JIYIK CN >

At this stage, load balancing is a widely used technology. Nginx, as a load balancing server for http, is being used more and more widely.

There are three ways to set up Nginx load balancing:

Round-robin - This method distributes access requests to application servers in a round-robin manner.

least-connect - The next request is dispatched to the application server with the fewest active connections.

ip-hash ——Use the hash method to determine which application server to distribute the next request to.

Let’s look at these three methods separately.

Round-robin default mode

This is the simplest distribution method and the easiest to configure. Let's look at the following example.

Example 1

http {
    upstream onmpw {
        server 192.168.144.128;
        server 192.168.144.132;
        server 192.168.144.131;
    }
    server {
        listen 80;
        location / {
            proxy_pass http://onmpw;
        }
    }
}

In Example 1, there are three application instances. Since we did not specify the load mode in the upstream, the default is round-robin mode. That is, for the first request, nginx load balancing will distribute the request to application server 128; the second request will be distributed to application server 132; the third request will be distributed to application server 131; and the fourth request will be distributed to application server 128 again. In this way, the request will be distributed to each application server in a round-robin manner.

Least connect load balancing setting method

Another load balancing method is least-connect. This method seems to be fairer to application servers. This method is advantageous when some requests take a long time to complete. In this method, the load service will try not to respond to too many requests to an overloaded application server. It will distribute new requests to application servers with less load.

The setting is as follows (just slightly change the example 1)

Example 2

upstream onmpw {
least_conn;
    server 192.168.144.128;
    server 192.168.144.132;
    server 192.168.144.131;
}

Session persistence - load balancing

It should be noted here that whether it is round-robin or least-connect, when processing requests, both methods may distribute subsequent requests to different application servers. They cannot guarantee that the requests of the same client will always be directed to the same application server.

If we need to bind a client to a specific application server, that is, to make the client session persistent by distributing specific client requests to a specific application server, then the ip-hash load method is needed.

In this way, the client's IP address is used as the hash key to determine which application server to distribute the client's request to. This method ensures that requests from the same client will always be distributed to a specific application server, unless this specific server stops working.

To configure the ip-hash load mode, just add ip_hash to the upstream in Example 1.

Example 3

upstream onmpw {
ip_hash;
    server 192.168.144.128;
    server 192.168.144.132;
    server 192.168.144.131;
}

Temporarily remove an application server

If an application server needs to be temporarily removed and requests are not allowed to access it, we can use down after it to mark this server, which means that nginx will not distribute requests to this application server. When the client previously responded to by this server makes a request again, nginx will automatically distribute it to other application servers.

Example 4

upstream onmpw {
    server 192.168.144.128;
    server 192.168.144.132;
    server 192.168.144.131 down;
}  

Application server weight

By default, Nginx distributes requests to the application server group in a round-robin manner. We can set the weight of each application server through the weight parameter. By default, the weight of each application is 1.

Example 5

upstream onmpw {
    server 192.168.144.128 weight=5;
    server 192.168.144.132;
    server 192.168.144.131;
}

In Example 5, the weight of 128 is set to 5, and the weights of the other two application servers are set to 1 by default. According to this situation, for every 7 requests, 5 will be distributed to application 128, and the other two will be distributed to applications 132 and 131 respectively.
This is the purpose of setting weights for servers.

Backup application server

In actual applications, not all of our application servers need to participate. Among them, we can use one or several applications as standby servers. When other application servers have problems and cannot be accessed, Nginx will automatically start the standby application server.

Let's make some changes based on Example 5.

Example 6

upstream onmpw {
    server 192.168.144.128 weight=5;
    server 192.168.144.132;
    server 192.168.144.131 backup;
}

The difference between this example and Example 5 is that application 131 is used as a backup server. When there are no 6 requests, 5 will be distributed to application 128 and 1 will be distributed to application 132. Only when both application servers 128 and 132 are inaccessible, Nginx will automatically distribute the request to application 131.

The above is a brief introduction to several ways to configure load balancing. I hope it will be helpful to everyone.