Microservice avalanche problems and Solutions

Avalanche problem

Avalanche problem

In microservice , The relationship between services is complicated , A microservice often depends on multiple other microservices .

Pictured , If the service provider I Something went wrong , Part of the business of the current application depends on services I, Therefore, it will also be blocked . here , Others do not depend on services I Our business seems unaffected .

however , dependent service I Your service request is blocked , The user will not get a response , be tomcat This thread will not release , As a result, more and more user requests come , More and more threads will block ：

The server supports a limited number of threads and concurrency , Requests keep blocking , Will cause the server to run out of resources , As a result, all other services are unavailable , Then the current service is unavailable .

that , Other services that depend on the current service over time , It will eventually become unavailable , Cascading failed , An avalanche happened ：

Solution

There are four common ways to solve the avalanche problem ：

1. timeout handler

timeout handler ： Set the timeout , If the request exceeds a certain time and there is no response, an error message is returned , Will not wait endlessly

2. Bin wall mode

programme 2： Bin wall mode

The silo wall model comes from the design of the cabin ：

The cabin will be separated into multiple independent spaces by partitions , When the hull is damaged , It will only cause part of the space to enter , Control the fault within a certain range , Avoid flooding the whole hull .

It's similar to , The warehouse wall mode can limit the number of threads that can be used by each business , Avoid exhausting the whole tomcat Resources for , Therefore, it is also called thread isolation .

3. Circuit breaker

Breaker mode ： from Circuit breaker Statistics the abnormal proportion of business execution , If the threshold is exceeded, it will Fuse The business , Intercept all requests to access the service .

The circuit breaker counts the number of requests to access a service , Abnormal proportion ：

When found accessing the service D When the proportion of abnormal requests is too high , Think service D There is a risk of avalanches , Will block access to the service D All your requests , Form a fuse ：

4. Current limiting

flow control ： Restrictions on business access QPS（ Number of requests per second ）, Avoid service failure due to sudden increase of traffic .

summary

What is the avalanche problem ？

• Micro services call each other , Because a service in the call chain failed , Cause the whole link to be inaccessible .

It can be said that ：

Current limiting Is the protection of services , Avoid service failure caused by instantaneous high concurrent traffic , To avoid avalanches . It's a kind of The prevention of measures .

timeout handler 、 Thread isolation 、 Degraded fuse When some services fail , Control the fault within a certain range , Avoid avalanches . It's a kind of Remedy measures .