The health check is performed on the backend server from the listening dimension.

TCP listening You can use TCP or HTTP to perform health checks;

UDP listening Use UDP for health check;

HTTP/HTTPS monitoring Use HTTP for health check.

The following are the most representative HTTP Health check of:

It is recommended to leave the "domain name", "check port" and "check path" as default. If there are special requirements, please refer to the parameter description.

Parameter description:

For TCP/HTTP/HTTPS health check

Health check status

Whether to enable health check? Since health check is very important to ensure service availability, it is recommended to enable it.

domain name

The check item of HTTP protocol health check. The load balancer performs health checks in the form of http head requests.

Some application servers will verify the host field in the client request, that is, the request header must contain the host field, or the request will be rejected. Configuring the domain name here means attaching the host field. It depends on the back-end application server. To prevent the back-end server from rejecting the health check request of load balancing.

Check port

The service port of the back-end server. It defaults to "Back end Protocol [Port]" in "Basic Configuration" and can be changed.

Check path

Also called URI, it is the access path of the back-end server.

For example, the page address to be checked is: //192.168.1.2:8080/help/2453.html , the check path (URI) is/help/2453.html.

Normal status code

The check item of HTTP protocol health check. Listen for the status code returned when checking the backend node through the head request method of the http protocol. The most common ones are 200 (normal), 301 (redirect), 404 (page not found), etc.

Response timeout

The time to wait for a response when the load balancing listener checks the backend server. If the response time is exceeded, the health check is considered as failed. The time range is 1-50 seconds.

Health Check Interval

The interval between health checks. The time range is 1-5 seconds.

Unhealthy threshold

The number of times to continuously check whether the backend ECS is abnormal. In the process of ECS health check, if the backend ECS is checked for abnormal status for n consecutive times, it is considered that the backend ECS is invalid and the load cluster is kicked out. Range 1-10 times.

Health threshold

The number of times the backend ECS is checked to be normal. In the process of ECS health check, if the backend ECS is checked for normal status for n consecutive times, it is considered that the backend ECS is back to normal and rejoins the load cluster. Range 1-10 times.

Supplementary notes

By default, the load balancing instance will use the default value

The default values for TCP/UDP/HTTP/HTTPS listening are:

Configure default value Response timeout 3 seconds Health check interval 3 seconds Unhealthy threshold 2 times Health threshold 4 times

Health check failure time window =3 second response timeout × 2 checks+3 second check interval × 1 interval= 9 seconds

Health check success time window =3 second inspection interval × 4 inspections= 12 seconds

In other words, if the backend server does not respond to the health check of listening within 9 seconds, it will be kicked out of the load cluster; If the server kicked out of the cluster can respond to the listening health check within 12 seconds, it should be re added to the load cluster.

be careful:

If you have higher requirements, you can configure to reduce the check failure time window, but you must first ensure that the request processing time of the service in the normal state is less than this value.

For UDP health check

Health Check Request&Health Check Return Result

Due to the particularity of UDP protocol, for users who use UDP health check, you need to know how UDP back-end applications work. That is to say, for a client request (customized), you know what kind of return results can be obtained after the backend server responds. Load balancing only forwards packets to UDP, and health check matches the request and result defined by you.