Hystrix服务降级,其实就是线程池中单个线程障处理,防止单个线程请求时间太长,导致资源长期被占有而得不到释放,从而导致线程池被快速占用完,导致服务崩溃。
Hystrix能解决如下问题:
1.请求超时降级,线程资源不足降级,降级之后可以返回自定义数据
2.线程池隔离降级,分布式服务可以针对不同的服务使用不同的线程池,从而互不影响
3.自动触发降级与恢复
4.实现请求缓存和请求合并
下面我们主要展示一下超时降级
测试1:设置getCourse sleep 3s,打印“getFallback HystrixCommand course.”
测试2:设置getCourse sleep 1s,打印“get course success.”
HystrixCommand实现熔断
public class GetCourseHystrixCommand extends HystrixCommand<String> {
private CourseService courseService;
protected GetCourseHystrixCommand(CourseService courseService) {
super(setter());
this.courseService= courseService;
}
@Override protected String run() throws Exception {
return courseService.getCourse();
}
@Override
protected String getFallback(){
return "getFallback HystrixCommand course.";
}
private static Setter setter(){
HystrixCommandGroupKey commandGroupKey= HystrixCommandGroupKey.Factory.asKey("course");
HystrixCommandProperties.Setter commandProperties= HystrixCommandProperties.Setter()
.withExecutionIsolationStrategy(HystrixCommandProperties.ExecutionIsolationStrategy.THREAD)
.withFallbackEnabled(true)//是否启用降级
.withFallbackIsolationSemaphoreMaxConcurrentRequests(10)//设置getFallback并发请求信号量,如果超过信号量值,不会再走直getFallback而直接快速失败
.withExecutionIsolationThreadInterruptOnFutureCancel(true)
.withExecutionIsolationThreadInterruptOnTimeout(true)//当隔离策略为Thread时候,当线程超时时,是否中断处理
.withExecutionTimeoutEnabled(true)//是否启用超时机制
.withExecutionTimeoutInMilliseconds(2000);//超时时间设置
return HystrixCommand.Setter.withGroupKey(commandGroupKey)
.andCommandPropertiesDefaults(commandProperties);
}
public static void main(String[] args) throws ExecutionException, InterruptedException {
GetCourseHystrixCommand courseHystrixCommand= new GetCourseHystrixCommand(new CourseService());
Future<String> future= courseHystrixCommand.queue();
System.out.println(future.get());
}
}
//辅助类
public class CourseService {
public String getCourse() throws InterruptedException {
Thread.sleep(3000L);
return "get course success.";
}
}
网友评论