What 1 hystrix that?

Netflix (the largest foreign similar, love Fantastic Art, Youku) video site five or six years ago, too, felt the system, the entire site, often a failure, availability is not too high

Sometimes some of the vip members can not pay, sometimes watching video on Caton, could not watch the video. . .

Affect the company's revenue !!!

Fifty-six years ago, netflix, api team, to enhance the high-availability, the development of a framework similar to the spring, mybatis, hibernate, and so on such a framework

High availability framework, hystrix

hystrix, framework, offers a variety of features related to high availability, and then make sure to say under the protection hystrix, the entire system can be in the long-term availability of state

Under ideal conditions, the failure of the software, it should not say leads to some failures collapse of the entire system, server hardware, redundant service

The only likely to lead to the total collapse of the system is similar to before, the accident Alipay, construction workers, Waduan cable, leading to several rooms have blackout

Not available, and have some fault or bug distinction

2 hystrix functionality provided

Resource isolation, current limiting, fuse, demotion, operation and maintenance monitoring

Resource isolation

Let your system, a piece of something, in case of failure, the system will not run out of resources, such as thread resources

I was actually a case of the project, there is a thing, is to use multiple threads to do something, junior partner doing the project, did not pay attention too, resource isolation, that piece of code, in the case of encounter some failures, each threads running at the time, because the bug, directly on the cycle of death, resulting in a large piece of something to start threads, each thread infinite loop

I eventually leading to system resource depletion, collapse, does not work, is not available, was destroyed

Resource isolation, that piece of code, up to a maximum of 10 threads is spent, not more, and to destroy, and limited number of good resources

Limiting

High concurrent influx of traffic coming in, for example, suddenly a second one million QPS, was destroyed, 100,000 QPS into the system, other 900 000 QPS was rejected

Fuse

Some systems rely on back-end, which some fault, for example, hang mysql, each request is given, and a fuse, not directly over the subsequent requests received, denied access, and then try to see mysql After 10 minutes no recovery

Demote

mysql hung up, the system is found, automatically downgrade, small amounts of data stored in the memory, to extract some data out

Operation and maintenance monitoring

+ + Optimized alarm monitoring, various abnormal situations, there are problems timely alarm, and parameter optimization of some system configuration, or the code

3 How to explain this piece of content?

(1) How eshop-cache, cache core service transformed into a highly available architecture
(2) part of hystrix in a single pull out, made a free small courses, as benefits paid out
(3) eshop-cache , write code, eshop-cache-ha, business scenarios, join up with before, to re-write the code
(4) hystrix do a service availability of this content, just explain to a business context, to re-write the code, independent

eshop-cache, cache data in case of failure at all levels, will re-call interface from the source system, depending on the source system to query the mysql database to retrieve data

If you depend on a variety of services with failure, then it may cause your system unusable

hystrix the system a variety of high-availability systems strengthening to cope with various situations unavailable

4 design principles of Hystrix

Dependence on service calls occur when call delay call fails and control and fault protection.
In the complex distributed system, one of its dependencies prevent service failure spread throughout the system. For example, one of service failure, leading to other services also followed the failure.
Provide fail-fast (fast failure) and rapid recovery support.
Providing graceful degradation fallback support.
Support for near real-time monitoring, alarm and operation and maintenance operations.

for example:

There is a distributed system, dependent on the service A service B, service B depends on the service C / D / E. In such a sophisticated system, for example, it may be at most 100 thread resources. Under normal circumstances, 40 concurrent threads calling service C, each 30 concurrent threads to call D / E.

C call service, just 20ms, and now because of the failure of the service C, such as delay, or hung up, then thread Diaozhu about 2s. 40 threads all been stuck due to the continuous influx of requests, other threads can also be used to invoke the service C, the same will be stuck. This leads to thread B service resources are exhausted, can not receive a new request, even as a large number of threads continue to operate, leading to their downtime. A service also hung up.

Hystrix resources can be isolated, such as restrictions on service B is only 40 thread calls Services C. When this 40 thread hang live, the other 60 threads are still normal call work. In order to ensure that the entire system does not collapse.

5 Hystrix design principles more details

Prevent any dependent services run out of resources, such as resources, all threads in the tomcat.
Request queuing and avoiding a backlog, current-limiting and fail fast fault controlled.
Downgrade to provide fallback mechanism to deal with failure.
Use resource isolation techniques, such as bulkhead (bulkhead isolation technology), swimlane (lane technology), circuit breaker (circuit technology) to limit the impact of any failure of a dependent service.
Through near real-time statistics / monitoring / alarm function, to improve the speed of faults found.
Through near real-time thermal properties and configuration modification capabilities to improve troubleshooting and speed recovery.
Protect all fault conditions depend on service calls, not just of a network failure.

Reference:
https://www.nowcoder.com/discuss/188033

　　https://www.cnblogs.com/yzh-blog/p/11703037.html#1817129157

springcloud-hystrix face questions