升级后监控脚本

#!/bin/bash
# post_upgrade_monitor.sh
# from:www.itpux.com.qq113257174.wx:itpux-com
# web: http://www.fgedu.net.cn

# 定义变量
LOG_FILE="/opengauss/logs/post_upgrade_monitor_$(date +'%Y%m%d').log"
METRICS_FILE="/opengauss/metrics/metrics_$(date +'%Y%m%d_%H%M%S').json"

# 日志函数
log() {
    echo "[$(date +'%Y-%m-%d %H:%M:%S')] $1" >> $LOG_FILE
    echo "[$(date +'%Y-%m-%d %H:%M:%S')] $1"
}

# 监控数据库状态
monitor_database() {
    log "监控数据库状态..."
    
    # 检查数据库是否运行
    gs_ctl status -D /opengauss/data > /tmp/db_status.txt
    if grep -q "server is running" /tmp/db_status.txt; then
        log "数据库运行正常"
    else
        log "警告：数据库未运行"
    fi
    
    # 检查连接数
    connections=$(gsql -U fgedu -d postgres -t -c "SELECT count(*) FROM pg_stat_activity; 
")
    log "当前连接数：$connections"
    
    # 检查系统负载
    load=$(uptime | awk '{print $10 $11 $12}')
    log "系统负载：$load"
    
    # 检查磁盘空间
    disk_usage=$(df -h | grep /opengauss | awk '{print $5}')
    log "磁盘使用率：$disk_usage"
    
    # 收集性能指标
    gsql -U fgedu -d postgres -t -c "SELECT datname, numbackends, xact_commit, xact_rollback, blks_read, blks_hit FROM pg_stat_database; 
" > /tmp/performance_metrics.txt
    
    # 生成JSON格式的指标
    cat > $METRICS_FILE << EOF
{
    "timestamp": "$(date +'%Y-%m-%d %H:%M:%S')",
    "database_status": "$(grep -q "server is running" /tmp/db_status.txt && echo "running" || echo "stopped")",
    "connections": "$connections",
    "system_load": "$load",
    "disk_usage": "$disk_usage",
    "performance_metrics": "$(cat /tmp/performance_metrics.txt | tr '\n' ' ' | sed 's/  */ /g')"
}
EOF
    
    log "数据库状态监控完成"
}

# 监控应用服务
monitor_application() {
    log "监控应用服务..."
    
    # 检查应用服务是否运行
    systemctl status application > /tmp/app_status.txt
    if grep -q "active (running)" /tmp/app_status.txt; then
        log "应用服务运行正常"
    else
        log "警告：应用服务未运行"
    fi
    
    # 检查应用响应时间
    response_time=$(curl -o /dev/null -s -w "%{time_total}" http://localhost:8080)
    log "应用响应时间：$response_time 秒"
    
    # 检查应用错误率
    error_rate=$(grep -c "ERROR" /var/log/application.log | tail -n 100 | wc -l | awk '{print $1/100}')
    log "应用错误率：$error_rate"
    
    log "应用服务监控完成"
}

# 发送监控报告
send_report() {
    log "发送监控报告..."
    
    # 发送邮件通知
    echo "升级后监控报告" | mail -s "openGauss升级后监控报告" -a $METRICS_FILE admin@fgedu.net.cn
    
    log "监控报告发送完成"
}

# 主流程
log "=== 开始升级后监控 ==="

monitor_database
monitor_application
send_report

log "=== 升级后监控完成 ==="