高並發linux核心網路參數調優

調整linux(ubuntu8.0.4)核心參數以便滿足高並發訪問，解決大量time_wait和syn攻擊問題： 
1 sudo vi /etc/sysctl.conf 
在最下面編輯添加： 
net.ipv4.tcp_fin_timeout = 30 
net.ipv4.tcp_keepalive_time = 1200 
net.ipv4.route.gc_timeout = 100 
net.ipv4.ip_local_port_range = 1024 65000 
net.ipv4.tcp_tw_reuse = 1 
net.ipv4.tcp_tw_recycle = 1 
net.ipv4.tcp_syn_retries = 1 
net.ipv4.tcp_synack_retries = 1 
net.ipv4.tcp_max_syn_backlog = 262144 
net.core.netdev_max_backlog = 262144 
net.core.somaxconn = 262144 
net.ipv4.tcp_mem = 94500000 915000000 927000000 
儲存退出 
2 sudo /sbin/sysctl -p 
立刻生效(實踐中有遇到不能立刻生效的，需要重啟機子才會生效) 

使用如下命令監控當前各狀態串連： 
1 netstat -n | awk '/^tcp/ {++S[$NF]} END {for(a in S) print a, S[a]}' 
2 cat /proc/net/sockstat 
3 查看是否有異常日誌：dmesg 
4 查看當前IP個數和每IP串連數： 
netstat -an | grep 80 | awk '{print $5}' | awk 'BEGIN {FS=":"} NF==2 {print $1} NF==5 {print $4}' | sort | uniq -c | sort -n   

預設值恢複： 
net.ipv4.tcp_fin_timeout = 60 
net.ipv4.tcp_keepalive_time = 7200 
net.ipv4.route.gc_timeout = 300 
net.ipv4.ip_local_port_range = 32768 61000 
net.ipv4.tcp_tw_reuse = 0 
net.ipv4.tcp_tw_recycle = 0 
net.ipv4.tcp_syn_retries = 5 
net.ipv4.tcp_synack_retries = 5 
net.ipv4.tcp_max_syn_backlog = 1024 
net.core.netdev_max_backlog = 1000 
net.core.somaxconn = 128 
net.ipv4.tcp_mem = 853440 1137920 1706880 

以上配置在實際生產環境中單機1.7w串連數情況下nginx工作正常。 

伺服器最佳化推薦參考文章：http://blog.sina.com.cn/s/blog_466c66400100bi2n.html~type=v5_one&label=rela_prevarticle