继承方式
HystrixCommand
public class UserSelectAllCommand extends HystrixCommand<List<User>> { private RestTemplate restTemplate; /** * 设置线程组 和命令名用于仪表盘统计信息 * 设置线程组 可以使同一个组使用同一个线程池 * .andThreadPoolKey(HystrixThreadPoolKey.Factory.asKey("ThreadPoolKey") 可以跟细粒度的线程池划分 * @param restTemplate */ public UserSelectAllCommand(RestTemplate restTemplate){ super(HystrixCommand.Setter.withGroupKey(HystrixCommandGroupKey.Factory.asKey("YouGroupName")) .andCommandKey(HystrixCommandKey.Factory.asKey("YouCommandName")) .andCommandPropertiesDefaults(HystrixCommandProperties.Setter() .withExecutionTimeoutInMilliseconds(10000)//设置超时时间。我这边全局设置无效 就对应设置 )); this.restTemplate=restTemplate; } @Override protected List<User> run() throws Exception { return restTemplate.getForObject("http://PROVIDER/user/findAll",List.class); } }
@Controller @RequestMapping("/UserHystrixCommand") public class UserHystrixCommandContorller { @Autowired RestTemplate restTemplate; //同步执行 @RequestMapping("/findAll") @ResponseBody public List<User> findAll() { UserSelectAllCommand userSelectAllCommand = new UserSelectAllCommand(restTemplate); return userSelectAllCommand.execute(); } //异步 @RequestMapping("/findAllAsyn") @ResponseBody public List<User> findAllAsyn() throws ExecutionException, InterruptedException { UserSelectAllCommand userSelectAllCommand = new UserSelectAllCommand(restTemplate); return userSelectAllCommand.queue().get(); } }
调用execute同步执行 queue 返回Future 异步执行
还可以通过执行
Observable observable= userSelectAllCommand.toObservable();//订阅的时候发起请求
Observable observable=userSelectAllCommand.observe();//立即发起请求
通过订阅获得请求结果
observable.subscribe(new Subscriber() { @Override public void onCompleted() { } @Override public void onError(Throwable throwable) { } @Override public void onNext(Object o) { } });
HystrixObservableCommand
与HystrixCommand是可以发射多次结果
public class UserSelectObservableCommand extends HystrixObservableCommand<User> { /** * 设置线程组 和命令名用于仪表盘统计信息 * 设置线程组 可以使同一个组使用同一个线程池 * .andThreadPoolKey(HystrixThreadPoolKey.Factory.asKey("ThreadPoolKey") 可以跟细粒度的线程池划分 * @param restTemplate */ @Autowired RestTemplate restTemplate; private List<Integer> ids; public UserSelectObservableCommand(List<Integer> ids, RestTemplate restTemplate) { super(Setter.withGroupKey(HystrixCommandGroupKey.Factory.asKey("YouGroupName")) .andCommandKey(HystrixCommandKey.Factory.asKey("YouCommandName")) .andCommandPropertiesDefaults(HystrixCommandProperties.Setter() .withExecutionTimeoutInMilliseconds(10000)//设置超时时间。我这边全局设置无效 就对应设置 )); this.restTemplate=restTemplate; this.ids=ids; } @Override protected Observable<User> construct() { return Observable.create(new Observable.OnSubscribe<User>() { @Override public void call(Subscriber<? super User> subscriber) { try{ if(!subscriber.isUnsubscribed()){ for (Integer id: ids) { MultiValueMap<String, String> map= new LinkedMultiValueMap<String, String>(); map.add("id",id.toString()); HttpHeaders headers = new HttpHeaders(); headers.setContentType(MediaType.APPLICATION_FORM_URLENCODED); HttpEntity<MultiValueMap<String, String>> request = new HttpEntity<MultiValueMap<String, String>>(map, headers); //调用多次服务 User user=restTemplate.postForEntity("http://PROVIDER/user/findById", request,User.class).getBody(); subscriber.onNext(user); } } subscriber.onCompleted(); }catch (Exception e){ e.printStackTrace(); subscriber.onError(e); } } }).subscribeOn(Schedulers.io()); } /** * 服务降级执行逻辑 * 错误 超时 线程池拒绝 断路器熔断 执行 * @return */ @Override protected Observable<User> resumeWithFallback() { return Observable.create(new Observable.OnSubscribe<User>() { @Override public void call(Subscriber<? super User> observer) { try { if (!observer.isUnsubscribed()) { User u = new User(); observer.onNext(u); observer.onCompleted(); } } catch (Exception e) { observer.onError(e); } } }).subscribeOn(Schedulers.io()); } }
//异步 @RequestMapping("/process") @ResponseBody public void process() throws ExecutionException, InterruptedException { UserSelectObservableCommand userSelectObservableCommand=new UserSelectObservableCommand(Arrays.asList(new Integer[]{1,2,3,4,6}),restTemplate); Observable<User> observable= userSelectObservableCommand.observe(); observable.subscribe(new Subscriber<User>(){ List<User> users=new ArrayList<User>(); @Override public void onCompleted() { users.stream().forEach(c->{ System.out.println(c.getName()); }); } @Override public void onError(Throwable throwable) { } @Override public void onNext(User user) { users.add(user); } }); }
注解方式
@Component public class UserService { @Autowired RestTemplate restTemplate; @HystrixCommand(groupKey = "userService",commandKey = "findAll") public List<User> findAll(){ return restTemplate.getForObject("http://PROVIDER/user/findAll",List.class); } @HystrixCommand public Future<List<User>> findAllAsyn(){ return new AsyncResult<List<User>>() { @Override public List<User> invoke() { return findAll(); } }; } /** * ObservableExecutionMode.EAGER observe () * ObservableExecutionMode.LAZY toObservable () * ignoreExceptions 排除异常 * @param id * @return */ @HystrixCommand(observableExecutionMode = ObservableExecutionMode.EAGER,ignoreExceptions = BusinessException.class) public Observable<User> findUserByIdA(Integer id){ return Observable.create(new Observable.OnSubscribe<User>(){ @Override public void call(Subscriber<? super User> subscriber) { //判断是否取消订阅 if(subscriber.isUnsubscribed()){ User user=restTemplate.getForObject("http://PROVIDER/user/findById",User.class); subscriber.onNext(user); subscriber.onCompleted(); } } }); } }
@Controller @RequestMapping("/UserHystrixCommandAnotation") public class UserHystrixCommandAnotationContorller { @Autowired UserService userService; //同步执行 @RequestMapping("/findAll") @ResponseBody public List<User> findAll(){ return userService.findAll(); } //异步 @RequestMapping("/findAllAsyn") @ResponseBody public List<User> findAllAsyn() throws ExecutionException, InterruptedException { return userService.findAllAsyn().get(); } }
请求缓存
继承方式
public class UserSelectAllCommand extends HystrixCommand<List<User>> { private RestTemplate restTemplate; public static final HystrixCommandKey hystrixCommandKey=HystrixCommandKey.Factory.asKey("findAll");; /** * 设置线程组 和命令名用于仪表盘统计信息 * 设置线程组 可以使同一个组使用同一个线程池 * .andThreadPoolKey(HystrixThreadPoolKey.Factory.asKey("ThreadPoolKey") 可以跟细粒度的线程池划分 * @param restTemplate */ public UserSelectAllCommand(RestTemplate restTemplate){ super(HystrixCommand.Setter.withGroupKey(HystrixCommandGroupKey.Factory.asKey("UserService")) .andCommandKey(hystrixCommandKey) .andCommandPropertiesDefaults(HystrixCommandProperties.Setter() .withExecutionTimeoutInMilliseconds(10000)//设置超时时间。我这边全局设置无效 就对应设置 )); this.restTemplate=restTemplate; } @Override protected List<User> run() throws Exception { System.out.println("执行了...."); return restTemplate.getForObject("http://PROVIDER/user/findAll",List.class); } /** * 只需要重写这个方法 将开启缓存 * @return */ @Override protected String getCacheKey() { return "UserSelectAllCommand";//因为没有参数所以key是用类名 } /** * 清除缓存 */ public static void clearKey(){ HystrixRequestCache.getInstance(UserSelectAllCommand.hystrixCommandKey,HystrixConcurrencyStrategyDefault.getInstance() ).clear("UserSelectAllCommand"); } }
@Controller @RequestMapping("/UserHystrixCommand") public class UserHystrixCommandContorller { @Autowired RestTemplate restTemplate; //同步执行 @RequestMapping("/findAll") @ResponseBody public List<User> findAll() { //开启缓存后 必须初始化一个context HystrixRequestContext.initializeContext(); UserSelectAllCommand userSelectAllCommand = new UserSelectAllCommand(restTemplate); userSelectAllCommand.execute(); userSelectAllCommand = new UserSelectAllCommand(restTemplate); userSelectAllCommand.execute(); userSelectAllCommand = new UserSelectAllCommand(restTemplate); //清空缓存 UserSelectAllCommand.clearKey(); return userSelectAllCommand.execute(); } }
注意每个HystrixCommand命令只能调用一次 多次调用会报错
注解方式
@CacheResult | 标识结果被缓存 必须配合@HystrixCommand使用 可以使用cacheKeyMethod或者CacheKey设置缓存key | cacheKeyMethod 标识获得缓存key的方法名 参数形式必须与目标方法一致 |
@CacheRemove | 标识将清除指定key的缓存 commandKey 必须指定 用于定位到清除指定命令的缓存cacheKeyMethod 指定清除缓存key或者使用CacheKey指定 | cacheKeyMethod 获得清除缓存key的方法名 参数形式必须与目标方法一致 commandKey 指定需要清除指定命令的缓存 |
@Cachekey | 标识指定目标为缓存的key优先级比cacheKeyMethod低 |
指定缓存key的几种方式
@Component public class UserService { @Autowired RestTemplate restTemplate; /** * 如果没有指定cacheKey 则默认是参数 * * @param id * @return */ @CacheResult @HystrixCommand(ignoreExceptions = BusinessException.class) public User findUserById(Integer id) { System.out.println("执行了。。。。"); User user = restTemplate.getForObject("http://PROVIDER/user/findById?id={id}", User.class, id); return user; } }
@RequestMapping("/findById") @ResponseBody public User findById(Integer id){ //开启缓存后 必须初始化一个context 可以在过滤器统一实现 HystrixRequestContext.initializeContext(); userService.findUserById(id); userService.findUserById(id); userService.findUserById(id); return userService.findUserById(id); }
使用cacheMethod定义缓存key
/** * 如果没有指定cacheKey 则默认是参数 * @param id * @return */ @CacheResult(cacheKeyMethod= "getFindUserByIdKey") @HystrixCommand(ignoreExceptions = BusinessException.class) public User findUserById(Integer id) { System.out.println("执行了。。。。"); User user = restTemplate.getForObject("http://PROVIDER/user/findById?id={id}", User.class, id); return user; } /** * 参数要与上面指定方法的一致 * @param id * @return */ public String getFindUserByIdKey(Integer id){ return String.valueOf(id); }
使用cacheKey定义缓存的key
/** * 如果没有指定cacheKey 则默认是参数 * @param id * @return */ @CacheResult @HystrixCommand(ignoreExceptions = BusinessException.class) public User findUserById(@CacheKey Integer id) { System.out.println("执行了。。。。"); User user = restTemplate.getForObject("http://PROVIDER/user/findById?id={id}", User.class, id); return user; }
如果是对象可以使用
@CacheResult @HystrixCommand(ignoreExceptions = BusinessException.class) public User findUserById(@CacheKey("{属性名字}") User user) { System.out.println("执行了。。。。"); User user = restTemplate.getForObject("http://PROVIDER/user/findById?id={id}", User.class, id); return user; }
清除缓存
/** * CacheRemove.commandKey必须指定 通过他能够找到缓存的位置然后通过key删除 * * @param user * @return */ @CacheRemove(commandKey = "findUserById") @HystrixCommand(ignoreExceptions = BusinessException.class) public boolean saveEdit(@CacheKey("id") User user) { return true; }
commandKey必须指定 用于定位某个命令的key 没显式指定命令 则为方法名 几种指定的key的方式和缓存一致
请求合并
继承的方式
1.准备一个批量查询的Service
@Service public class UserService { @Autowired private RestTemplate restTemplate; public List<User> findAll(List<Long> ids){ List<User> users=restTemplate.getForObject("http://PROVIDER/user?ids={1}", List.class, StringUtils.join(ids,",")); return users; } }
2.准备一个批处理Command
public class UserBatchCommand extends HystrixCommand<List<User>> { UserService userService; List<Long> ids; public UserBatchCommand(UserService userService,List<Long> userIds){ super(Setter.withGroupKey(asKey("userServiceCommand")).andCommandPropertiesDefaults(HystrixCommandProperties.Setter() .withExecutionTimeoutInMilliseconds(10000)));//设置超时时间。我这边全局设置无效 就对应设置 this.userService=userService; this.ids=userIds; } @Override protected List<User> run() throws Exception { return userService.findAll(ids); } }
3.定义请求合并器
/** * 第一个泛型参数 为批量处理的请求的返回类型 * 第二个泛型参数 为单个请求的返回类型 * 第三个泛型参数 为参数类型 */ public class UserCollapseCommand extends HystrixCollapser<List<User>,User ,Long> { UserService userService; Long userId; /** * 用于获取请求参数 * @return */ @Override public Long getRequestArgument() { return userId; } public UserCollapseCommand(UserService userService,Long userId){ super(Setter.withCollapserKey(HystrixCollapserKey.Factory.asKey("userCollapsercommand")) .andCollapserPropertiesDefaults(HystrixCollapserProperties.Setter().withTimerDelayInMilliseconds(100))); this.userService=userService; this.userId=userId; } /** * 合并请求产生批量处理的方法 * @param collection * @return */ @Override protected HystrixCommand<List<User>> createCommand(Collection<CollapsedRequest<User, Long>> collection) { List<Long> userIds=new ArrayList<Long>(collection.size()); userIds.addAll(collection.stream().map(CollapsedRequest::getArgument).collect(Collectors.toList())); return new UserBatchCommand(userService,userIds); } /** * 批量请求获得结果后 将结果拆分 返回给各个原子请求 * @param users * @param collection */ @Override protected void mapResponseToRequests(List<User> users, Collection<CollapsedRequest<User, Long>> collection) { int count=0; ObjectMapper objectMapper=new ObjectMapper(); for(CollapsedRequest<User,Long> collapsedRequest:collection) { User user =objectMapper.convertValue(users.get(count++),User.class); collapsedRequest.setResponse(user); } } }
4.测试
@RunWith(SpringJUnit4ClassRunner .class) @SpringBootTest(classes={SpringcloudConsumerApplication.class, hystrixCollapse.hystrixCollapserTest.class}) public class hystrixCollapserTest { @Autowired UserService userService; @Test public void simpleTest() throws ExecutionException, InterruptedException { HystrixRequestContext context = HystrixRequestContext.initializeContext(); List<Future<User>> user=new ArrayList<Future<User>>(); for(long id=0;id<10;id++){ UserCollapseCommand userCollapseCommand=new UserCollapseCommand(userService,id); User user1=userCollapseCommand.queue().get(); System.out.print(user1.getId()); } Thread.sleep(4000); } }
当在一定窗口期内 的请求 会合并成一个请求 通过HystrixCollapserProperties.Setter().withTimerDelayInMilliseconds(100) 默认100毫秒
注解方式
@Service public class UserSerivice { @Autowired private RestTemplate restTemplate; @HystrixCollapser(batchMethod = "findAll",collapserProperties = {@HystrixProperty(name = "timerDelayInMilliseconds",value="100")}) public User find(Long id){ return restTemplate.getForObject("http://PROVIDER/user/{1}", User.class, id); } /** * 直接返回list会转为linkendHashMap 所以这里手动转了一次 正式环境 都用包装类封装一次 ProcessResult7yt g * @param ids * @return */ @HystrixCommand(commandProperties={ @HystrixProperty(name = "execution.isolation.thread.timeoutInMilliseconds", value = "10000")}) public List<User> findAll(List<Long> ids){ List<User> users=restTemplate.getForObject("http://PROVIDER/user?ids={1}", List.class, StringUtils.join(ids,",")); ObjectMapper objectMapper=new ObjectMapper(); List<User> userList=new ArrayList<User>(); for (Object obj:users ) { userList.add(objectMapper.convertValue(obj,User.class)); } return userList; } }
合并器原理图
未使用合并器
使用合并器后
请求合并器虽然能节省线程池的开销 但是因为有窗口期 如果窗口10m 一个请求耗时需要5m 也会等到窗口期过才发起请求
窗口期内有3个以上请求 才推荐使用请求合并