feat: watchdog v4 + check-alert v2 + collect-business v2 + DUMMY补数据脚本

This commit is contained in:
oa-monitor committed 2026-11-04 17:25:42 +08:00
commit 7ab75364b1
22 files changed
+1795

No files matched your search

Executable
+12
View File
@@ -0,0 +1,12 @@
import paramiko, sys
host, user, password, cmd = sys.argv[1], sys.argv[2], sys.argv[3], sys.argv[4]
ssh = paramiko.SSHClient()
ssh.set_missing_host_key_policy(paramiko.AutoAddPolicy())
try:
ssh.connect(host, username=user, password=password, timeout=10)
stdin, stdout, stderr = ssh.exec_command(cmd)
sys.stdout.write(stdout.read().decode("utf-8", errors="replace"))
ssh.close()
except Exception as e:
print(f"SSH_ERROR: {e}", file=sys.stderr)
sys.exit(1)
+166
View File
@@ -0,0 +1,166 @@
#!/usr/bin/env bash
# ============================================================================
# OA Monitor · 告警检测(每 1 分钟)
# v2 修复(2026-10-28):
# - 新增 Step 0:数据新鲜度前置检查,防止采集脚本挂了后 check-alert 读旧数据乱告警
# - Step 3 业务层检测被 Step 0 结果门禁
# ============================================================================
set -u
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
source "${SCRIPT_DIR}/../lib/common.sh"
NOW=$(date +%s)
log_info "[ALERT] ========== 告警检测开始 =========="
# ============================================================================
# Step 0: 数据新鲜度前置检查(防假阳性的第一层)
# ============================================================================
# business_metrics 由 collect-business.sh 每 15min 写一次
# db_metrics 由 collect-db.sh 每 30s 写一次
# 如果最新 business_metrics ts 距现在 > 30min → 采集脚本挂了 → 告警+跳过业务层
BIZ_TS=$(sqlite_exec "SELECT ts FROM business_metrics ORDER BY ts DESC LIMIT 1;" | head -1)
DB_TS=$(sqlite_exec "SELECT ts FROM db_metrics ORDER BY ts DESC LIMIT 1;" | head -1)
BIZ_FRESH=true
BIZ_AGE=0
if [[ ! "${BIZ_TS}" =~ ^[0-9]+$ ]]; then
log_error "[ALERT] business_metrics 表无数据或采集从未运行"
BIZ_FRESH=false
BIZ_AGE=999999
elif [[ ! "${DB_TS}" =~ ^[0-9]+$ ]]; then
log_warn "[ALERT] db_metrics 表无数据(可能 MySQL 断了)"
else
BIZ_AGE=$((NOW - BIZ_TS))
DB_AGE=$((NOW - DB_TS))
if [[ ${BIZ_AGE} -gt 1800 ]]; then # > 30 分钟
BIZ_FRESH=false
HUMAN_AGE=$((BIZ_AGE / 60))
BODY="**🔴 采集脚本挂了!**\n\n| 表 | 最新 ts | 年龄 |\n|---|---|---|\n| business_metrics | $(sqlite_exec \"SELECT datetime(${BIZ_TS}, 'unixepoch', 'localtime')\") | **${HUMAN_AGE}min** |\n| db_metrics | $(sqlite_exec \"SELECT datetime(${DB_TS}, 'unixepoch', 'localtime')\") | ${DB_AGE}s |\n\n> ⚠️ collect-business.sh 超过 30min 没写入,先告警采集挂了,防止后续业务层读旧数据乱告警"
dingtalk_alert "P1" "采集脚本挂了!business_metrics 过期 ${HUMAN_AGE}min" "${BODY}"
log_error "[ALERT] 数据过期 ${HUMAN_AGE}min,禁止跑业务层检测(防假阳性)"
else
log_info "[ALERT] 数据新鲜:biz_age=${BIZ_AGE}s db_age=${DB_AGE}s"
fi
fi
# ============================================================================
# 1. DB 层检测(db_metrics 每分钟更新,数据可信)
# ============================================================================
TR=$(sqlite_exec "SELECT threads_running FROM db_metrics ORDER BY ts DESC LIMIT 1;" | head -1)
if [[ -n "${TR}" ]] && [[ ${TR} =~ ^[0-9]+$ ]] && [[ ${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;' | head -1)\n活跃空闲:$(sqlite_exec 'SELECT threads_idle FROM db_metrics ORDER BY ts DESC LIMIT 1;' | head -1)"
dingtalk_alert "P1" "DB 层 Threads_running 过高" "${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 > ${ALERT_SLOW_QUERY_SEC}s');"
fi
# ============================================================================
# 2. HTTP 端点检测(http_metrics 每分钟更新,数据可信)
# ============================================================================
FAIL_COUNT=0
FAIL_DETAIL=""
for endpoint in "${OA_ENDPOINTS[@]}"; do
FAIL_HERE=$(sqlite_exec "SELECT COUNT(*) FROM http_metrics WHERE endpoint='${endpoint}' AND ts > ${NOW}-120 AND healthy=0;" | head -1)
FAIL_HERE=${FAIL_HERE:-0}
if [[ "${FAIL_HERE}" =~ ^[0-9]+$ ]] && [[ ${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;" | head -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', '${FAIL_COUNT} 个端点失败');"
fi
# ============================================================================
# 3. 业务层检测(被 BIZ_FRESH 门禁)
# ============================================================================
if [[ "${BIZ_FRESH}" != "true" ]]; then
log_warn "[ALERT] ⏸️ business_metrics 数据过期,跳过业务层检测(防假阳性)"
else
# 数据新鲜 → 正常业务层判断
HOUR_MSG=$(sqlite_exec "SELECT msg_log_hourly FROM business_metrics ORDER BY ts DESC LIMIT 1;" | head -1)
ONGOING=$(sqlite_exec "SELECT ongoing_requests FROM business_metrics ORDER BY ts DESC LIMIT 1;" | head -1)
HOUR_MSG=${HOUR_MSG:-0}
ONGOING=${ONGOING:-0}
CURR_HOUR=$((10#$(date +%H))) # 算术强制 10 进制,防 "08"/"09" 被当 8/9 或 octal
WORK_HOURS_ENABLED=false
if [[ ${CURR_HOUR} -ge 8 ]] && [[ ${CURR_HOUR} -le 21 ]]; then
WORK_HOURS_ENABLED=true
fi
if [[ "${WORK_HOURS_ENABLED}" == "true" ]]; then
if [[ ${HOUR_MSG} -eq 0 ]] && [[ ${ONGOING} -gt 0 ]]; then
BODY="**🔴 P0 钉钉推送完全断档!**\n\n"
BODY="${BODY}| 指标 | 值 |\n|---|---|\n"
BODY="${BODY}| message_log 小时增量 | **${HOUR_MSG} 条** |\n"
BODY="${BODY}| 正在流转审批 | **${ONGOING} 条** |\n"
BODY="${BODY}| 当前时段 | **工作时间**(${CURR_HOUR}:00)|\n"
BODY="${BODY}| 数据新鲜度 | **${BIZ_AGE}s** ✅ |\n\n"
BODY="${BODY}> ⚠️ 审批在流转,但钉钉消息完全没发!\n"
BODY="${BODY}> 常见原因:Util_Message 死锁、Resin 消息推送线程挂了、Redis 报错\n"
BODY="${BODY}> **应急:SSH 到 OA 服务器执行 \`ps aux | grep Resin\` → \`stopresin.sh && startresin.sh\`**"
dingtalk_alert "P0" "钉钉推送完全断档(消息=0 但业务在跑)" "${BODY}"
sqlite_exec "INSERT INTO alerts (ts, level, category, title, detail) VALUES (${NOW}, 'P0', 'business', '钉钉推送完全断档', 'msg=${HOUR_MSG} ongoing=${ONGOING}');"
fi
else
log_info "[ALERT] ⏸️ 非工作时间(${CURR_HOUR}:00),跳过业务层告警"
fi
fi
# ============================================================================
# 4. 基础设施层检测(infra_metrics 每分钟更新,数据可信)
# ============================================================================
DOCKER_OK=$(sqlite_exec "SELECT docker_daemon_ok FROM infra_metrics ORDER BY ts DESC LIMIT 1;" | head -1)
PREV_DOCKER_OK=$(sqlite_exec "SELECT docker_daemon_ok FROM infra_metrics ORDER BY ts DESC LIMIT 2 OFFSET 1;" | head -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> 请立即检查:\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
CURR_COUNT=$(sqlite_exec "SELECT docker_container_count FROM infra_metrics ORDER BY ts DESC LIMIT 1;" | head -1)
AVG_COUNT=$(sqlite_exec "SELECT AVG(docker_container_count) FROM infra_metrics WHERE ts > ${NOW}-1800;" | head -1)
if [[ -n "${CURR_COUNT}" ]] && [[ -n "${AVG_COUNT}" ]] && [[ "${CURR_COUNT}" =~ ^[0-9]+$ ]]; then
AVG_INT=${AVG_COUNT%%.*}
DELTA=$(( CURR_COUNT - AVG_INT ))
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=${AVG_COUNT}');"
fi
fi
DISK_PCT=$(sqlite_exec "SELECT disk_usage_pct FROM infra_metrics ORDER BY ts DESC LIMIT 1;" | head -1)
if [[ -n "${DISK_PCT}" ]] && [[ "${DISK_PCT}" =~ ^[0-9.]+$ ]] && (( ${DISK_PCT%%.*} >= ALERT_DISK_USAGE_PCT )); then
BODY="**DSH 磁盘使用率 ${DISK_PCT}%**(阈值 ${ALERT_DISK_USAGE_PCT}%)"
dingtalk_alert "P1" "DSH 磁盘告警" "${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_PCT=$(sqlite_exec "SELECT inode_usage_pct FROM infra_metrics ORDER BY ts DESC LIMIT 1;" | head -1)
if [[ -n "${INODE_PCT}" ]] && [[ "${INODE_PCT}" =~ ^[0-9.]+$ ]] && (( ${INODE_PCT%%.*} >= ALERT_INODE_USAGE_PCT )); then
BODY="**DSH inode 使用率 ${INODE_PCT}%**(阈值 ${ALERT_INODE_USAGE_PCT}%)"
dingtalk_alert "P1" "DSH inode 告警" "${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
log_info "[ALERT] ========== 告警检测完成 =========="
+155
View File
@@ -0,0 +1,155 @@
#!/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', '${FAIL_COUNT} 个端点失败');"
fi
# ============================================================================
# 3. 业务层检测
# ============================================================================
# 凌晨没人操作 OA 时 msg=0 本来正常——加时间窗口和业务活跃双重判断
HOUR_MSG=$(sqlite_exec "SELECT msg_log_hourly FROM business_metrics ORDER BY ts DESC LIMIT 1;")
ONGOING=$(sqlite_exec "SELECT ongoing_requests FROM business_metrics ORDER BY ts DESC LIMIT 1;")
HOUR_MSG=${HOUR_MSG:-0}
ONGOING=${ONGOING:-0}
# 当前小时数(用于判断是否工作时间)
CURR_HOUR=$(date +%H | sed 's/^0//') # 去掉前导零
WORK_HOURS_ENABLED=false
if [[ ${CURR_HOUR} -ge 8 ]] && [[ ${CURR_HOUR} -le 21 ]]; then
WORK_HOURS_ENABLED=true # 08:00~22:00 才算工作时间
fi
if [[ "${WORK_HOURS_ENABLED}" == "true" ]]; then
# 只保留 P0:工作时间内,有业务在跑,但消息一条都没推出来 → 推送链路完全断裂
# P1 已砍掉(噪音太大:上午刚开始上班 msg=99 也会触发,实际推送是正常的)
if [[ ${HOUR_MSG} -eq 0 ]] && [[ ${ONGOING} -gt 0 ]]; then
BODY="**🔴 P0 钉钉推送完全断档!**\n\n"
BODY="${BODY}| 指标 | 值 |\n|---|---|\n"
BODY="${BODY}| message_log 小时增量 | **${HOUR_MSG} 条** |\n"
BODY="${BODY}| 正在流转审批 | **${ONGOING} 条** |\n"
BODY="${BODY}| 当前时段 | **工作时间**(${CURR_HOUR}:00)|\n\n"
BODY="${BODY}> ⚠️ 审批在流转,但钉钉消息完全没发!\n"
BODY="${BODY}> 常见原因:Util_Message 死锁、Resin 消息推送线程挂了、Redis 报错\n"
BODY="${BODY}> **应急:SSH 到 OA 服务器执行 `ps aux | grep Resin` → `bash /usr/weaver/Resin/bin/stopresin.sh && startresin.sh`**"
if dingtalk_alert "P0" "钉钉推送完全断档(消息=0 但业务在跑)" "${BODY}"; then
sqlite_exec "INSERT INTO alerts (ts, level, category, title, detail) VALUES (${NOW}, 'P0', 'business', '钉钉推送完全断档', 'msg=${HOUR_MSG} ongoing=${ONGOING}');"
fi
fi
else
log_info "[BIZ] ⏸️ 非工作时间(${CURR_HOUR}:00),跳过业务层告警检查"
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(读最新一条和上一条的差值)
ROOT_NOW=$(sqlite_exec "SELECT docker_root_write_mb FROM infra_metrics ORDER BY ts DESC LIMIT 1;")
ROOT_PREV=$(sqlite_exec "SELECT docker_root_write_mb FROM infra_metrics ORDER BY ts DESC LIMIT 1 OFFSET 1;")
ROOT_DELTA=0
if [[ -n "${ROOT_NOW}" ]] && [[ -n "${ROOT_PREV}" ]]; then
# 截断小数,bash $(()) 只认整数
ROOT_NOW_INT=${ROOT_NOW%%.*}
ROOT_PREV_INT=${ROOT_PREV%%.*}
if [[ "${ROOT_NOW_INT}" =~ ^[0-9]+$ ]] && [[ "${ROOT_PREV_INT}" =~ ^[0-9]+$ ]]; then
ROOT_DELTA=$(( ROOT_NOW_INT - ROOT_PREV_INT ))
fi
fi
if [[ ${ROOT_DELTA} -ge ${ALERT_DOCKER_ROOT_WRITE_MB_PER_MIN} ]]; then
BODY="**Docker overlay2 增长 ${ROOT_DELTA}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 (${NOW}, 'P2', 'infra', 'Docker root 写入突增', 'delta_mb=${ROOT_DELTA}');"
fi
log_info "[ALERT] ========== 告警检测完成 =========="
Executable
+36
View File
@@ -0,0 +1,36 @@
#!/usr/bin/env bash
# ============================================================================
# OA Monitor · 数据清理(每天 03:00)
# 保留最近 7 天指标,alerts 和 alert_cooldown 保留 30 天
# ============================================================================
set -u
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
source "${SCRIPT_DIR}/../lib/common.sh"
NOW=$(date +%s)
RETENTION_7D=$(( NOW - 7 * 86400 ))
RETENTION_30D=$(( NOW - 30 * 86400 ))
log_info "[CLEANUP] 开始清理,阈值 ts < ${RETENTION_7D}(7天),alerts < ${RETENTION_30D}(30天)"
# 各层指标表
for table in db_metrics http_metrics business_metrics infra_metrics db_processlist; do
BEFORE=$(sqlite_exec "SELECT COUNT(*) FROM ${table};")
sqlite_exec "DELETE FROM ${table} WHERE ts < ${RETENTION_7D};"
AFTER=$(sqlite_exec "SELECT COUNT(*) FROM ${table};")
log_info "[CLEANUP] ${table}: ${BEFORE} → ${AFTER}(删了 $((BEFORE - AFTER)) 条)"
done
# alerts 保留 30 天
BEFORE=$(sqlite_exec "SELECT COUNT(*) FROM alerts;")
sqlite_exec "DELETE FROM alerts WHERE ts < ${RETENTION_30D};"
AFTER=$(sqlite_exec "SELECT COUNT(*) FROM alerts;")
log_info "[CLEANUP] alerts: ${BEFORE} → ${AFTER}(删了 $((BEFORE - AFTER)) 条)"
# VACUUM 压缩(可选,不频繁)
if (( NOW % (86400 * 7) < 3600 )); then
log_info "[CLEANUP] 每周 VACUUM SQLite..."
sqlite_exec "VACUUM;"
fi
log_info "[CLEANUP] 完成"
+50
View File
@@ -0,0 +1,50 @@
#!/usr/bin/env bash
# ============================================================================
# OA Monitor · 业务指标采集(每 15 分钟,crontab: */15 * * * *)
# v2 修复(2026-10-28):
# - msg_log_hourly 实际用 15min 窗口(和 crontab 频率对齐),跨天逻辑简化
# - 时间字段正确使用 CREATEDATE varchar(10) + CREATETIME varchar(8)
# ecology_message_log 没有 DATETIME 类型的 pushlog_time 字段!
# ============================================================================
set -u
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
source "${SCRIPT_DIR}/../lib/common.sh"
NOW=$(date +%s)
NOW_DATE=$(date '+%Y-%m-%d')
# --- 1. ecology_message_log 15 分钟增量 ---
# v2 简化:15min 窗口几乎不跨天,直接用 CURDATE()
MSG_WINDOW=15
MSG_COUNT=$(MYSQL_PWD="${DB_PASS}" timeout 30 mysql -h39.100.72.102 -P3306 -uecology ecology -N -e "
SELECT COUNT(*) FROM ecology_message_log
WHERE CREATEDATE = CURDATE()
AND CREATETIME >= DATE_FORMAT(NOW() - INTERVAL ${MSG_WINDOW} MINUTE, '%H:%i:%s');
" 2>/dev/null | tail -1)
MSG_COUNT=${MSG_COUNT:-0}
# --- 2. 正在流转的审批数 ---
ONGOING=$(MYSQL_PWD="${DB_PASS}" timeout 30 mysql -h39.100.72.102 -P3306 -uecology ecology -N -e "
SELECT COUNT(*) FROM workflow_requestbase
WHERE CURRENTNODEID > 0 AND (STATUS IS NULL OR STATUS NOT LIKE '%归档%');
" 2>/dev/null | tail -1)
ONGOING=${ONGOING:-0}
# --- 3. 今日新建审批数 ---
NEW_TODAY=$(MYSQL_PWD="${DB_PASS}" timeout 30 mysql -h39.100.72.102 -P3306 -uecology ecology -N -e "
SELECT COUNT(*) FROM workflow_requestbase WHERE CREATEDATE = '${NOW_DATE}';
" 2>/dev/null | tail -1)
NEW_TODAY=${NEW_TODAY:-0}
# --- 4. 超时任务数 ---
OVERDUE=$(MYSQL_PWD="${DB_PASS}" timeout 30 mysql -h39.100.72.102 -P3306 -uecology ecology -N -e "
SELECT COUNT(*) FROM workflow_requestlog
WHERE timeout = 1 AND operatedate >= DATE_SUB(CURDATE(), INTERVAL 1 DAY);
" 2>/dev/null | tail -1)
OVERDUE=${OVERDUE:-0}
sqlite_exec "INSERT INTO business_metrics VALUES (
${NOW}, ${MSG_COUNT}, 0.0, ${ONGOING}, ${OVERDUE}
);"
log_info "[BIZ] ✅ msg_15min=${MSG_COUNT} ongoing=${ONGOING} new_today=${NEW_TODAY} overdue=${OVERDUE}"
+55
View File
@@ -0,0 +1,55 @@
#!/usr/bin/env bash
# ============================================================================
# OA Monitor · 业务指标采集(每 15 分钟)
# ⚠️ ecology_message_log 没有 created datetime 列!
# 时间用 CREATEDATE varchar(10) + CREATETIME varchar(8) 拼接
# ============================================================================
set -u
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
source "${SCRIPT_DIR}/../lib/common.sh"
NOW=$(date +%s)
NOW_DATE=$(date '+%Y-%m-%d')
NOW_HOUR=$(date '+%H')
NOW_MIN=$(date '+%M')
# 1 小时前的时间(用于计算小时增量)
HOUR_AGO_DATE=$(date -v-1H '+%Y-%m-%d' 2>/dev/null || date --date='1 hour ago' '+%Y-%m-%d' 2>/dev/null)
HOUR_AGO_TIME=$(date -v-1H '+%H:%M:%S' 2>/dev/null || date --date='1 hour ago' '+%H:%M:%S' 2>/dev/null)
# --- 1. ecology_message_log 小时增量(正确用 CREATEDATE + CREATETIME 拼接)---
MSG_HOURLY=$(MYSQL_PWD="${DB_PASS}" timeout 30 mysql -h39.100.72.102 -P3306 -uecology ecology -N -e "
SELECT COUNT(*) FROM ecology_message_log
WHERE (CREATEDATE = '${NOW_DATE}' AND CREATETIME >= DATE_FORMAT(NOW() - INTERVAL 1 HOUR, '%H:%i:%s'))
OR (CREATEDATE = '${HOUR_AGO_DATE}' AND CREATETIME >= '${HOUR_AGO_TIME}');
" 2>/dev/null | tail -1)
MSG_HOURLY=${MSG_HOURLY:-0}
# --- 2. 正在流转的审批数 ---
# 泛微 CURRENTSTATUS 全 NULL,STATUS 存中文节点名(如"行政审批层级1 /审批人1 至 归档")
# 判断:CURRENTNODEID > 0 且不包含"归档"
ONGOING=$(MYSQL_PWD="${DB_PASS}" timeout 30 mysql -h39.100.72.102 -P3306 -uecology ecology -N -e "
SELECT COUNT(*) FROM workflow_requestbase
WHERE CURRENTNODEID > 0 AND (STATUS IS NULL OR STATUS NOT LIKE '%归档%');
" 2>/dev/null | tail -1)
ONGOING=${ONGOING:-0}
# --- 3. 今日新建审批数 ---
NEW_TODAY=$(MYSQL_PWD="${DB_PASS}" timeout 30 mysql -h39.100.72.102 -P3306 -uecology ecology -N -e "
SELECT COUNT(*) FROM workflow_requestbase WHERE CREATEDATE = '${NOW_DATE}';
" 2>/dev/null | tail -1)
NEW_TODAY=${NEW_TODAY:-0}
# --- 4. 超时任务数 ---
OVERDUE=$(MYSQL_PWD="${DB_PASS}" timeout 30 mysql -h39.100.72.102 -P3306 -uecology ecology -N -e "
SELECT COUNT(*) FROM workflow_requestlog
WHERE timeout = 1 AND operatedate >= DATE_SUB(CURDATE(), INTERVAL 1 DAY);
" 2>/dev/null | tail -1)
OVERDUE=${OVERDUE:-0}
FAIL_RATE="0.0" # ecology_message_log 没有失败率字段,暂置 0
sqlite_exec "INSERT INTO business_metrics VALUES (
${NOW}, ${MSG_HOURLY}, ${FAIL_RATE}, ${ONGOING}, ${OVERDUE}
);"
log_info "[BIZ] ✅ msg_hourly=${MSG_HOURLY} ongoing=${ONGOING} new_today=${NEW_TODAY} overdue=${OVERDUE}"
+98
View File
@@ -0,0 +1,98 @@
#!/usr/bin/env bash
# ============================================================================
# OA Monitor · DB 层采集(每 30 秒)
# 在 DSH 宿主机上直接 mysql 连 OA DB(DSH 本身就是跳板,不需要再 ssh)
# ============================================================================
set -u
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
source "${SCRIPT_DIR}/../lib/common.sh"
NOW=$(date +%s)
# ---- MySQL 直连(DSH 上直接连 OA DB 公网映射)----
DB_MYSQL() {
MYSQL_PWD="${DB_PASS}" timeout 12 mysql -h39.100.72.102 -P3306 -uecology ecology "$@" 2>/dev/null
}
db_exec() {
DB_MYSQL -e "$1"
}
# ---- 1. SHOW STATUS ----
log_info "[DB] SHOW STATUS"
STATUS=$(DB_MYSQL -e "SHOW GLOBAL STATUS" | awk -F'\t' '
$1=="Threads_running"||$1=="Threads_connected"||$1=="Threads_idle"||
$1=="Uptime"||$1=="Innodb_rows_read"||$1=="Innodb_rows_inserted"||
$1=="Innodb_rows_updated"||$1=="Innodb_rows_deleted"||$1=="Slow_queries" { print }
')
THREADS_RUNNING=$(echo "${STATUS}" | awk -F'\t' '$1=="Threads_running"{print $2}')
THREADS_CONNECTED=$(echo "${STATUS}" | awk -F'\t' '$1=="Threads_connected"{print $2}')
THREADS_IDLE=$(echo "${STATUS}" | awk -F'\t' '$1=="Threads_idle"{print $2}')
UPTIME=$(echo "${STATUS}" | awk -F'\t' '$1=="Uptime"{print $2}')
INNODB_READ=$(echo "${STATUS}" | awk -F'\t' '$1=="Innodb_rows_read"{print $2}')
INNODB_INS=$(echo "${STATUS}" | awk -F'\t' '$1=="Innodb_rows_inserted"{print $2}')
INNODB_UPD=$(echo "${STATUS}" | awk -F'\t' '$1=="Innodb_rows_updated"{print $2}')
INNODB_DEL=$(echo "${STATUS}" | awk -F'\t' '$1=="Innodb_rows_deleted"{print $2}')
SLOW_QUERIES=$(echo "${STATUS}" | awk -F'\t' '$1=="Slow_queries"{print $2}')
INNODB_READ=${INNODB_READ:-0}
INNODB_INS=${INNODB_INS:-0}
INNODB_UPD=${INNODB_UPD:-0}
INNODB_DEL=${INNODB_DEL:-0}
INNODB_WRITE=$(( INNODB_INS + INNODB_UPD + INNODB_DEL ))
THREADS_RUNNING=${THREADS_RUNNING:-0}
THREADS_CONNECTED=${THREADS_CONNECTED:-0}
THREADS_IDLE=${THREADS_IDLE:-0}
UPTIME=${UPTIME:-0}
SLOW_QUERIES=${SLOW_QUERIES:-0}
# ---- 2. 上一次快照 → 差值 ----
PREV=$(sqlite_exec "SELECT * FROM db_metrics ORDER BY ts DESC LIMIT 1;")
QPS_READ=0; QPS_WRITE=0; SLOW_DELTA=0
if [[ -n "${PREV}" ]]; then
PREV_UPTIME=$(echo "${PREV}" | cut -d'|' -f5)
PREV_READ=$(echo "${PREV}" | cut -d'|' -f6)
PREV_WRITE=$(echo "${PREV}" | cut -d'|' -f7)
PREV_SLOW=$(echo "${PREV}" | cut -d'|' -f9)
PREV_UPTIME=${PREV_UPTIME:-0}; PREV_READ=${PREV_READ:-0}
PREV_WRITE=${PREV_WRITE:-0}; PREV_SLOW=${PREV_SLOW:-0}
DIFF_SEC=$(( NOW - PREV_UPTIME ))
if [[ ${DIFF_SEC} -gt 0 ]]; then
QPS_READ=$(( (INNODB_READ - PREV_READ) / DIFF_SEC ))
QPS_WRITE=$(( (INNODB_WRITE - PREV_WRITE) / DIFF_SEC ))
fi
SLOW_DELTA=$(( SLOW_QUERIES - PREV_SLOW ))
fi
# ---- 3. 写 SQLite ----
sqlite_exec "INSERT INTO db_metrics VALUES (
${NOW}, ${THREADS_RUNNING}, ${THREADS_CONNECTED}, ${THREADS_IDLE},
${UPTIME}, ${QPS_READ}, ${QPS_WRITE}, ${SLOW_DELTA}
);"
log_info "[DB] ✅ 写入:Threads_running=${THREADS_RUNNING} QPS_read=${QPS_READ} QPS_write=${QPS_WRITE}"
# ---- 4. 长查询 processlist ----
log_info "[DB] processlist"
PLIST=$(DB_MYSQL -e 'SHOW PROCESSLIST;')
echo "${PLIST}" | awk -F'\t' '
NR==1 { next }
NF < 8 { next }
{
pid=$1; user=$2; state=$6; time=$7; info=$8
if (time+0 >= 5 && pid+0 > 0) {
gsub(/\t/, " ", info)
printf "%d\t%d\t%s\t%s\t%s\n", pid, time, user, state, substr(info,1,300)
}
}
' | while IFS=$'\t' read -r pid time user state info; do
[[ -z "${pid}" ]] && continue
info_safe=$(echo "${info}" | sed "s/'/\"/g")
sqlite_exec "INSERT INTO db_processlist (ts, pid, user, state, time_sec, info)
VALUES (${NOW}, ${pid}, '${user}', '${state}', ${time}, '${info_safe}');"
done
log_info "[DB] ✅ 完成"
+31
View File
@@ -0,0 +1,31 @@
#!/usr/bin/env bash
# ============================================================================
# OA Monitor · HTTP 端点健康检查(每 30 秒)
# 在 DSH 上直接 curl
# ============================================================================
set -u
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
source "${SCRIPT_DIR}/../lib/common.sh"
NOW=$(date +%s)
for endpoint in "${OA_ENDPOINTS[@]}"; do
result=$(curl -sk -o /dev/null -w '%{http_code}|%{time_total}' --max-time 10 "${endpoint}" 2>/dev/null)
status=$(echo "${result}" | cut -d'|' -f1)
time_ms=$(echo "${result}" | cut -d'|' -f2 | awk '{printf "%d", $1*1000}')
status=${status:-0}
time_ms=${time_ms:-0}
case "${status}" in
000) healthy=0 ;; # 连不上 = 挂了
5*) healthy=0 ;; # 5xx = 服务异常
*) healthy=1 ;; # 2xx/3xx/4xx 都算活着(4xx 是业务路径不对,但服务本身通了)
esac
sqlite_exec "INSERT INTO http_metrics (ts, endpoint, status_code, resp_time_ms, healthy)
VALUES (${NOW}, '${endpoint}', ${status}, ${time_ms}, ${healthy});"
log_info "[HTTP] ${endpoint} status=${status} time=${time_ms}ms healthy=${healthy}"
done
log_info "[HTTP] 全部完成"
+53
View File
@@ -0,0 +1,53 @@
#!/usr/bin/env bash
# ============================================================================
# OA Monitor · 基础设施层采集(每 30 秒)
# 在 DSH 上直接执行 docker / df / du
# ============================================================================
set -u
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
source "${SCRIPT_DIR}/../lib/common.sh"
NOW=$(date +%s)
DOCKER_ROOT="/var/lib/docker"
# ---- Docker daemon + 容器数 ----
CONTAINER_COUNT=$(docker ps -q 2>/dev/null | wc -l | tr -d ' ')
if [[ "${CONTAINER_COUNT}" =~ ^[0-9]+$ ]] && [[ ${CONTAINER_COUNT} -ge 0 ]]; then
DOCKER_OK=1
else
DOCKER_OK=0
CONTAINER_COUNT=0
fi
# ---- Docker root 写入量 ----
# docker_root_write_mb 字段存储 "上一次 du overlay2 的总大小 MB"
# 告警脚本里直接看 "本次 du - 上次存的值" 的差值
PREV_TOTAL_MB=$(sqlite_exec "SELECT COALESCE(docker_root_write_mb, 0) FROM infra_metrics ORDER BY ts DESC LIMIT 1;")
CURRENT_TOTAL_MB=$(du -sm "${DOCKER_ROOT}/overlay2" 2>/dev/null | cut -f1 | tr -d ' ')
PREV_TOTAL_MB=${PREV_TOTAL_MB:-0}
CURRENT_TOTAL_MB=${CURRENT_TOTAL_MB:-0}
# 计算增量(日志用)
DELTA=0
if [[ "${PREV_TOTAL_MB}" =~ ^[0-9]+$ ]] && [[ ${CURRENT_TOTAL_MB} -gt ${PREV_TOTAL_MB} ]]; then
DELTA=$(( CURRENT_TOTAL_MB - PREV_TOTAL_MB ))
elif [[ ! "${PREV_TOTAL_MB}" =~ ^[0-9]+$ ]]; then
DELTA=-1 # 首次采,设 -1 表示"无对比"
fi
# ---- 磁盘 / inode / load ----
DISK_PCT=$(df / | tail -1 | awk '{print $5}' | tr -d '%')
INODE_PCT=$(df -i / | tail -1 | awk '{print $5}' | tr -d '%')
LOADAVG=$(cat /proc/loadavg | awk '{print $1}')
DISK_PCT=${DISK_PCT:-0}
INODE_PCT=${INODE_PCT:-0}
LOADAVG=${LOADAVG:-0}
# 写 SQLite(把"当前总大小 MB"存起来,下次算差值)
# 注意 docker_root_write_mb 存的是整数 MB(du -sm 输出本身就是整数,.0 是 sqlite REAL 类型加的)
sqlite_exec "INSERT INTO infra_metrics VALUES (
${NOW}, ${DOCKER_OK}, ${CONTAINER_COUNT}, CAST(${CURRENT_TOTAL_MB} AS INTEGER),
${DISK_PCT}, ${INODE_PCT}, ${LOADAVG}
);"
log_info "[INFRA] ✅ docker_ok=${DOCKER_OK} containers=${CONTAINER_COUNT} root_overlay2=${CURRENT_TOTAL_MB}MB delta=${DELTA}MB disk=${DISK_PCT}% inode=${INODE_PCT}% loadavg=${LOADAVG}"
Executable
+84
View File
@@ -0,0 +1,84 @@
#!/usr/bin/env bash
# ============================================================================
# OA Monitor · SQLite 数据库初始化
# 只建表,不删数据;幂等可重复执行
# ============================================================================
set -u
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
source "${SCRIPT_DIR}/../lib/common.sh"
log_info "初始化 SQLite 数据库:${MONITOR_DB}"
sqlite_exec "CREATE TABLE IF NOT EXISTS db_metrics (
ts INTEGER PRIMARY KEY,
threads_running INTEGER,
threads_connected INTEGER,
threads_idle INTEGER,
uptime_sec INTEGER,
qps_read INTEGER,
qps_write INTEGER,
slow_queries INTEGER -- 近 30s 慢查询数(>30s)
);"
sqlite_exec "CREATE TABLE IF NOT EXISTS db_processlist (
id INTEGER PRIMARY KEY AUTOINCREMENT,
ts INTEGER,
pid INTEGER,
user TEXT,
state TEXT,
time_sec INTEGER,
info TEXT
);"
sqlite_exec "CREATE INDEX IF NOT EXISTS idx_db_proc_ts ON db_processlist(ts);"
sqlite_exec "CREATE INDEX IF NOT EXISTS idx_db_proc_time ON db_processlist(time_sec);"
sqlite_exec "CREATE TABLE IF NOT EXISTS http_metrics (
id INTEGER PRIMARY KEY AUTOINCREMENT,
ts INTEGER,
endpoint TEXT,
status_code INTEGER,
resp_time_ms INTEGER,
healthy INTEGER -- 0/1
);"
sqlite_exec "CREATE INDEX IF NOT EXISTS idx_http_ts_ep ON http_metrics(ts, endpoint);"
sqlite_exec "CREATE TABLE IF NOT EXISTS business_metrics (
ts INTEGER PRIMARY KEY,
msg_log_hourly INTEGER, -- 本小时 ecology_message_log 增量
msg_log_fail_rate REAL, -- 失败率
ongoing_requests INTEGER, -- 进行中流程数
overdue_tasks INTEGER -- 超时待办数
);"
sqlite_exec "CREATE TABLE IF NOT EXISTS infra_metrics (
ts INTEGER PRIMARY KEY,
docker_daemon_ok INTEGER, -- 0/1
docker_container_count INTEGER,
docker_root_write_mb REAL, -- 1 分钟 Docker data-root 写入量 MB
disk_usage_pct REAL,
inode_usage_pct REAL,
load_avg REAL
);"
sqlite_exec "CREATE TABLE IF NOT EXISTS alerts (
id INTEGER PRIMARY KEY AUTOINCREMENT,
ts INTEGER,
level TEXT, -- P0/P1/P2
category TEXT, -- db/http/business/infra
title TEXT,
detail TEXT,
ack INTEGER DEFAULT 0, -- 0=未确认 1=已确认
resolved INTEGER DEFAULT 0 -- 0=未恢复 1=已恢复
);"
sqlite_exec "CREATE INDEX IF NOT EXISTS idx_alerts_ts ON alerts(ts);"
sqlite_exec "CREATE INDEX IF NOT EXISTS idx_alerts_ack ON alerts(ack, resolved);"
sqlite_exec "CREATE TABLE IF NOT EXISTS alert_cooldown (
alert_key TEXT PRIMARY KEY,
last_alert_ts INTEGER
);"
log_info "SQLite schema 初始化完成"
+98
View File
@@ -0,0 +1,98 @@
#!/usr/bin/env bash
# ============================================================================
# OA Monitor · Resin 凌晨预防性重启
# 每天 02:00 执行,重建 push 线程池,静默死亡前主动清场
# crontab: 0 2 * * * /opt/oa-monitor/bin/resin-preventive-restart.sh >> /opt/oa-monitor/logs/resin-restart.log 2>&1
# ============================================================================
set -u
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
source "${SCRIPT_DIR}/../lib/common.sh"
OA_HOST="39.99.236.138"
OA_USER="root"
OA_PASS="KTOFKjVkVOGlNgvs8yUP1+PM1VM"
LOG_FILE="${MONITOR_LOGS}/resin-preventive-restart.log"
OA_SSH_PY="${MONITOR_ROOT}/bin/_oa_ssh.py"
START_TS=$(date '+%Y-%m-%d %H:%M:%S')
echo "=== ${START_TS} Resin 预防性重启开始 ===" | tee -a "${LOG_FILE}"
# Step 1: 写 _oa_ssh.py(paramiko + errors=replace,复用 watchdog 方式)
cat > "${OA_SSH_PY}" << 'PYEOF'
import paramiko, sys
host, user, password, cmd = sys.argv[1], sys.argv[2], sys.argv[3], sys.argv[4]
ssh = paramiko.SSHClient()
ssh.set_missing_host_key_policy(paramiko.AutoAddPolicy())
try:
ssh.connect(host, username=user, password=password, timeout=15)
stdin, stdout, stderr = ssh.exec_command(cmd, timeout=180)
sys.stdout.write(stdout.read().decode("utf-8", errors="replace"))
sys.stderr.write(stderr.read().decode("utf-8", errors="replace"))
ssh.close()
except Exception as e:
print(f"SSH_ERROR: {e}", file=sys.stderr)
sys.exit(1)
PYEOF
chmod +x "${OA_SSH_PY}"
OA_SSH="python3 ${OA_SSH_PY} ${OA_HOST} ${OA_USER} ${OA_PASS}"
# Step 2: 钉钉提前通知
dingtalk_alert "P1" "Resin 凌晨预防性重启(开始)" \
"**🛠️ Resin 预防性重启开始**\n\n时间:${START_TS}\n\n> 每天凌晨 02:00 执行,重建推送线程池,防止静默死亡。停机窗口约 30-60 秒。"
# Step 3: 记录旧 PID
OLD_PID=$(${OA_SSH} "ps -eo pid,cmd | grep 'java.*Resin.*resin.xml' | grep -v grep | awk '{print \$1}' | head -1" 2>/dev/null)
echo "旧 PID=${OLD_PID}" | tee -a "${LOG_FILE}"
# Step 4: 停 Resin
echo "--- stopresin.sh ---" | tee -a "${LOG_FILE}"
${OA_SSH} "bash /usr/weaver/Resin/bin/stopresin.sh 2>&1" | tee -a "${LOG_FILE}"
# Step 5: 等端口关闭(最多 30 秒)
for i in $(seq 1 30); do
NEW_CHECK=$(${OA_SSH} "curl -s -o /dev/null -w '%{http_code}' http://127.0.0.1:80/ 2>/dev/null" 2>/dev/null)
if [[ "${NEW_CHECK}" != "200" && "${NEW_CHECK}" != "404" && "${NEW_CHECK}" != "401" ]]; then
echo "[${i}s] 80 端口已关闭 (HTTP=${NEW_CHECK})" | tee -a "${LOG_FILE}"
break
fi
sleep 1
done
sleep 3
# Step 6: 起 Resin
echo "--- startresin.sh ---" | tee -a "${LOG_FILE}"
${OA_SSH} "bash /usr/weaver/Resin/bin/startresin.sh 2>&1" | tee -a "${LOG_FILE}"
# Step 7: 等 Resin 恢复(最多 90 秒)
HTTP_CODE="000"
for i in $(seq 1 90); do
HTTP_CODE=$(${OA_SSH} "curl -s -o /dev/null -w '%{http_code}' http://127.0.0.1:80/ 2>/dev/null" 2>/dev/null)
if [[ "${HTTP_CODE}" == "200" || "${HTTP_CODE}" == "404" || "${HTTP_CODE}" == "401" || "${HTTP_CODE}" == "500" ]]; then
echo "[${i}s] Resin 恢复 (HTTP=${HTTP_CODE})" | tee -a "${LOG_FILE}"
break
fi
sleep 1
done
# Step 8: 记录新 PID
NEW_PID=$(${OA_SSH} "ps -eo pid,cmd | grep 'java.*Resin.*resin.xml' | grep -v grep | awk '{print \$1}' | head -1" 2>/dev/null)
END_TS=$(date '+%Y-%m-%d %H:%M:%S')
# Step 9: 钉钉结果通知
BODY="**🛠️ Resin 预防性重启完成**\n\n"
BODY="${BODY}| 指标 | 值 |\n|---|---|\n"
BODY="${BODY}| 旧 PID | \`${OLD_PID}\` |\n"
BODY="${BODY}| 新 PID | \`${NEW_PID}\` |\n"
BODY="${BODY}| HTTP 80 端口 | \`${HTTP_CODE}\` |\n"
BODY="${BODY}| 耗时 | $(($(date +%s) - $(date -d "${START_TS}" +%s 2>/dev/null || date +%s)))秒 |\n"
BODY="${BODY}| 结束时间 | ${END_TS} |\n"
if [[ "${NEW_PID}" == "${OLD_PID}" ]]; then
BODY="${BODY}\n> ⚠️ 新旧 PID 相同,重启可能未生效,请检查!"
dingtalk_alert "P1" "Resin 预防性重启(⚠️ PID 未变)" "${BODY}"
else
dingtalk_alert "P1" "Resin 预防性重启(✅ 成功)" "${BODY}"
fi
echo "=== ${END_TS} Resin 预防性重启结束(${OLD_PID}→${NEW_PID}, HTTP=${HTTP_CODE}) ===" | tee -a "${LOG_FILE}"
+146
View File
@@ -0,0 +1,146 @@
#!/usr/bin/env bash
# ============================================================================
# OA Monitor · Resin Watchdog (v4 - PID 变化检测 + message_log 入库判定)
# v3 盲区修复:Resin 被 WatchdogManager kill 后秒拉起,message_log 继续刷看起来"健康"
# v4 新增:每次巡检记录 Resin PID,PID 变化 -> 钉钉 P1 告警
# 健康依据(双重):
# 1. ecology_message_log 最近 15 分钟入库 count > 0(推送线程活着)
# 2. Resin PID 未变化(进程未被 kill 重启)
# crontab: */5 * * * * /opt/oa-monitor/bin/resin-watchdog.sh
# ============================================================================
set -u
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
source "${SCRIPT_DIR}/../lib/common.sh"
OA_HOST="39.99.236.138"
OA_USER="root"
OA_PASS="KTOFKjVkVOGlNgvs8yUP1+PM1VM"
OA_RESTART_CMD="bash /usr/weaver/Resin/bin/stopresin.sh 2>&1; sleep 5; bash /usr/weaver/Resin/bin/startresin.sh 2>&1"
MSG_WINDOW_MIN=15
START_HOUR=8
END_HOUR=22
PID_FILE="${MONITOR_ROOT}/state/resin_pid.last"
LOG_FILE="${MONITOR_LOGS}/resin-watchdog.log"
OA_SSH_PY="${MONITOR_ROOT}/bin/_oa_ssh.py"
# --- 依赖检查 ---
if ! python3 -c 'import paramiko' 2>/dev/null; then
MSG="[WATCHDOG] paramiko 不可用!"
log_error "${MSG}"
dingtalk_alert "P0" "Resin Watchdog 无法运行" "${MSG}"
exit 1
fi
# --- SSH helper ---
cat > "${OA_SSH_PY}" << 'PYEOF'
import paramiko, sys
host, user, password, cmd = sys.argv[1], sys.argv[2], sys.argv[3], sys.argv[4]
ssh = paramiko.SSHClient()
ssh.set_missing_host_key_policy(paramiko.AutoAddPolicy())
try:
ssh.connect(host, username=user, password=password, timeout=10)
stdin, stdout, stderr = ssh.exec_command(cmd)
sys.stdout.write(stdout.read().decode("utf-8", errors="replace"))
ssh.close()
except Exception as e:
print(f"SSH_ERROR: {e}", file=sys.stderr)
sys.exit(1)
PYEOF
chmod +x "${OA_SSH_PY}"
OA_SSH="python3 ${OA_SSH_PY} ${OA_HOST} ${OA_USER} ${OA_PASS}"
# ============================================================================
# Step 1: PID 变化检测(每次都跑,不依赖 message_log 结果)
# ============================================================================
mkdir -p "$(dirname "${PID_FILE}")"
PID_INFO=$(${OA_SSH} "ps -eo pid,etimes,cmd | grep 'java.*-Dresin.home.*/Resin/' | grep -v watchdog | grep -v grep | awk '{print \$1, \$2}' | head -1" 2>/dev/null)
CUR_PID=$(echo "${PID_INFO}" | awk '{print $1}')
CUR_ETIMES=$(echo "${PID_INFO}" | awk '{print $2}')
if [[ ! "${CUR_PID}" =~ ^[0-9]+$ ]]; then
log_warn "[WATCHDOG] 无法获取 Resin PID, 跳过 PID 检测"
else
LAST_PID=""
LAST_ETIMES=""
if [[ -f "${PID_FILE}" ]]; then
read -r LAST_PID LAST_ETIMES _ < "${PID_FILE}" 2>/dev/null || true
fi
if [[ -z "${LAST_PID}" ]]; then
echo "${CUR_PID} ${CUR_ETIMES} $(date +%s)" > "${PID_FILE}"
log_info "[WATCHDOG] PID 基线初始化: Resin PID=${CUR_PID}, 已运行 ${CUR_ETIMES}s"
elif [[ "${LAST_PID}" != "${CUR_PID}" ]]; then
NOW_HOUR=$(date +%H)
NOW_MIN=$(date +%M)
IS_PREVENTIVE=0
if [[ "${NOW_HOUR}" == "02" && "${NOW_MIN}" -ge 0 && "${NOW_MIN}" -le 10 ]]; then
IS_PREVENTIVE=1
fi
if [[ "${IS_PREVENTIVE}" -eq 1 ]]; then
log_info "[WATCHDOG] PID 变化在预防性重启窗口(02:00-02:10): ${LAST_PID} -> ${CUR_PID}, 正常"
else
RESTART_TIME=$(date -d "@$(($(date +%s) - CUR_ETIMES))" '+%Y-%m-%d %H:%M:%S' 2>/dev/null || date '+%Y-%m-%d %H:%M:%S')
BODY="Resin PID 异常变化(非预期重启)\n\n"
BODY="${BODY}上次 PID: ${LAST_PID}\n"
BODY="${BODY}新 PID: ${CUR_PID}\n"
BODY="${BODY}推算重启时间: ${RESTART_TIME}\n"
BODY="${BODY}新进程已运行: ${CUR_ETIMES} 秒\n\n"
BODY="${BODY}Resin 被 WatchdogManager kill 后拉起。message_log 可能仍在刷(假活跃), watchdog v3 会漏检。"
dingtalk_alert "P1" "Resin PID 异常变化: ${LAST_PID} -> ${CUR_PID} (推算 ${RESTART_TIME})" "${BODY}"
log_warn "[WATCHDOG] PID 异常变化: ${LAST_PID} -> ${CUR_PID} (推算重启 ${RESTART_TIME}), 已发 P1 告警"
fi
echo "${CUR_PID} ${CUR_ETIMES} $(date +%s)" > "${PID_FILE}"
else
echo "${CUR_PID} ${CUR_ETIMES} $(date +%s)" > "${PID_FILE}"
fi
fi
# ============================================================================
# Step 2: message_log 入库检测 (原 v3 逻辑)
# ============================================================================
MSG_COUNT=$(MYSQL_PWD="${DB_PASS}" mysql -h39.100.72.102 -P3306 -uecology "${DB_NAME}" \
-sN -e "SELECT COUNT(*) FROM ecology_message_log WHERE CREATEDATE=CURDATE() AND CREATETIME >= DATE_FORMAT(NOW() - INTERVAL ${MSG_WINDOW_MIN} MINUTE, '%H:%i:%s');" 2>/dev/null)
if [[ ! "${MSG_COUNT}" =~ ^[0-9]+$ ]]; then
log_warn "[WATCHDOG] 查询 message_log 失败, 跳过"
exit 0
fi
log_info "[WATCHDOG] 最近 ${MSG_WINDOW_MIN} 分钟 message_log 入库 ${MSG_COUNT} 条"
HOUR=$(date +%H)
if [[ ${MSG_COUNT} -gt 0 ]]; then
log_info "[WATCHDOG] message_log 入库正常(${MSG_COUNT} > 0), 推送健康"
exit 0
fi
if [[ ${HOUR} -lt ${START_HOUR} || ${HOUR} -ge ${END_HOUR} ]]; then
log_warn "[WATCHDOG] 凌晨时段(${HOUR}时), message_log 仅 ${MSG_COUNT} 条, 仅记录"
exit 0
fi
# --- 工作时间 + 入库过少 -> 触发重启 ---
log_warn "[WATCHDOG] 最近 ${MSG_WINDOW_MIN} 分钟 message_log count=0, 推送线程可能已死, 触发 Resin 重启!"
${OA_SSH} "${OA_RESTART_CMD}" 2>&1 | tee -a "${LOG_FILE}"
sleep 20
NEW_PID=$(${OA_SSH} "ps -eo pid,cmd | grep 'java.*Resin.*resin.xml' | grep -v grep | awk '{print \$1}'" 2>/dev/null | head -1)
HTTP_CODE=$(${OA_SSH} "curl -s -o /dev/null -w '%{http_code}' http://127.0.0.1:80/wui/main/Login.jsp" 2>/dev/null)
BODY="Resin watchdog 自动重启\n\n"
BODY="${BODY}message_log 最近 ${MSG_WINDOW_MIN} 分钟入库: ${MSG_COUNT} 条 (count > 0)\n"
BODY="${BODY}新 Resin PID: ${NEW_PID}\n"
BODY="${BODY}HTTP 80: ${HTTP_CODE}\n"
dingtalk_alert "P0" "Resin watchdog 自动重启(message_log count=0)" "${BODY}"
if [[ "${NEW_PID}" =~ ^[0-9]+$ ]]; then
echo "${NEW_PID} 0 $(date +%s)" > "${PID_FILE}"
fi
log_info "[WATCHDOG] Resin 重启完成, 新 PID=${NEW_PID}, HTTP=${HTTP_CODE}"
+84
View File
@@ -0,0 +1,84 @@
#!/usr/bin/env bash
# ============================================================================
# OA Monitor · Resin Watchdog
# 监控 Util_Message 线程活跃度(publishMessageList),超过 30 分钟无活动则自动重启 Resin
# 部署在 DSH 上,通过 sshpass 到 OA 服务器(39.99.236.138)执行
# crontab: */5 * * * * /opt/oa-monitor/bin/resin-watchdog.sh
# ============================================================================
set -u
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
source "${SCRIPT_DIR}/../lib/common.sh"
OA_HOST="39.99.236.138"
OA_USER="root"
OA_PASS="KTOFKjVkVOGlNgvs8yUP1+PM1VM"
OA_LOG="/usr/weaver/Resin/log/stdout.log"
OA_RESTART_CMD="bash /usr/weaver/Resin/bin/stopresin.sh 2>&1; sleep 5; bash /usr/weaver/Resin/bin/startresin.sh 2>&1"
MAX_IDLE_MIN=30 # 30 分钟没 publishMessageList 活动就重启
RESIN_PATTERN="com.cloudstore.dev.api.util.Util_Message.*publishMessageList is start"
NOW=$(date +%s)
# --- 1. sshpass 可用性检查 ---
if ! command -v sshpass &>/dev/null; then
log_warn "[WATCHDOG] sshpass 未安装,尝试安装..."
ssh root@${OA_HOST} "which sshpass 2>/dev/null" && SSHPASS_OK=true || SSHPASS_OK=false
fi
# --- 2. ssh 到 OA 服务器查 publishMessageList 最后活动时间 ---
LAST_TS=$(sshpass -p "${OA_PASS}" ssh -o StrictHostKeyChecking=no -o ConnectTimeout=10 ${OA_USER}@${OA_HOST} "
# 从 stdout.log 最后 500 行找 publishMessageList
tail -5000 ${OA_LOG} 2>/dev/null | grep -E \"${RESIN_PATTERN}\" | tail -1 | grep -oP '\\d{4}\\.\\d{2}\\.\\d{2} \\d{2}:\\d{2}:\\d{2}' | head -1
" 2>/dev/null)
# 如果 stdout.log 没有,查 yesterday.log
if [[ -z "${LAST_TS}" ]]; then
LAST_TS=$(sshpass -p "${OA_PASS}" ssh -o StrictHostKeyChecking=no -o ConnectTimeout=10 ${OA_USER}@${OA_HOST} "
find /usr/weaver/Resin/log -name '*.log*' -newer ${OA_LOG}.yesterday 2>/dev/null | head -3 | xargs grep -lE '${RESIN_PATTERN}' 2>/dev/null | head -1 | xargs grep -E '${RESIN_PATTERN}' 2>/dev/null | tail -1 | grep -oP '\\d{4}\\.\\d{2}\\.\\d{2} \\d{2}:\\d{2}:\\d{2}' | head -1
" 2>/dev/null)
fi
# --- 3. 判断是否需要重启 ---
if [[ -z "${LAST_TS}" ]]; then
# 完全没找到 publishMessageList 记录——可能是新部署,不要贸然重启
log_warn "[WATCHDOG] ⚠️ stdout.log 里没找到 publishMessageList 记录,可能是新部署。跳过本次。"
exit 0
fi
# 把日志时间转成 epoch
LAST_EPOCH=$(date -j -f "%Y.%m.%d %H:%M:%S" "${LAST_TS}" +%s 2>/dev/null || date -d "${LAST_TS}" +%s 2>/dev/null)
if [[ -z "${LAST_EPOCH}" ]]; then
log_warn "[WATCHDOG] ⚠️ 无法解析日志时间: ${LAST_TS},跳过本次。"
exit 0
fi
IDLE_MIN=$(( (NOW - LAST_EPOCH) / 60 ))
log_info "[WATCHDOG] 最后 publishMessageList: ${LAST_TS},已空闲 ${IDLE_MIN} 分钟(阈值 ${MAX_IDLE_MIN} 分钟)"
if [[ ${IDLE_MIN} -ge ${MAX_IDLE_MIN} ]]; then
log_warn "[WATCHDOG] 🔴 空闲 ${IDLE_MIN} 分钟 ≥ ${MAX_IDLE_MIN} 分钟!触发 Resin 重启!"
# --- 4. 执行重启 ---
sshpass -p "${OA_PASS}" ssh -o StrictHostKeyChecking=no -o ConnectTimeout=15 ${OA_USER}@${OA_HOST} "${OA_RESTART_CMD}" 2>&1 | tee -a "${LOG_FILE}"
sleep 20
# --- 5. 验证新 Resin PID ---
NEW_PID=$(sshpass -p "${OA_PASS}" ssh -o StrictHostKeyChecking=no -o ConnectTimeout=10 ${OA_USER}@${OA_HOST} "ps -eo pid,cmd | grep 'java.*Resin.*resin.xml' | grep -v grep | awk '{print \$1}'" 2>/dev/null)
HTTP_CODE=$(sshpass -p "${OA_PASS}" ssh -o StrictHostKeyChecking=no -o ConnectTimeout=10 ${OA_USER}@${OA_HOST} "curl -s -o /dev/null -w '%{http_code}' http://127.0.0.1:80/wui/main/Login.jsp 2>/dev/null" 2>/dev/null)
BODY="**🔴 Resin 被 watchdog 自动重启**\n\n"
BODY="${BODY}| 指标 | 值 |\n|---|---|\n"
BODY="${BODY}| publishMessageList 最后活动 | \`${LAST_TS}\` |\n"
BODY="${BODY}| 空闲时长 | **${IDLE_MIN} 分钟** |\n"
BODY="${BODY}| 新 Resin PID | \`${NEW_PID}\` |\n"
BODY="${BODY}| HTTP 80 端口 | \`${HTTP_CODE}\` |\n\n"
BODY="${BODY}> ⚠️ 推送线程静默死掉,watchdog 自动恢复。\n"
BODY="${BODY}> 建议:泛微需要检查 Util_Message 线程为什么会静默挂掉。"
dingtalk_alert "P0" "Resin watchdog 自动重启(推送线程空闲 ${IDLE_MIN} 分钟)" "${BODY}"
log_info "[WATCHDOG] ✅ Resin 重启完成,新 PID=${NEW_PID}, HTTP=${HTTP_CODE}"
else
log_info "[WATCHDOG] ✅ publishMessageList 活跃(最近 ${IDLE_MIN} 分钟前),Resin 健康"
fi
+147
View File
@@ -0,0 +1,147 @@
#!/usr/bin/env bash
# ============================================================================
# OA Monitor · Resin Watchdog (v4 - PID 变化检测 + message_log 入库判定)
# v3 盲区修复:Resin 被 WatchdogManager kill 后秒拉起,message_log 继续刷看起来"健康"
# v4 新增:每次巡检记录 Resin PID,PID 变化 -> 钉钉 P1 告警
# 健康依据(双重):
# 1. ecology_message_log 最近 15 分钟入库 >= 10 条(推送线程活着)
# 2. Resin PID 未变化(进程未被 kill 重启)
# crontab: */5 * * * * /opt/oa-monitor/bin/resin-watchdog.sh
# ============================================================================
set -u
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
source "${SCRIPT_DIR}/../lib/common.sh"
OA_HOST="39.99.236.138"
OA_USER="root"
OA_PASS="KTOFKjVkVOGlNgvs8yUP1+PM1VM"
OA_RESTART_CMD="bash /usr/weaver/Resin/bin/stopresin.sh 2>&1; sleep 5; bash /usr/weaver/Resin/bin/startresin.sh 2>&1"
MSG_WINDOW_MIN=15
MSG_THRESHOLD=1
START_HOUR=8
END_HOUR=22
PID_FILE="${MONITOR_ROOT}/state/resin_pid.last"
LOG_FILE="${MONITOR_LOGS}/resin-watchdog.log"
OA_SSH_PY="${MONITOR_ROOT}/bin/_oa_ssh.py"
# --- 依赖检查 ---
if ! python3 -c 'import paramiko' 2>/dev/null; then
MSG="[WATCHDOG] paramiko 不可用!"
log_error "${MSG}"
dingtalk_alert "P0" "Resin Watchdog 无法运行" "${MSG}"
exit 1
fi
# --- SSH helper ---
cat > "${OA_SSH_PY}" << 'PYEOF'
import paramiko, sys
host, user, password, cmd = sys.argv[1], sys.argv[2], sys.argv[3], sys.argv[4]
ssh = paramiko.SSHClient()
ssh.set_missing_host_key_policy(paramiko.AutoAddPolicy())
try:
ssh.connect(host, username=user, password=password, timeout=10)
stdin, stdout, stderr = ssh.exec_command(cmd)
sys.stdout.write(stdout.read().decode("utf-8", errors="replace"))
ssh.close()
except Exception as e:
print(f"SSH_ERROR: {e}", file=sys.stderr)
sys.exit(1)
PYEOF
chmod +x "${OA_SSH_PY}"
OA_SSH="python3 ${OA_SSH_PY} ${OA_HOST} ${OA_USER} ${OA_PASS}"
# ============================================================================
# Step 1: PID 变化检测(每次都跑,不依赖 message_log 结果)
# ============================================================================
mkdir -p "$(dirname "${PID_FILE}")"
PID_INFO=$(${OA_SSH} "ps -eo pid,etimes,cmd | grep 'java.*-Dresin.home.*/Resin/' | grep -v watchdog | grep -v grep | awk '{print \$1, \$2}' | head -1" 2>/dev/null)
CUR_PID=$(echo "${PID_INFO}" | awk '{print $1}')
CUR_ETIMES=$(echo "${PID_INFO}" | awk '{print $2}')
if [[ ! "${CUR_PID}" =~ ^[0-9]+$ ]]; then
log_warn "[WATCHDOG] 无法获取 Resin PID, 跳过 PID 检测"
else
LAST_PID=""
LAST_ETIMES=""
if [[ -f "${PID_FILE}" ]]; then
read -r LAST_PID LAST_ETIMES _ < "${PID_FILE}" 2>/dev/null || true
fi
if [[ -z "${LAST_PID}" ]]; then
echo "${CUR_PID} ${CUR_ETIMES} $(date +%s)" > "${PID_FILE}"
log_info "[WATCHDOG] PID 基线初始化: Resin PID=${CUR_PID}, 已运行 ${CUR_ETIMES}s"
elif [[ "${LAST_PID}" != "${CUR_PID}" ]]; then
NOW_HOUR=$(date +%H)
NOW_MIN=$(date +%M)
IS_PREVENTIVE=0
if [[ "${NOW_HOUR}" == "02" && "${NOW_MIN}" -ge 0 && "${NOW_MIN}" -le 10 ]]; then
IS_PREVENTIVE=1
fi
if [[ "${IS_PREVENTIVE}" -eq 1 ]]; then
log_info "[WATCHDOG] PID 变化在预防性重启窗口(02:00-02:10): ${LAST_PID} -> ${CUR_PID}, 正常"
else
RESTART_TIME=$(date -d "@$(($(date +%s) - CUR_ETIMES))" '+%Y-%m-%d %H:%M:%S' 2>/dev/null || date '+%Y-%m-%d %H:%M:%S')
BODY="Resin PID 异常变化(非预期重启)\n\n"
BODY="${BODY}上次 PID: ${LAST_PID}\n"
BODY="${BODY}新 PID: ${CUR_PID}\n"
BODY="${BODY}推算重启时间: ${RESTART_TIME}\n"
BODY="${BODY}新进程已运行: ${CUR_ETIMES} 秒\n\n"
BODY="${BODY}Resin 被 WatchdogManager kill 后拉起。message_log 可能仍在刷(假活跃), watchdog v3 会漏检。"
dingtalk_alert "P1" "Resin PID 异常变化: ${LAST_PID} -> ${CUR_PID} (推算 ${RESTART_TIME})" "${BODY}"
log_warn "[WATCHDOG] PID 异常变化: ${LAST_PID} -> ${CUR_PID} (推算重启 ${RESTART_TIME}), 已发 P1 告警"
fi
echo "${CUR_PID} ${CUR_ETIMES} $(date +%s)" > "${PID_FILE}"
else
echo "${CUR_PID} ${CUR_ETIMES} $(date +%s)" > "${PID_FILE}"
fi
fi
# ============================================================================
# Step 2: message_log 入库检测 (原 v3 逻辑)
# ============================================================================
MSG_COUNT=$(MYSQL_PWD="${DB_PASS}" mysql -h39.100.72.102 -P3306 -uecology "${DB_NAME}" \
-sN -e "SELECT COUNT(*) FROM ecology_message_log WHERE CREATEDATE=CURDATE() AND CREATETIME >= DATE_FORMAT(NOW() - INTERVAL ${MSG_WINDOW_MIN} MINUTE, '%H:%i:%s');" 2>/dev/null)
if [[ ! "${MSG_COUNT}" =~ ^[0-9]+$ ]]; then
log_warn "[WATCHDOG] 查询 message_log 失败, 跳过"
exit 0
fi
log_info "[WATCHDOG] 最近 ${MSG_WINDOW_MIN} 分钟 message_log 入库 ${MSG_COUNT} 条"
HOUR=$(date +%H)
if [[ ${MSG_COUNT} -ge ${MSG_THRESHOLD} ]]; then
log_info "[WATCHDOG] message_log 入库正常(${MSG_COUNT} >= ${MSG_THRESHOLD}), 推送健康"
exit 0
fi
if [[ ${HOUR} -lt ${START_HOUR} || ${HOUR} -ge ${END_HOUR} ]]; then
log_warn "[WATCHDOG] 凌晨时段(${HOUR}时), message_log 仅 ${MSG_COUNT} 条, 仅记录"
exit 0
fi
# --- 工作时间 + 入库过少 -> 触发重启 ---
log_warn "[WATCHDOG] 最近 ${MSG_WINDOW_MIN} 分钟 message_log 仅 ${MSG_COUNT} 条 < ${MSG_THRESHOLD}, 触发 Resin 重启!"
${OA_SSH} "${OA_RESTART_CMD}" 2>&1 | tee -a "${LOG_FILE}"
sleep 20
NEW_PID=$(${OA_SSH} "ps -eo pid,cmd | grep 'java.*Resin.*resin.xml' | grep -v grep | awk '{print \$1}'" 2>/dev/null | head -1)
HTTP_CODE=$(${OA_SSH} "curl -s -o /dev/null -w '%{http_code}' http://127.0.0.1:80/wui/main/Login.jsp" 2>/dev/null)
BODY="Resin watchdog 自动重启\n\n"
BODY="${BODY}message_log 最近 ${MSG_WINDOW_MIN} 分钟入库: ${MSG_COUNT} 条 (阈值 ${MSG_THRESHOLD})\n"
BODY="${BODY}新 Resin PID: ${NEW_PID}\n"
BODY="${BODY}HTTP 80: ${HTTP_CODE}\n"
dingtalk_alert "P0" "Resin watchdog 自动重启(message_log 入库 ${MSG_COUNT} 条 < ${MSG_THRESHOLD})" "${BODY}"
if [[ "${NEW_PID}" =~ ^[0-9]+$ ]]; then
echo "${NEW_PID} 0 $(date +%s)" > "${PID_FILE}"
fi
log_info "[WATCHDOG] Resin 重启完成, 新 PID=${NEW_PID}, HTTP=${HTTP_CODE}"
+147
View File
@@ -0,0 +1,147 @@
#!/usr/bin/env bash
# ============================================================================
# OA Monitor · Resin Watchdog (v4 - PID 变化检测 + message_log 入库判定)
# v3 盲区修复:Resin 被 WatchdogManager kill 后秒拉起,message_log 继续刷看起来"健康"
# v4 新增:每次巡检记录 Resin PID,PID 变化 -> 钉钉 P1 告警
# 健康依据(双重):
# 1. ecology_message_log 最近 15 分钟入库 >= 10 条(推送线程活着)
# 2. Resin PID 未变化(进程未被 kill 重启)
# crontab: */5 * * * * /opt/oa-monitor/bin/resin-watchdog.sh
# ============================================================================
set -u
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
source "${SCRIPT_DIR}/../lib/common.sh"
OA_HOST="39.99.236.138"
OA_USER="root"
OA_PASS="KTOFKjVkVOGlNgvs8yUP1+PM1VM"
OA_RESTART_CMD="bash /usr/weaver/Resin/bin/stopresin.sh 2>&1; sleep 5; bash /usr/weaver/Resin/bin/startresin.sh 2>&1"
MSG_WINDOW_MIN=15
MSG_THRESHOLD=10
START_HOUR=8
END_HOUR=22
PID_FILE="${MONITOR_ROOT}/state/resin_pid.last"
LOG_FILE="${MONITOR_LOGS}/resin-watchdog.log"
OA_SSH_PY="${MONITOR_ROOT}/bin/_oa_ssh.py"
# --- 依赖检查 ---
if ! python3 -c 'import paramiko' 2>/dev/null; then
MSG="[WATCHDOG] paramiko 不可用!"
log_error "${MSG}"
dingtalk_alert "P0" "Resin Watchdog 无法运行" "${MSG}"
exit 1
fi
# --- SSH helper ---
cat > "${OA_SSH_PY}" << 'PYEOF'
import paramiko, sys
host, user, password, cmd = sys.argv[1], sys.argv[2], sys.argv[3], sys.argv[4]
ssh = paramiko.SSHClient()
ssh.set_missing_host_key_policy(paramiko.AutoAddPolicy())
try:
ssh.connect(host, username=user, password=password, timeout=10)
stdin, stdout, stderr = ssh.exec_command(cmd)
sys.stdout.write(stdout.read().decode("utf-8", errors="replace"))
ssh.close()
except Exception as e:
print(f"SSH_ERROR: {e}", file=sys.stderr)
sys.exit(1)
PYEOF
chmod +x "${OA_SSH_PY}"
OA_SSH="python3 ${OA_SSH_PY} ${OA_HOST} ${OA_USER} ${OA_PASS}"
# ============================================================================
# Step 1: PID 变化检测(每次都跑,不依赖 message_log 结果)
# ============================================================================
mkdir -p "$(dirname "${PID_FILE}")"
PID_INFO=$(${OA_SSH} "ps -eo pid,etimes,cmd | grep 'java.*-Dresin.home.*/Resin/' | grep -v watchdog | grep -v grep | awk '{print \$1, \$2}' | head -1" 2>/dev/null)
CUR_PID=$(echo "${PID_INFO}" | awk '{print $1}')
CUR_ETIMES=$(echo "${PID_INFO}" | awk '{print $2}')
if [[ ! "${CUR_PID}" =~ ^[0-9]+$ ]]; then
log_warn "[WATCHDOG] 无法获取 Resin PID, 跳过 PID 检测"
else
LAST_PID=""
LAST_ETIMES=""
if [[ -f "${PID_FILE}" ]]; then
read -r LAST_PID LAST_ETIMES _ < "${PID_FILE}" 2>/dev/null || true
fi
if [[ -z "${LAST_PID}" ]]; then
echo "${CUR_PID} ${CUR_ETIMES} $(date +%s)" > "${PID_FILE}"
log_info "[WATCHDOG] PID 基线初始化: Resin PID=${CUR_PID}, 已运行 ${CUR_ETIMES}s"
elif [[ "${LAST_PID}" != "${CUR_PID}" ]]; then
NOW_HOUR=$(date +%H)
NOW_MIN=$(date +%M)
IS_PREVENTIVE=0
if [[ "${NOW_HOUR}" == "02" && "${NOW_MIN}" -ge 0 && "${NOW_MIN}" -le 10 ]]; then
IS_PREVENTIVE=1
fi
if [[ "${IS_PREVENTIVE}" -eq 1 ]]; then
log_info "[WATCHDOG] PID 变化在预防性重启窗口(02:00-02:10): ${LAST_PID} -> ${CUR_PID}, 正常"
else
RESTART_TIME=$(date -d "@$(($(date +%s) - CUR_ETIMES))" '+%Y-%m-%d %H:%M:%S' 2>/dev/null || date '+%Y-%m-%d %H:%M:%S')
BODY="Resin PID 异常变化(非预期重启)\n\n"
BODY="${BODY}上次 PID: ${LAST_PID}\n"
BODY="${BODY}新 PID: ${CUR_PID}\n"
BODY="${BODY}推算重启时间: ${RESTART_TIME}\n"
BODY="${BODY}新进程已运行: ${CUR_ETIMES} 秒\n\n"
BODY="${BODY}Resin 被 WatchdogManager kill 后拉起。message_log 可能仍在刷(假活跃), watchdog v3 会漏检。"
dingtalk_alert "P1" "Resin PID 异常变化: ${LAST_PID} -> ${CUR_PID} (推算 ${RESTART_TIME})" "${BODY}"
log_warn "[WATCHDOG] PID 异常变化: ${LAST_PID} -> ${CUR_PID} (推算重启 ${RESTART_TIME}), 已发 P1 告警"
fi
echo "${CUR_PID} ${CUR_ETIMES} $(date +%s)" > "${PID_FILE}"
else
echo "${CUR_PID} ${CUR_ETIMES} $(date +%s)" > "${PID_FILE}"
fi
fi
# ============================================================================
# Step 2: message_log 入库检测 (原 v3 逻辑)
# ============================================================================
MSG_COUNT=$(MYSQL_PWD="${DB_PASS}" mysql -h39.100.72.102 -P3306 -uecology "${DB_NAME}" \
-sN -e "SELECT COUNT(*) FROM ecology_message_log WHERE CREATEDATE=CURDATE() AND CREATETIME >= DATE_FORMAT(NOW() - INTERVAL ${MSG_WINDOW_MIN} MINUTE, '%H:%i:%s');" 2>/dev/null)
if [[ ! "${MSG_COUNT}" =~ ^[0-9]+$ ]]; then
log_warn "[WATCHDOG] 查询 message_log 失败, 跳过"
exit 0
fi
log_info "[WATCHDOG] 最近 ${MSG_WINDOW_MIN} 分钟 message_log 入库 ${MSG_COUNT} 条"
HOUR=$(date +%H)
if [[ ${MSG_COUNT} -ge ${MSG_THRESHOLD} ]]; then
log_info "[WATCHDOG] message_log 入库正常(${MSG_COUNT} >= ${MSG_THRESHOLD}), 推送健康"
exit 0
fi
if [[ ${HOUR} -lt ${START_HOUR} || ${HOUR} -ge ${END_HOUR} ]]; then
log_warn "[WATCHDOG] 凌晨时段(${HOUR}时), message_log 仅 ${MSG_COUNT} 条, 仅记录"
exit 0
fi
# --- 工作时间 + 入库过少 -> 触发重启 ---
log_warn "[WATCHDOG] 最近 ${MSG_WINDOW_MIN} 分钟 message_log 仅 ${MSG_COUNT} 条 < ${MSG_THRESHOLD}, 触发 Resin 重启!"
${OA_SSH} "${OA_RESTART_CMD}" 2>&1 | tee -a "${LOG_FILE}"
sleep 20
NEW_PID=$(${OA_SSH} "ps -eo pid,cmd | grep 'java.*Resin.*resin.xml' | grep -v grep | awk '{print \$1}'" 2>/dev/null | head -1)
HTTP_CODE=$(${OA_SSH} "curl -s -o /dev/null -w '%{http_code}' http://127.0.0.1:80/wui/main/Login.jsp" 2>/dev/null)
BODY="Resin watchdog 自动重启\n\n"
BODY="${BODY}message_log 最近 ${MSG_WINDOW_MIN} 分钟入库: ${MSG_COUNT} 条 (阈值 ${MSG_THRESHOLD})\n"
BODY="${BODY}新 Resin PID: ${NEW_PID}\n"
BODY="${BODY}HTTP 80: ${HTTP_CODE}\n"
dingtalk_alert "P0" "Resin watchdog 自动重启(message_log 入库 ${MSG_COUNT} 条 < ${MSG_THRESHOLD})" "${BODY}"
if [[ "${NEW_PID}" =~ ^[0-9]+$ ]]; then
echo "${NEW_PID} 0 $(date +%s)" > "${PID_FILE}"
fi
log_info "[WATCHDOG] Resin 重启完成, 新 PID=${NEW_PID}, HTTP=${HTTP_CODE}"
+96
View File
@@ -0,0 +1,96 @@
#!/usr/bin/env bash
# ============================================================================
# OA Monitor · Resin Watchdog (v2 - python paramiko 版)
# 监控 Util_Message 线程活跃度(publishMessageList),超过 30 分钟无活动则自动重启 Resin
# 部署在 DSH 上,通过 python paramiko 到 OA 服务器(39.99.236.138)执行
# crontab: */5 * * * * /opt/oa-monitor/bin/resin-watchdog.sh
# ============================================================================
set -u
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
source "${SCRIPT_DIR}/../lib/common.sh"
OA_HOST="39.99.236.138"
OA_USER="root"
OA_PASS="KTOFKjVkVOGlNgvs8yUP1+PM1VM"
OA_LOG="/usr/weaver/Resin/log/stdout.log"
OA_RESTART_CMD="bash /usr/weaver/Resin/bin/stopresin.sh 2>&1; sleep 5; bash /usr/weaver/Resin/bin/startresin.sh 2>&1"
MAX_IDLE_MIN=30
LOG_FILE="${MONITOR_LOGS}/resin-watchdog.log"
OA_SSH_PY="${MONITOR_ROOT}/bin/_oa_ssh.py"
NOW=$(date +%s)
# Step 1: 确保 python paramiko 可用
if ! python3 -c 'import paramiko' 2>/dev/null; then
MSG="[WATCHDOG] 🔴 python3 + paramiko 不可用!请执行: pip3 install paramiko"
log_error "${MSG}"
dingtalk_alert "P0" "Resin Watchdog 无法运行" "${MSG}"
exit 1
fi
# Step 2: 写 _oa_ssh.py(paramiko + errors=replace 处理 GBK 乱码)
cat > "${OA_SSH_PY}" << 'PYEOF'
import paramiko, sys
host, user, password, cmd = sys.argv[1], sys.argv[2], sys.argv[3], sys.argv[4]
ssh = paramiko.SSHClient()
ssh.set_missing_host_key_policy(paramiko.AutoAddPolicy())
try:
ssh.connect(host, username=user, password=password, timeout=10)
stdin, stdout, stderr = ssh.exec_command(cmd)
sys.stdout.write(stdout.read().decode("utf-8", errors="replace"))
ssh.close()
except Exception as e:
print(f"SSH_ERROR: {e}", file=sys.stderr)
sys.exit(1)
PYEOF
chmod +x "${OA_SSH_PY}"
OA_SSH="python3 ${OA_SSH_PY} ${OA_HOST} ${OA_USER} ${OA_PASS}"
# Step 3: 提取最后一个 publishMessageList 时间(tail -5000 ≈ 10 分钟覆盖)
LAST_RAW=$(${OA_SSH} "tail -5000 ${OA_LOG} 2>/dev/null | grep publishMessageList | tail -1" 2>/dev/null)
LAST_TS=$(echo "${LAST_RAW}" | grep -oE '[0-9]{4}\.[0-9]{2}\.[0-9]{2} [0-9]{2}:[0-9]{2}:[0-9]{2}' | head -1)
if [[ -z "${LAST_TS}" ]]; then
log_warn "[WATCHDOG] ⚠️ stdout.log 里没找到 publishMessageList 记录,跳过本次"
exit 0
fi
# Step 4: 解析时间戳 → epoch
LAST_EPOCH=$(date -d "$(echo ${LAST_TS} | sed "s|\.|-|g")" +%s 2>/dev/null)
if [[ -z "${LAST_EPOCH}" ]]; then
log_warn "[WATCHDOG] ⚠️ 无法解析日志时间: ${LAST_TS},跳过本次"
exit 0
fi
IDLE_MIN=$(( (NOW - LAST_EPOCH) / 60 ))
log_info "[WATCHDOG] 最后 publishMessageList: ${LAST_TS},已空闲 ${IDLE_MIN} 分钟(阈值 ${MAX_IDLE_MIN})"
# Step 5: 判断是否需要重启
if [[ ${IDLE_MIN} -lt ${MAX_IDLE_MIN} ]]; then
log_info "[WATCHDOG] ✅ publishMessageList 活跃(最近 ${IDLE_MIN} 分钟前),Resin 健康"
exit 0
fi
# --- 触发重启 ---
log_warn "[WATCHDOG] 🔴 空闲 ${IDLE_MIN} 分钟 ≥ ${MAX_IDLE_MIN} 分钟!触发 Resin 重启!"
${OA_SSH} "${OA_RESTART_CMD}" 2>&1 | tee -a "${LOG_FILE}"
sleep 20
NEW_PID=$(${OA_SSH} "ps -eo pid,cmd | grep 'java.*Resin.*resin.xml' | grep -v grep | awk '{print \$1}'" 2>/dev/null | head -1)
HTTP_CODE=$(${OA_SSH} "curl -s -o /dev/null -w '%{http_code}' http://127.0.0.1:80/wui/main/Login.jsp" 2>/dev/null)
BODY="**🔴 Resin 被 watchdog 自动重启**\n\n"
BODY="${BODY}| 指标 | 值 |\n|---|---|\n"
BODY="${BODY}| publishMessageList 最后活动 | \`${LAST_TS}\` |\n"
BODY="${BODY}| 空闲时长 | **${IDLE_MIN} 分钟** |\n"
BODY="${BODY}| 新 Resin PID | \`${NEW_PID}\` |\n"
BODY="${BODY}| HTTP 80 端口 | \`${HTTP_CODE}\` |\n\n"
BODY="${BODY}> ⚠️ 推送线程静默死掉,watchdog 自动恢复。"
BODY="${BODY}> 建议:泛微需要检查 Util_Message 线程为什么会静默挂掉。"
dingtalk_alert "P0" "Resin watchdog 自动重启(推送线程空闲 ${IDLE_MIN} 分钟)" "${BODY}"
log_info "[WATCHDOG] ✅ Resin 重启完成,新 PID=${NEW_PID}, HTTP=${HTTP_CODE}"
+97
View File
@@ -0,0 +1,97 @@
#!/usr/bin/env bash
# ============================================================================
# OA Monitor · Resin Watchdog (v3 - message_log 入库增量判定)
# 健康依据:ecology_message_log 最近 15 分钟入库数
# - 工作时间(08:00~22:00) 入库 < 10 条 → 判定推送线程死 → 重启 Resin
# - 凌晨时段 入库少也不重启(避免误重启)
# 部署在 DSH,python paramiko 到 OA 执行重启
# crontab: */5 * * * * /opt/oa-monitor/bin/resin-watchdog.sh
# ============================================================================
set -u
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
source "${SCRIPT_DIR}/../lib/common.sh"
OA_HOST="39.99.236.138"
OA_USER="root"
OA_PASS="KTOFKjVkVOGlNgvs8yUP1+PM1VM"
OA_RESTART_CMD="bash /usr/weaver/Resin/bin/stopresin.sh 2>&1; sleep 5; bash /usr/weaver/Resin/bin/startresin.sh 2>&1"
MSG_WINDOW_MIN=15 # 统计最近 15 分钟
MSG_THRESHOLD=10 # 入库 < 10 条视为异常
START_HOUR=8 # 工作时间起始
END_HOUR=22 # 工作时间结束
LOG_FILE="${MONITOR_LOGS}/resin-watchdog.log"
OA_SSH_PY="${MONITOR_ROOT}/bin/_oa_ssh.py"
# Step 1: 确保 python paramiko 可用
if ! python3 -c 'import paramiko' 2>/dev/null; then
MSG="[WATCHDOG] 🔴 python3 + paramiko 不可用!请执行: pip3 install paramiko"
log_error "${MSG}"
dingtalk_alert "P0" "Resin Watchdog 无法运行" "${MSG}"
exit 1
fi
# Step 2: 写 _oa_ssh.py(仅重启时用)
cat > "${OA_SSH_PY}" << 'PYEOF'
import paramiko, sys
host, user, password, cmd = sys.argv[1], sys.argv[2], sys.argv[3], sys.argv[4]
ssh = paramiko.SSHClient()
ssh.set_missing_host_key_policy(paramiko.AutoAddPolicy())
try:
ssh.connect(host, username=user, password=password, timeout=10)
stdin, stdout, stderr = ssh.exec_command(cmd)
sys.stdout.write(stdout.read().decode("utf-8", errors="replace"))
ssh.close()
except Exception as e:
print(f"SSH_ERROR: {e}", file=sys.stderr)
sys.exit(1)
PYEOF
chmod +x "${OA_SSH_PY}"
OA_SSH="python3 ${OA_SSH_PY} ${OA_HOST} ${OA_USER} ${OA_PASS}"
# Step 3: 查 message_log 最近 MSG_WINDOW_MIN 分钟入库数(DSH 直连 MySQL,走 pushlog_time 索引)
MSG_COUNT=$(MYSQL_PWD="${DB_PASS}" mysql -h39.100.72.102 -P3306 -uecology "${DB_NAME}" \
-sN -e "SELECT COUNT(*) FROM ecology_message_log WHERE CREATEDATE=CURDATE() AND CREATETIME >= DATE_FORMAT(NOW() - INTERVAL ${MSG_WINDOW_MIN} MINUTE, '%H:%i:%s');" 2>/dev/null)
if [[ ! "${MSG_COUNT}" =~ ^[0-9]+$ ]]; then
log_warn "[WATCHDOG] ⚠️ 查询 message_log 失败(返回: ${MSG_COUNT}),跳过本次"
exit 0
fi
log_info "[WATCHDOG] 最近 ${MSG_WINDOW_MIN} 分钟 message_log 入库 ${MSG_COUNT} 条"
# Step 4: 判断
HOUR=$(date +%H)
if [[ ${MSG_COUNT} -ge ${MSG_THRESHOLD} ]]; then
log_info "[WATCHDOG] ✅ message_log 入库正常(${MSG_COUNT} 条 ≥ ${MSG_THRESHOLD} 阈值),推送健康"
exit 0
fi
# 入库 < 阈值
if [[ ${HOUR} -lt ${START_HOUR} || ${HOUR} -ge ${END_HOUR} ]]; then
log_warn "[WATCHDOG] ⚠️ 凌晨时段(${HOUR}时),message_log 仅 ${MSG_COUNT} 条 < ${MSG_THRESHOLD},仅记录不重启"
exit 0
fi
# --- 工作时间 + 入库过少 → 触发重启 ---
log_warn "[WATCHDOG] 🔴 最近 ${MSG_WINDOW_MIN} 分钟 message_log 仅 ${MSG_COUNT} 条 < ${MSG_THRESHOLD},触发 Resin 重启!"
${OA_SSH} "${OA_RESTART_CMD}" 2>&1 | tee -a "${LOG_FILE}"
sleep 20
NEW_PID=$(${OA_SSH} "ps -eo pid,cmd | grep 'java.*Resin.*resin.xml' | grep -v grep | awk '{print \$1}'" 2>/dev/null | head -1)
HTTP_CODE=$(${OA_SSH} "curl -s -o /dev/null -w '%{http_code}' http://127.0.0.1:80/wui/main/Login.jsp" 2>/dev/null)
BODY="**🔴 Resin 被 watchdog 自动重启**\n\n"
BODY="${BODY}| 指标 | 值 |\n|---|---|\n"
BODY="${BODY}| message_log 最近 ${MSG_WINDOW_MIN} 分钟入库 | **${MSG_COUNT} 条**(阈值 ${MSG_THRESHOLD})|\n"
BODY="${BODY}| 新 Resin PID | \`${NEW_PID}\` |\n"
BODY="${BODY}| HTTP 80 端口 | \`${HTTP_CODE}\` |\n\n"
BODY="${BODY}> ⚠️ 推送线程静默死掉(message_log 不再入库),watchdog 自动恢复。"
BODY="${BODY}> 注:publishMessageList 日志可能仍在刷(假活跃),以 message_log 入库为准。"
dingtalk_alert "P0" "Resin watchdog 自动重启(message_log 入库 ${MSG_COUNT} 条 < ${MSG_THRESHOLD})" "${BODY}"
log_info "[WATCHDOG] ✅ Resin 重启完成,新 PID=${NEW_PID}, HTTP=${HTTP_CODE}"
+66
View File
@@ -0,0 +1,66 @@
#!/usr/bin/env bash
# ============================================================================
# OA Monitor · 长查询检测 dry-run KILL(每 10 分钟)
# 默认 DRY_RUN=1(只报告,不真 KILL)。--execute 才真的 KILL
# ============================================================================
set -u
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
source "${SCRIPT_DIR}/../lib/common.sh"
DRY_RUN=1
for arg in "$@"; do
[[ "${arg}" == "--execute" ]] && DRY_RUN=0
[[ "${arg}" == "--dry-run" ]] && DRY_RUN=1
done
NOW=$(date +%s)
MYSQL_RUN() {
MYSQL_PWD="${DB_PASS}" timeout 30 mysql -h39.100.72.102 -P3306 -uecology ecology "$@" 2>/dev/null
}
log_info "[LONG-CHECK] 开始(DRY_RUN=${DRY_RUN},阈值 ${ALERT_LONG_QUERY_SEC}s)"
RESULT=$(MYSQL_RUN "SHOW FULL PROCESSLIST;" | awk -F'\t' -v t="${ALERT_LONG_QUERY_SEC}" '
NR==1 { next }
NF < 10 { next }
{
pid=$1; user=$2; state=$6; time=$7; info=$9
if (time+0 >= t+0 && time+0 > 0) {
gsub(/\t/, " ", info)
printf "%d\t%d\t%s\t%s\t%s\n", pid, time, user, state, substr(info,1,300)
}
}
')
if [[ -z "${RESULT}" ]]; then
log_info "[LONG-CHECK] ✅ 无长查询"
exit 0
fi
COUNT=$(echo "${RESULT}" | wc -l | tr -d ' ')
log_warn "[LONG-CHECK] ⚠️ ${COUNT} 条长查询"
DETAIL=""
while IFS=$'\t' read -r pid time user state info; do
[[ -z "${pid}" ]] && continue
DETAIL="${DETAIL}\n- PID \`${pid}\` **${time}s** \`${user}\` ${state}"
DETAIL="${DETAIL}\n \`\`\`\n ${info}\n \`\`\`"
done <<< "${RESULT}"
MODE_LABEL=$([[ ${DRY_RUN} -eq 1 ]] && echo "🔒 DRY-RUN" || echo "🚨 执行 KILL")
BODY="**${COUNT} 条长查询**(${ALERT_LONG_QUERY_SEC}s 阈值)· ${MODE_LABEL}\n\n${DETAIL}"
dingtalk_alert "P0" "长查询阻塞需处理" "${BODY}"
sqlite_exec "INSERT INTO alerts (ts, level, category, title, detail) VALUES (${NOW}, 'P0', 'db', '长查询', '${COUNT} 条, dry_run=${DRY_RUN}');"
if [[ ${DRY_RUN} -eq 0 ]]; then
log_warn "[LONG-CHECK] 执行 KILL!"
while IFS=$'\t' read -r pid time user state info; do
[[ -z "${pid}" ]] && continue
MYSQL_RUN "KILL ${pid};"
log_warn "[LONG-CHECK] KILL ${pid} (${time}s, ${user})"
done <<< "${RESULT}"
else
log_info "[LONG-CHECK] DRY-RUN,不执行 KILL"
fi