延迟任务
目录
- 延迟任务
- 技术对比:
- Redis实现定时任务:编辑
- 新增任务:
- 取消任务:
- 拉取任务:
- Zset定时刷新数据到List中:
- 分布式锁实现定时任务只刷新一次:
技术对比:
Redis实现定时任务:
问题1:为什么任务需要存储在数据库中?
问题2:为什么使用两种Redis的数据结构(List和Zset)?
问题3:添加Zset数据时,为什么需要预加载?
新增任务:
添加任务到数据库进行持久化----->添加任务到Redis队列中------>如果任务的执行时间小于等于当前时间,则添加到当前任务List队列中------>如果任务的执行时间大于当前时间并且小于预设时间,就把其放到未来队列Zset中
@Service
@Transactional
@Slf4j
public class TaskServiceImpl implements TaskService {/*** 添加延迟任务** @param task* @return*/@Overridepublic long addTask(Task task) {//1.添加任务到数据库中boolean success = addTaskToDb(task);if (success) {//2.添加任务到redisaddTaskToCache(task);}return task.getTaskId();}@Autowiredprivate CacheService cacheService;/*** 把任务添加到redis中** @param task*/private void addTaskToCache(Task task) {String key = task.getTaskType() + "_" + task.getPriority();//获取5分钟之后的时间 毫秒值Calendar calendar = Calendar.getInstance();calendar.add(Calendar.MINUTE, 5);long nextScheduleTime = calendar.getTimeInMillis();//2.1 如果任务的执行时间小于等于当前时间,存入listif (task.getExecuteTime() <= System.currentTimeMillis()) {cacheService.lLeftPush(ScheduleConstants.TOPIC + key, JSON.toJSONString(task));} else if (task.getExecuteTime() <= nextScheduleTime) {//2.2 如果任务的执行时间大于当前时间 && 小于等于预设时间(未来5分钟) 存入zset中cacheService.zAdd(ScheduleConstants.FUTURE + key, JSON.toJSONString(task), task.getExecuteTime());}}@Autowiredprivate TaskinfoMapper taskinfoMapper;@Autowiredprivate TaskinfoLogsMapper taskinfoLogsMapper;/*** 添加任务到数据库中** @param task* @return*/private boolean addTaskToDb(Task task) {boolean flag = false;try {//保存任务表Taskinfo taskinfo = new Taskinfo();BeanUtils.copyProperties(task, taskinfo);taskinfo.setExecuteTime(new Date(task.getExecuteTime()));taskinfoMapper.insert(taskinfo);//设置taskIDtask.setTaskId(taskinfo.getTaskId());//保存任务日志数据TaskinfoLogs taskinfoLogs = new TaskinfoLogs();BeanUtils.copyProperties(taskinfo, taskinfoLogs);taskinfoLogs.setVersion(1);taskinfoLogs.setStatus(ScheduleConstants.SCHEDULED);taskinfoLogsMapper.insert(taskinfoLogs);flag = true;} catch (Exception e) {e.printStackTrace();}return flag;}
}
取消任务:
把数据库的任务删除----->更新任务的日志------>删除redis队列中的数据
/*** 取消任务* @param taskId* @return*/
@Override
public boolean cancelTask(long taskId) {boolean flag = false;//删除任务,更新日志Task task = updateDb(taskId,ScheduleConstants.EXECUTED);//删除redis的数据if(task != null){removeTaskFromCache(task);flag = true;}return false;
}/*** 删除redis中的任务数据* @param task*/
private void removeTaskFromCache(Task task) {String key = task.getTaskType()+"_"+task.getPriority();if(task.getExecuteTime()<=System.currentTimeMillis()){cacheService.lRemove(ScheduleConstants.TOPIC+key,0,JSON.toJSONString(task));}else {cacheService.zRemove(ScheduleConstants.FUTURE+key, JSON.toJSONString(task));}
}/*** 删除任务,更新任务日志状态* @param taskId* @param status* @return*/
private Task updateDb(long taskId, int status) {Task task = null;try {//删除任务taskinfoMapper.deleteById(taskId);//更新日志TaskinfoLogs taskinfoLogs = taskinfoLogsMapper.selectById(taskId);taskinfoLogs.setStatus(status);taskinfoLogsMapper.updateById(taskinfoLogs);task = new Task();BeanUtils.copyProperties(taskinfoLogs,task);task.setExecuteTime(taskinfoLogs.getExecuteTime().getTime());}catch (Exception e){log.error("task cancel exception taskid={}",taskId);}return task;
}
拉取任务:
/*** 按照类型和优先级拉取任务* @return*/
@Override
public Task poll(int type,int priority) {Task task = null;try {String key = type+"_"+priority;//从List中取出任务String task_json = cacheService.lRightPop(ScheduleConstants.TOPIC + key);if(StringUtils.isNotBlank(task_json)){task = JSON.parseObject(task_json, Task.class);//更新数据库信息,删除任务并且更新日志updateDb(task.getTaskId(),ScheduleConstants.EXECUTED);}}catch (Exception e){e.printStackTrace();log.error("poll task exception");}return task;
}
Zset定时刷新数据到List中+分布式锁实现定时任务只刷新一次:
如何获取Zset中所有的key。方案一:keys()获取,不推荐。方案二:scan()获取。
通过定时任务,使用redis管道,把Zset中的数据往List中传输
/*** 未来数据定时刷新*/
@Scheduled(cron = "0 */1 * * * ?")
public void refresh(){//获取锁String token = cacheService.tryLock("FUTURE_TASK_SYNC", 1000 * 30);if(StringUtils.isNotBlank(token)){log.info("未来数据定时刷新---定时任务");//获取所有未来数据的集合keySet<String> futureKeys = cacheService.scan(ScheduleConstants.FUTURE + "*");for (String futureKey : futureKeys) {//future_100_50//获取当前数据的key topicString topicKey = ScheduleConstants.TOPIC+futureKey.split(ScheduleConstants.FUTURE)[1];//按照key和分值查询符合条件的数据Set<String> tasks = cacheService.zRangeByScore(futureKey, 0, System.currentTimeMillis());//同步数据if(!tasks.isEmpty()){cacheService.refreshWithPipeline(futureKey,topicKey,tasks);log.info("成功的将"+futureKey+"刷新到了"+topicKey);}}}
}
数据库同步到Redis中:
清除缓存----->查询出小于未来5分钟的所有任务------>新增任务到Redis中
@Scheduled(cron = "0 */5 * * * ?")
@PostConstruct
public void reloadData() {clearCache();log.info("数据库数据同步到缓存");Calendar calendar = Calendar.getInstance();calendar.add(Calendar.MINUTE, 5);//查看小于未来5分钟的所有任务List<Taskinfo> allTasks = taskinfoMapper.selectList(Wrappers.<Taskinfo>lambdaQuery().lt(Taskinfo::getExecuteTime,calendar.getTime()));if(allTasks != null && allTasks.size() > 0){for (Taskinfo taskinfo : allTasks) {Task task = new Task();BeanUtils.copyProperties(taskinfo,task);task.setExecuteTime(taskinfo.getExecuteTime().getTime());addTaskToCache(task);}}
}private void clearCache(){// 删除缓存中未来数据集合和当前消费者队列的所有keySet<String> futurekeys = cacheService.scan(ScheduleConstants.FUTURE + "*");// future_Set<String> topickeys = cacheService.scan(ScheduleConstants.TOPIC + "*");// topic_cacheService.delete(futurekeys);cacheService.delete(topickeys);
}
其他微服务通过Fegin远程调用Schedule服务 :
@FeignClient("leadnews-schedule")
public interface IScheduleClient {/*** 添加任务* @param task 任务对象* @return 任务id*/@PostMapping("/api/v1/task/add")public ResponseResult addTask(@RequestBody Task task);/*** 取消任务* @param taskId 任务id* @return 取消结果*/@GetMapping("/api/v1/task/cancel/{taskId}")public ResponseResult cancelTask(@PathVariable("taskId") long taskId);/*** 按照类型和优先级来拉取任务* @param type* @param priority* @return*/@GetMapping("/api/v1/task/poll/{type}/{priority}")public ResponseResult poll(@PathVariable("type") int type,@PathVariable("priority") int priority);
}
文章发布和审核:
发布:
@Autowired
private WmNewsTaskService wmNewsTaskService;/*** 发布修改文章或保存为草稿* @param dto* @return*/
@Override
public ResponseResult submitNews(WmNewsDto dto) {//0.条件判断if(dto == null || dto.getContent() == null){return ResponseResult.errorResult(AppHttpCodeEnum.PARAM_INVALID);}//1.保存或修改文章WmNews wmNews = new WmNews();//属性拷贝 属性名词和类型相同才能拷贝BeanUtils.copyProperties(dto,wmNews);//封面图片 list---> stringif(dto.getImages() != null && dto.getImages().size() > 0){//[1dddfsd.jpg,sdlfjldk.jpg]--> 1dddfsd.jpg,sdlfjldk.jpgString imageStr = StringUtils.join(dto.getImages(), ",");wmNews.setImages(imageStr);}//如果当前封面类型为自动 -1if(dto.getType().equals(WemediaConstants.WM_NEWS_TYPE_AUTO)){wmNews.setType(null);}saveOrUpdateWmNews(wmNews);//2.判断是否为草稿 如果为草稿结束当前方法if(dto.getStatus().equals(WmNews.Status.NORMAL.getCode())){return ResponseResult.okResult(AppHttpCodeEnum.SUCCESS);}//3.不是草稿,保存文章内容图片与素材的关系//获取到文章内容中的图片信息List<String> materials = ectractUrlInfo(dto.getContent());saveRelativeInfoForContent(materials,wmNews.getId());//4.不是草稿,保存文章封面图片与素材的关系,如果当前布局是自动,需要匹配封面图片saveRelativeInfoForCover(dto,wmNews,materials);//审核文章// wmNewsAutoScanService.autoScanWmNews(wmNews.getId());wmNewsTaskService.addNewsToTask(wmNews.getId(),wmNews.getPublishTime());return ResponseResult.okResult(AppHttpCodeEnum.SUCCESS);}
消费:
@Autowired
private WmNewsAutoScanServiceImpl wmNewsAutoScanService;/*** 消费延迟队列数据*/
@Scheduled(fixedRate = 1000)
@Override
@SneakyThrows
public void scanNewsByTask() {log.info("文章审核---消费任务执行---begin---");ResponseResult responseResult = scheduleClient.poll(TaskTypeEnum.NEWS_SCAN_TIME.getTaskType(), TaskTypeEnum.NEWS_SCAN_TIME.getPriority());if(responseResult.getCode().equals(200) && responseResult.getData() != null){String json_str = JSON.toJSONString(responseResult.getData());Task task = JSON.parseObject(json_str, Task.class);byte[] parameters = task.getParameters();WmNews wmNews = ProtostuffUtil.deserialize(parameters, WmNews.class);System.out.println(wmNews.getId()+"-----------");wmNewsAutoScanService.autoScanWmNews(wmNews.getId());}log.info("文章审核---消费任务执行---end---");
}