')"
dingtalk_alert "P1" "DB 层 Threads_running 过高" "${BODY}"
sqlite_exec "INSERT INTO alerts (ts, level, category, title, detail) VALUES (${NOW}, 'P1', 'db', 'Threads_running 过高', 'threads_running=${TR}');"
fi
# 长查询(processlist 中 Time >= ALERT_SLOW_QUERY_SEC 的)
LONG_QUERIES=$(sqlite_exec "SELECT pid, time_sec, user, state, substr(info,1,100) FROM db_processlist WHERE ts > ${NOW}-120 AND time_sec >= ${ALERT_SLOW_QUERY_SEC} ORDER BY time_sec DESC LIMIT 5;")
if [[ -n "${LONG_QUERIES}" ]]; then
COUNT=$(echo "${LONG_QUERIES}" | wc -l | tr -d ' ')
DETAIL=""
while IFS='|' read -r pid time user state info; do
DETAIL="${DETAIL}\n- PID \`${pid}\` Time **${time}s** User \`${user}\` State \`${state}\`\n \`\`\`sql\n ${info}\n \`\`\`"
done <<< "${LONG_QUERIES}"
BODY="**检测到 ${COUNT} 条长查询**(阈值 ${ALERT_SLOW_QUERY_SEC}s)\n\n${DETAIL}\n\n> ⚠️ 长查询会阻塞消息推送线程,建议优先处理"
dingtalk_alert "P0" "DB 层长查询阻塞(${COUNT} 条)" "${BODY}"
sqlite_exec "INSERT INTO alerts (ts, level, category, title, detail) VALUES (${NOW}, 'P0', 'db', '长查询阻塞', '${COUNT} 条 > ${ALERT_SLOW_QUERY_SEC}s');"
fi
# ============================================================================
# 2. HTTP 端点检测
# ============================================================================
# 最近 3 次的失败次数
FAIL_COUNT=0
FAIL_DETAIL=""
for endpoint in "${OA_ENDPOINTS[@]}"; do
FAIL_HERE=$(sqlite_exec "SELECT COUNT(*) FROM http_metrics WHERE endpoint='${endpoint}' AND ts > ${NOW}-120 AND healthy=0;")
FAIL_HERE=${FAIL_HERE:-0}
if [[ ${FAIL_HERE} -ge ${ALERT_HTTP_FAIL_COUNT} ]]; then
FAIL_COUNT=$((FAIL_COUNT + 1))
LAST_STATUS=$(sqlite_exec "SELECT status_code FROM http_metrics WHERE endpoint='${endpoint}' ORDER BY ts DESC LIMIT 1;")
FAIL_DETAIL="${FAIL_DETAIL}\n- \`${endpoint}\` 最近 ${FAIL_HERE} 次失败,最后状态码 ${LAST_STATUS}"
fi
done
if [[ ${FAIL_COUNT} -gt 0 ]]; then
BODY="**${FAIL_COUNT} 个 OA 端点连续失败**(阈值 ${ALERT_HTTP_FAIL_COUNT} 次)\n\n${FAIL_DETAIL}\n\n> 请检查 OA 应用服务器 39.99.236.138 + nginx + e-cology 服务状态"
dingtalk_alert "P1" "OA HTTP 端点故障" "${BODY}"
sqlite_exec "INSERT INTO alerts (ts, level, category, title, detail) VALUES (${NOW}, 'P1', 'http', 'HTTP 端点故障', '${FAIL_COUNT} 个端点失败');"
fi
# ============================================================================
# 3. 业务层检测
# ============================================================================
# message_log 小时增量过低(钉钉推送可能断了)
HOUR_MSG=$(sqlite_exec "SELECT msg_log_hourly FROM business_metrics ORDER BY ts DESC LIMIT 1;")
if [[ -n "${HOUR_MSG}" ]] && [[ ${HOUR_MSG} -lt ${ALERT_MSG_LOG_HOUR_MIN} ]] && [[ ${HOUR_MSG} -gt 0 ]]; then
BODY="**ecology_message_log 小时增量仅 ${HOUR_MSG} 条**(阈值 ${ALERT_MSG_LOG_HOUR_MIN})\n\n> 钉钉推送可能阻塞。泛微云桥 e-Bridge 为黑盒线程,需登录 OA 后台检查「消息推送」日志或用 SHOW PROCESSLIST 看 active 线程"
dingtalk_alert "P1" "钉钉推送疑似停滞" "${BODY}"
sqlite_exec "INSERT INTO alerts (ts, level, category, title, detail) VALUES (${NOW}, 'P1', 'business', '钉钉推送疑似停滞', 'msg_hourly=${HOUR_MSG}');"
fi
# ============================================================================
# 4. 基础设施层检测
# ============================================================================
# Docker daemon 挂了
DOCKER_OK=$(sqlite_exec "SELECT docker_daemon_ok FROM infra_metrics ORDER BY ts DESC LIMIT 1;")
PREV_DOCKER_OK=$(sqlite_exec "SELECT docker_daemon_ok FROM infra_metrics ORDER BY ts DESC LIMIT 2 OFFSET 1;")
if [[ "${DOCKER_OK}" == "0" ]] && [[ "${PREV_DOCKER_OK}" == "0" ]]; then
BODY="**Docker daemon 连续 2 次检测为不可用**\n\n> ⚠️ 今天的 rm -rf /ipaas 事故中 Docker daemon 重启过。请立即登录 DSH 检查:\n\`\`\`\nsystemctl status docker\ndocker info | grep Root\ndf -h\n\`\`\`"
dingtalk_alert "P0" "Docker daemon 挂了!" "${BODY}"
sqlite_exec "INSERT INTO alerts (ts, level, category, title, detail) VALUES (${NOW}, 'P0', 'infra', 'Docker daemon 不可用', '连续 2 次失败');"
fi
# 容器数突变(对比最近 30 分钟的平均数)
CURR_COUNT=$(sqlite_exec "SELECT docker_container_count FROM infra_metrics ORDER BY ts DESC LIMIT 1;")
AVG_COUNT=$(sqlite_exec "SELECT AVG(docker_container_count) FROM infra_metrics WHERE ts > ${NOW}-1800;")
if [[ -n "${CURR_COUNT}" ]] && [[ -n "${AVG_COUNT}" ]]; then
DELTA=$(( CURR_COUNT - ${AVG_COUNT%%.*} ))
ABS_DELTA=${DELTA#-}
if [[ ${ABS_DELTA} -ge ${ALERT_DOCKER_CONTAINER_DELTA} ]]; then
BODY="**Docker 容器数突变**:当前 ${CURR_COUNT},30 分钟均值 ${AVG_COUNT}\n\n> ⚠️ 容器异常消失!请立刻检查 docker ps -a"
dingtalk_alert "P0" "Docker 容器数突变!" "${BODY}"
sqlite_exec "INSERT INTO alerts (ts, level, category, title, detail) VALUES (${NOW}, 'P0', 'infra', 'Docker 容器数突变', 'curr=${CURR_COUNT} avg=${AVG_COUNT}');"
fi
fi
# 磁盘
DISK_PCT=$(sqlite_exec "SELECT disk_usage_pct FROM infra_metrics ORDER BY ts DESC LIMIT 1;")
if [[ -n "${DISK_PCT}" ]] && (( ${DISK_PCT%%.*} >= ALERT_DISK_USAGE_PCT )); then
BODY="**DSH 磁盘使用率 ${DISK_PCT}%**(阈值 ${ALERT_DISK_USAGE_PCT}%)\n\n> 请检查:df -h → du -sh /var/lib/docker/* → pm2 logs 清理 → 旧 SQLite 数据归档"
dingtalk_alert "P1" "DSH 磁盘告警" "${BODY}"
sqlite_exec "INSERT INTO alerts (ts, level, category, title, detail) VALUES (${NOW}, 'P1', 'infra', '磁盘使用率过高', 'disk=${DISK_PCT}%');"
fi
# inode
INODE_PCT=$(sqlite_exec "SELECT inode_usage_pct FROM infra_metrics ORDER BY ts DESC LIMIT 1;")
if [[ -n "${INODE_PCT}" ]] && (( ${INODE_PCT%%.*} >= ALERT_INODE_USAGE_PCT )); then
BODY="**DSH inode 使用率 ${INODE_PCT}%**(阈值 ${ALERT_INODE_USAGE_PCT}%)\n\n> 可能有大量小文件,检查 find / -type f | wc -l"
dingtalk_alert "P1" "DSH inode 告警" "${BODY}"
sqlite_exec "INSERT INTO alerts (ts, level, category, title, detail) VALUES (${NOW}, 'P1', 'infra', 'inode 使用率过高', 'inode=${INODE_PCT}%');"
fi
# Docker root 写入突增
ROOT_WRITE=$(sqlite_exec "SELECT docker_root_write_mb FROM infra_metrics ORDER BY ts DESC LIMIT 1;")
if [[ -n "${ROOT_WRITE}" ]] && (( ${ROOT_WRITE%%.*} >= ALERT_DOCKER_ROOT_WRITE_MB_PER_MIN )); then
BODY="**Docker data-root 1 分钟内写入 ${ROOT_WRITE}MB**(阈值 ${ALERT_DOCKER_ROOT_WRITE_MB_PER_MIN}MB)\n\n> 可能有容器在疯狂写数据,检查:docker stats → docker logs"
dingtalk_alert "P2" "Docker data-root 写入突增" "${BODY}"
sqlite_exec "INSERT INTO alerts (ts, level, category, title, detail) VALUES (${NOW}, 'P2', 'infra', 'Docker root 写入突增', 'write_mb=${ROOT_WRITE}');"
fi
log_info "[ALERT] ========== 告警检测完成 =========="