Load balancer can be used with LVS scheme, but to prevent single point of failure, you can choose lvs+keepalived combination to ensure high availability
Focus : Each node synchronizes time
Ntpdate time.windows.com
1. Introduction to the Environment
Operating system: Centos6.6
Dr Tops: Ipvs management tools: ipvsadm,keepalived
Realserver tops: Nginx or Apache
650) this.width=650; "src=" Http://s3.51cto.com/wyfs02/M02/71/5B/wKiom1XMHTuy6mSoAAB-rc_cONg598.jpg "title=" Lvs.png "alt=" Wkiom1xmhtuy6msoaab-rc_cong598.jpg "/>
first, the theoretical chapter
composition of the 1.lvs cluster
The LVS server system consists of three parts
1) Load Balancer layer:
Located in the forefront of the entire system, to avoid a single point of failure is generally composed of 2 or more than 2 load scheduler
2) Server Group layer
is a group of truly service machines, real server can be: Web, Ftp, mail and so on. 4
3) shared storage tier
Provides a storage area for real servers that share storage space and content consistency
Directory is the core of the entire LVS and can only be installed on Linux and FreeBSD,
Real server can be all system platforms
2.lvs of three modes:
*NAT:
Principle: Dr received the request packet, according to the scheduling algorithm to find the corresponding RS, the destination IP of the packet is changed to RS IP, and the request is distributed to the RS,RS received the packet and processing completed, the data will be sent to dr,dr the original address of the packet to its own IP, the destination IP is changed to the client IP
Advantage: Save IP Address, because only Dr needs a public IP address
Disadvantage: When all the request packages and reply packets pass through the DR (load balancer) and the server nodes are too large, a lot of packets are processed by the load balancer slowly.
* Direct Connect route-DR:
Principle: Dr received the request packet, according to the scheduling algorithm to find the corresponding RS, the purpose of the packet Mac changed to Rs mac (because the IP is consistent), and the request was distributed to the Rs,rs received the packet and processing completed, due to IP consistent, the data is returned directly to the client
Pros: Dr is just a distribution request, and the response packet is returned directly to the customer by RS, reducing the load balancer's massive data flow, and the load balancer is no longer a system bottleneck
Disadvantage: Each machine needs a public IP,
* Tunnel-tun:
principle: (one IP packet is encapsulated in another IP packet, so it can cross the network segment.) After the DR receives the request packet, according to the scheduling algorithm to find the corresponding Rs,rs received, the message is unpacked to obtain the destination address of the VIP message, if the VIP address is configured locally, the request will be processed, processing the data directly back to the client
Advantage: If you run on the public network, you can distribute to different regions and cross the city.
3.lvs of four scheduling algorithms:
Polling (RR): In turn distributed to back-end RS in order of request
Weighted polling (WRR): Higher weights get more tasks
Minimum number of connections (LC): Dynamically creates a request on a RS with a smaller number of connections
Weighted minimum number of connections (WLC): The scheduler automatically asks for the real load situation of RS and dynamically adjusts the power
4.ipvsadm parameters
1) Add Virtual Server syntax:
Ipvsadm-a t|u|f [Vip_addr:port] [-s specify algorithm]
-A: Add
-T:TCP protocol
-U:UDP protocol
-F: Firewall tag
-D: Delete
-E: Modify
-L: View
2) Add Real server syntax
ipvsadm-a t|u|f [Vip_addr:port] [-R Rsip_addr] [-g|i|m] [-w weight]
-A: Add
-T:TCP protocol
-U:UDP protocol
-F: Firewall tag
-r: Specify Real server IP
-G:DR mode
-i:tun mode
-m:nat mode
-W: Specify Weights
-D: Delete
-E: Modify
-L: View
3) General:
IPVSADM-LN Viewing rules
Service ipvsadm Save rule
Ipvsadm-r < Path/to/somefile
4) Example:
Ipvsadm-a-T 192.168.1.10:80-s wrr ipvsadm-a-T 192.168.1.10:80-r 192.168.1.107-g-W 4ipvsadm-a-T 192.168.1.10:80- R 192.168.1.108-t-W 3
second, the actual combat chapter
1. Building the LVS
for Dr mode
1) directory: Installation Ipvsadm yum -y install ipvsadm Configuration vip ifconfig eth0:0 192.168.1.10 broadcast 192.168.1.10 netmask 255.255.255.255 Add Routes route add -host 192.168.1.10 dev eth0 adding a virtual server ipvsadm -A -t 192.168.1.10:80 -s wrr Add back-end rs1 Ipvsadm -a -t 192.168.1.10:80 -r 192.168.1.105 -g -w 2 Add back-end rs2 Ipvsadm -a -t 192.168.1.10:80 -r 192.168.1.106 -g -w 2 Preservation service ipvsadm Save<pre name= "code" class= "HTML" > View ipvsadm -ln 1) RS node: echo 2 > /proc/sys/net/ipv4/conf/all/arp_announce echo 2 > /proc/sys/net/ipv4/conf/eth0/ Arp_announce echo 1 > /proc/sys/net/ipv4/conf/all/arp_ignore echo 1 >/proc/sys/net/ipv4/conf/eth0/arp_ignore above configure tuning Linux kernel arp response parameters, prevent update VIP MAC address, avoid conflict Configuration vip ifconfig lo:0 192.168.1.10 braodcast 192.168.1.10 netmask 255.255.255.255 Add Route route add -host 192.168.1.10 dev lo:0 Start Services service httpd start
Build LVS Load Balancer Cluster