Shulou(Shulou.com)06/02 Report--

This article mainly introduces "how to use Lvs+Nginx cluster to build high concurrency architecture". In daily operation, I believe many people have doubts about how to use Lvs+Nginx cluster to build high concurrency architecture. The editor consulted all kinds of materials and sorted out simple and easy-to-use operation methods. I hope it will be helpful to answer the doubts about "how to use Lvs+Nginx cluster to build high concurrency architecture". Next, please follow the editor to study!

High concurrency sites need to consider not only the stability of the website's back-end service, but also whether the service can access and bear huge traffic, as shown below:

1: traffic access, you can use Lvs+Nginx cluster, which can access millions of QPS.

2: Nginx cluster is implemented through Lvs, and back-end service cluster is implemented by Nginx+Tomcat, which completes the high concurrency processing from access layer traffic processing to back-end service cluster.

1. Lvs introduction

LVS (Linux Virtual Server), which is the Linux virtual server. It is used for multi-server load balancing, works in the four layers of the network, can achieve high-performance, high-availability server cluster technology, it is stable and reliable, even if a server in the cluster can not work properly, it does not affect the overall effect. It is based on TCP/IP for routing and forwarding, with high stability and efficiency.

A LVS cluster often contains the following roles:

1:DS:Director Server . Virtual service, responsible for scheduling

2:RS:Real Server . The real working server at the back end.

3:VIP: the IP address that directly faces the user's request from the outside and is the target of the user's request

IP of 4:DIP:Director Server IP,DS

5:RIP:Real Server IP, the IP address of the backend server

6:CIP:Client IP, the IP address of the access client

2. Lvs load balancing mode

Lvs provides three load balancing modes, each of which is suitable for different scenarios. Let's explain these three load balancing modes.

2.1 NAT

After the user's request is sent to the dispatcher, the requested data packet is forwarded to the back-end RS through the preset iptables rules. RS needs to set the gateway as the private network IP of the dispatcher. The packets requested by the user and the packets returned to the user all pass through the dispenser, so the dispenser is called a bottleneck. In NAT mode, only the dispatcher needs to have public network IP, so it saves public network IP resources.

2.2 TUN

This mode requires a common IP configured on the dispenser and all the RS, which we call VIP. The target IP of the client request is VIP. After receiving the request packet, the dispatcher will process the packet and change the target IP to the IP of RS, so that the packet will be sent to the RS. After RS receives the packet, it restores the original packet so that the destination IP is VIP, and because this VIP is configured on all RS, he will think it is itself.

2.3 DR mode

Similar to IP Tunnel, except that it modifies the MAC address of the packet to the MAC address of RS. The real server returns the response directly to the customer.

This method does not have the cost of IP tunnel, and there is no requirement that the real service period in the cluster must support IP tunnel protocol, but it requires that the scheduler and the real server have a network card connected to the same physical network segment.

3. Lvs DR mode configuration

Based on the above analysis, we can draw a conclusion that DR mode has high performance efficiency and high security, so general companies recommend using DR mode. We also configure DR mode to implement Lvs+Nginx clusters here.

We prepared three machines: first, make sure that all three machines have Nginx installed.

1RV 192.168.183.133 (DS) 192.168.183.150 external services 2RV 192.168.183.134 (RS) 192.168.183.150 Real Service processing Business process 3RV 192.168.183.135 (RS) 192.168.183.150 Real Service processing Business process

VIP:192.168.183.150

3.1 Vip configuration

Close the network configuration manager (do it on every machine)

Systemctl stop NetworkManagersystemctl disable NetworkManager

Configure virtual IP (configured in VIP 192.168.183.133)

Create the file ifcfg-ens33:1 in / etc/sysconfig/network-scripts, as follows:

BOOTPROTO=staticDEVICE=ens33:1ONBOOT=yesIPADDR=192.168.183.150NETMASK=255.255.255.0

Restart the network service:

Service network restart

We can see that a virtual IP 150 has been added to the original network card.

At the same time, we need to build the virtual machine IP for 192.168.183.134 and 192.168.183.135, but it is only used to return data and cannot be accessed by the user, so we need to operate ifcfg-lo.

IPADDR=127.0.0.1, where 127.0.0.1 belongs to the local loopback address, does not belong to any classful address class. It represents the local virtual interface of the device, so it is considered by default to be an interface that will never go down.

NETMASK=255.255.255.255

192.168.183.134:

Copy the ifcfg-lo into an ifcfg-lo:1 and modify the ifcfg-lo:1 configuration as follows:

Refresh lo:

Ifup lo

If you look at IP, you can see that there is too much 150ip under lo.

192.168.100.133 know the same operation as above.

3.2 installation of LVS cluster management tools

Ipvsadm is used to manage the lvs cluster and needs to be installed manually. DS can be installed.

Installation commands:

Yum install ipvsadm

Version View:

Ipvsadm-Ln

The effect is as follows:

3.3 address Resolution Protocol

Operate in 192.168.183.134 and 192.168.183.135.

Both arp_ignore and arp_announce parameters are related to the ARP protocol and are mainly used to control the actions of the system when it returns an arp response and sends an arp request. These two parameters are very important, especially in the DR scenario of LVS, their configuration directly affects whether DR forwarding is normal.

The function of the arp-ignore:arp_ignore parameter is to control whether the system returns an arp response when it receives an external arp request.

Configuration file: / etc/sysctl.conf, copy the following files in:

Net.ipv4.conf.all.arp_ignore = 1net.ipv4.conf.default.arp_ignore = 1net.ipv4.conf.lo.arp_ignore = 1net.ipv4.conf.all.arp_announce = 2net.ipv4.conf.default.arp_announce = 2net.ipv4.conf.lo.arp_announce = 2

Refresh the configuration:

Sysctl-p

Add routing: if the route is not recognized at this time, you need to install the related tool yum install net-tools.

Route add-host 192.168.183.150 dev lo:1

An host address has been added to receive the data message, which will be handed over to the lo:1 for processing. (to prevent shutdown failure, you need to add the above command to / etc/rc.local)

After adding host, you can take a look at: route-n, you can see the effect obviously.

We will also configure the above configuration in 192.168.183.135.

3.4 Cluster configuration

The ipvsadm command explains:

Ipvsadm-A: used to create a cluster

Ipvsadm-E: used to modify the cluster

Ipvsadm-D: used to delete a cluster

Ipvsadm-C: used to clear cluster data

Ipvsadm-R: used to reset cluster configuration rules

Ipvsadm-S: used to save modified cluster rules

Ipvsadm-a: used to add a rs node

Ipvsadm-e: used to modify a rs node

Ipvsadm-d: used to delete a rs node

Add cluster TCP service address: (external requests are handled by the VIP specified by the configuration)

Ipvsadm-A-t 192.168.183.150 ipvsadm 80-s rr

Parameter description:

-A: add cluster configuration

-t:TCP request address (VIP)

-s: load balancing algorithm

Load balancing algorithm:

The algorithm describes the rr polling algorithm, which allocates the requests to different rs nodes in turn, that is, the equal allocation among the RS nodes. This algorithm is simple, but it is only suitable for RS weighted rotation scheduling when the processing performance of wrr nodes is similar. It will assign tasks according to the weights of different RS. The RS with higher weight will get the task first, and the number of connections assigned will be more than the RS with lower weight. RS with the same weight gets the same number of connections. Wlc weighted minimum number of connections scheduling, assuming that the full-time RS is Wi, and the current number of tcp connections is Ti. To Ti/Wi the smallest RS as the next assigned RSDh destination address hash scheduling (destination hashing) look up a static hash table with the destination address as the keyword to obtain the required RSSH source address hash scheduling (source hashing) use the source address as the keyword to find a static hash table to obtain the required RSLc minimum connection scheduling (least-connection), the IPVS table stores all active connections. LB compares address-based minimum connection scheduling (locality-based least-connection) for sending connection requests to the RS.Lblc with the least number of connections: assigning requests from the same destination address to the same RS when the server is not fully loaded. Otherwise, the request is assigned to the RS with the lowest number of connections as the first consideration for the next allocation.

Configure rs (2) nodes in DS:

Ipvsadm-a-t 192.168.183.150 purl 80-r 192.168.183.134 80-gipvsadm-a-t 192.168.183.150 purl 80-r 192.168.183.135 purl 80-g

Parameter description:

-a: add a node to the cluster

-t: specify the VIP address

-r: specify the real server address

-g: indicates that the mode of LVS is dr mode

After adding nodes, we can see that there are 2 more nodes through ipvsadm-Ln.

At this point, the client request data and TCP communication data in the cluster list will be persisted. To better see the effect, we can set the storage time to 2 seconds, as shown in the following command:

Ipvsadm-set 2 2 2

At this point, we request the http://192.168.183.150/ test

You can find that the request will be polled and switched between the two Nginx.

At this point, the study on "how to use Lvs+Nginx clusters to build a high concurrency architecture" is over. I hope to be able to solve your doubts. The collocation of theory and practice can better help you learn, go and try it! If you want to continue to learn more related knowledge, please continue to follow the website, the editor will continue to work hard to bring you more practical articles!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.