聊聊PowerJob的InstanceStatusCheckService

序

本文主要研究一下PowerJob的InstanceStatusCheckService

InstanceStatus

tech/powerjob/common/enums/InstanceStatus.java

@Getter
@AllArgsConstructor
public enum InstanceStatus {
    /**
     *
     */
    WAITING_DISPATCH(1, "等待派发"),
    WAITING_WORKER_RECEIVE(2, "等待Worker接收"),
    RUNNING(3, "运行中"),
    FAILED(4, "失败"),
    SUCCEED(5, "成功"),
    CANCELED(9, "取消"),
    STOPPED(10, "手动停止");

    private final int v;
    private final String des;

    /**
     * 广义的运行状态
     */
    public static final List GENERALIZED_RUNNING_STATUS = Lists.newArrayList(WAITING_DISPATCH.v, WAITING_WORKER_RECEIVE.v, RUNNING.v);
    /**
     * 结束状态
     */
    public static final List FINISHED_STATUS = Lists.newArrayList(FAILED.v, SUCCEED.v, CANCELED.v, STOPPED.v);

    public static InstanceStatus of(int v) {
        for (InstanceStatus is : values()) {
            if (v == is.v) {
                return is;
            }
        }
        throw new IllegalArgumentException("InstanceStatus has no item for value " + v);
    }
}

InstanceStatus定义了任务实例的状态，广义运行中的状态为WAITING_DISPATCH、WAITING_WORKER_RECEIVE、RUNNING；终态为FAILED、SUCCEED、CANCELED、STOPPED

InstanceStatusCheckService

tech/powerjob/server/core/scheduler/InstanceStatusCheckService.java

@Slf4j
@Service
@RequiredArgsConstructor
public class InstanceStatusCheckService {

    private static final int MAX_BATCH_NUM_APP = 10;
    private static final int MAX_BATCH_NUM_INSTANCE = 3000;
    private static final int MAX_BATCH_UPDATE_NUM = 500;
    private static final long DISPATCH_TIMEOUT_MS = 30000;
    private static final long RECEIVE_TIMEOUT_MS = 60000;
    private static final long RUNNING_TIMEOUT_MS = 60000;
    private static final long WORKFLOW_WAITING_TIMEOUT_MS = 60000;

    public static final long CHECK_INTERVAL = 10000;

    private final TransportService transportService;

    private final DispatchService dispatchService;

    private final InstanceManager instanceManager;

    private final WorkflowInstanceManager workflowInstanceManager;

    private final AppInfoRepository appInfoRepository;

    private final JobInfoRepository jobInfoRepository;

    private final InstanceInfoRepository instanceInfoRepository;

    private final WorkflowInfoRepository workflowInfoRepository;

    private final WorkflowInstanceInfoRepository workflowInstanceInfoRepository;

    //......
}    

InstanceStatusCheckService提供了checkRunningInstance、checkWaitingDispatchInstance、checkWaitingWorkerReceiveInstance、checkWorkflowInstance方法

checkRunningInstance

    public void checkRunningInstance() {
        Stopwatch stopwatch = Stopwatch.createStarted();
        // 查询 DB 获取该 Server 需要负责的 AppGroup
        List allAppIds = appInfoRepository.listAppIdByCurrentServer(transportService.defaultProtocol().getAddress());
        if (CollectionUtils.isEmpty(allAppIds)) {
            log.info("[InstanceStatusChecker] current server has no app's job to check");
            return;
        }
        try {
            // 检查 RUNNING 状态的任务（一定时间没收到 TaskTracker 的状态报告，视为失败）
            Lists.partition(allAppIds, MAX_BATCH_NUM_APP).forEach(this::handleRunningInstance);
        } catch (Exception e) {
            log.error("[InstanceStatusChecker] RunningInstance status check failed.", e);
        }
        log.info("[InstanceStatusChecker] RunningInstance status check using {}.", stopwatch.stop());
    }

    private void handleRunningInstance(List partAppIds) {
        // 3. 检查 RUNNING 状态的任务（一定时间没收到 TaskTracker 的状态报告，视为失败）
        long threshold = System.currentTimeMillis() - RUNNING_TIMEOUT_MS;
        List failedInstances = instanceInfoRepository.selectBriefInfoByAppIdInAndStatusAndGmtModifiedBefore(partAppIds, InstanceStatus.RUNNING.getV(), new Date(threshold), PageRequest.of(0, MAX_BATCH_NUM_INSTANCE));
        while (!failedInstances.isEmpty()) {
            // collect job id
            Set jobIds = failedInstances.stream().map(BriefInstanceInfo::getJobId).collect(Collectors.toSet());
            // query job info and map
            Map jobInfoMap = jobInfoRepository.findByIdIn(jobIds).stream().collect(Collectors.toMap(JobInfoDO::getId, e -> e));
            log.warn("[InstanceStatusCheckService] find some instances have not received status report for a long time : {}", failedInstances.stream().map(BriefInstanceInfo::getInstanceId).collect(Collectors.toList()));
            failedInstances.forEach(instance -> {
                Optional jobInfoOpt = Optional.ofNullable(jobInfoMap.get(instance.getJobId()));
                if (!jobInfoOpt.isPresent()) {
                    final Optional opt = instanceInfoRepository.findById(instance.getId());
                    opt.ifPresent(e -> updateFailedInstance(e, SystemInstanceResult.REPORT_TIMEOUT));
                    return;
                }
                TimeExpressionType timeExpressionType = TimeExpressionType.of(jobInfoOpt.get().getTimeExpressionType());
                SwitchableStatus switchableStatus = SwitchableStatus.of(jobInfoOpt.get().getStatus());
                // 如果任务已关闭，则不进行重试，将任务置为失败即可；秒级任务也直接置为失败，由派发器重新调度
                if (switchableStatus != SwitchableStatus.ENABLE || TimeExpressionType.FREQUENT_TYPES.contains(timeExpressionType.getV())) {
                    final Optional opt = instanceInfoRepository.findById(instance.getId());
                    opt.ifPresent(e -> updateFailedInstance(e, SystemInstanceResult.REPORT_TIMEOUT));
                    return;
                }
                // CRON 和 API一样，失败次数 + 1，根据重试配置进行重试
                if (instance.getRunningTimes() < jobInfoOpt.get().getInstanceRetryNum()) {
                    dispatchService.redispatchAsync(instance.getInstanceId(), InstanceStatus.RUNNING.getV());
                } else {
                    final Optional opt = instanceInfoRepository.findById(instance.getId());
                    opt.ifPresent(e -> updateFailedInstance(e, SystemInstanceResult.REPORT_TIMEOUT));
                }
            });
            threshold = System.currentTimeMillis() - RUNNING_TIMEOUT_MS;
            failedInstances = instanceInfoRepository.selectBriefInfoByAppIdInAndStatusAndGmtModifiedBefore(partAppIds, InstanceStatus.RUNNING.getV(), new Date(threshold), PageRequest.of(0, MAX_BATCH_NUM_INSTANCE));
        }

    }    

checkRunningInstance查询该server负责的appId，然后挨个遍历执行handleRunningInstance；handleRunningInstance查找一定时间没收到TaskTracker状态报告的任务实例，若任务已经关闭则不进行重试，若是秒级任务则更新为失败，其他的则判断运行次数市场超过重试次数，否则通过dispatchService.redispatchAsync重试

checkWaitingDispatchInstance

    public void checkWaitingDispatchInstance() {
        Stopwatch stopwatch = Stopwatch.createStarted();
        // 查询 DB 获取该 Server 需要负责的 AppGroup
        List allAppIds = appInfoRepository.listAppIdByCurrentServer(transportService.defaultProtocol().getAddress());
        if (CollectionUtils.isEmpty(allAppIds)) {
            log.info("[InstanceStatusChecker] current server has no app's job to check");
            return;
        }
        try {
            // 检查等待 WAITING_DISPATCH 状态的任务
            Lists.partition(allAppIds, MAX_BATCH_NUM_APP).forEach(this::handleWaitingDispatchInstance);
        } catch (Exception e) {
            log.error("[InstanceStatusChecker] WaitingDispatchInstance status check failed.", e);
        }
        log.info("[InstanceStatusChecker] WaitingDispatchInstance status check using {}.", stopwatch.stop());
    }

    private void handleWaitingDispatchInstance(List partAppIds) {
        // 1. 检查等待 WAITING_DISPATCH 状态的任务
        long threshold = System.currentTimeMillis() - DISPATCH_TIMEOUT_MS;
        List waitingDispatchInstances = instanceInfoRepository.findAllByAppIdInAndStatusAndExpectedTriggerTimeLessThan(partAppIds, InstanceStatus.WAITING_DISPATCH.getV(), threshold, PageRequest.of(0, MAX_BATCH_NUM_INSTANCE));
        while (!waitingDispatchInstances.isEmpty()) {
            List overloadAppIdList = new ArrayList<>();
            long startTime = System.currentTimeMillis();
            // 按照 appId 分组处理，方便处理超载的逻辑
            Map> waitingDispatchInstancesMap = waitingDispatchInstances.stream().collect(Collectors.groupingBy(InstanceInfoDO::getAppId));
            for (Map.Entry> entry : waitingDispatchInstancesMap.entrySet()) {
                final Long currentAppId = entry.getKey();
                final List currentAppWaitingDispatchInstances = entry.getValue();
                // collect job id
                Set jobIds = currentAppWaitingDispatchInstances.stream().map(InstanceInfoDO::getJobId).collect(Collectors.toSet());
                // query job info and map
                Map jobInfoMap = jobInfoRepository.findByIdIn(jobIds).stream().collect(Collectors.toMap(JobInfoDO::getId, e -> e));
                log.warn("[InstanceStatusChecker] find some instance in app({}) which is not triggered as expected: {}", currentAppId, currentAppWaitingDispatchInstances.stream().map(InstanceInfoDO::getInstanceId).collect(Collectors.toList()));
                final Holder overloadFlag = new Holder<>(false);
                // 先这么简单处理没问题，毕竟只有这一个地方用了 parallelStream
                currentAppWaitingDispatchInstances.parallelStream().forEach(instance -> {
                    if (overloadFlag.get()) {
                        // 直接忽略
                        return;
                    }
                    Optional jobInfoOpt = Optional.ofNullable(jobInfoMap.get(instance.getJobId()));
                    if (jobInfoOpt.isPresent()) {
                        // 处理等待派发的任务没有必要再重置一次状态，减少 io 次数
                        dispatchService.dispatch(jobInfoOpt.get(), instance.getInstanceId(), Optional.of(instance), Optional.of(overloadFlag));
                    } else {
                        log.warn("[InstanceStatusChecker] can't find job by jobId[{}], so redispatch failed, failed instance: {}", instance.getJobId(), instance);
                        final Optional opt = instanceInfoRepository.findById(instance.getId());
                        opt.ifPresent(instanceInfoDO -> updateFailedInstance(instanceInfoDO, SystemInstanceResult.CAN_NOT_FIND_JOB_INFO));
                    }
                });
                threshold = System.currentTimeMillis() - DISPATCH_TIMEOUT_MS;
                if (overloadFlag.get()) {
                    overloadAppIdList.add(currentAppId);
                }
            }
            log.info("[InstanceStatusChecker] process {} task,use {} ms", waitingDispatchInstances.size(), System.currentTimeMillis() - startTime);
            if (!overloadAppIdList.isEmpty()) {
                log.warn("[InstanceStatusChecker] app[{}] is overload, so skip check waiting dispatch instance", overloadAppIdList);
                partAppIds.removeAll(overloadAppIdList);
            }
            if (partAppIds.isEmpty()) {
                break;
            }
            waitingDispatchInstances = instanceInfoRepository.findAllByAppIdInAndStatusAndExpectedTriggerTimeLessThan(partAppIds, InstanceStatus.WAITING_DISPATCH.getV(), threshold, PageRequest.of(0, MAX_BATCH_NUM_INSTANCE));
        }

    }    

checkWaitingDispatchInstance查找待派发的任务实例，通过dispatchService.dispatch进行派发

checkWaitingWorkerReceiveInstance

    public void checkWaitingWorkerReceiveInstance() {
        Stopwatch stopwatch = Stopwatch.createStarted();
        // 查询 DB 获取该 Server 需要负责的 AppGroup
        List allAppIds = appInfoRepository.listAppIdByCurrentServer(transportService.defaultProtocol().getAddress());
        if (CollectionUtils.isEmpty(allAppIds)) {
            log.info("[InstanceStatusChecker] current server has no app's job to check");
            return;
        }
        try {
            // 检查 WAITING_WORKER_RECEIVE 状态的任务
            Lists.partition(allAppIds, MAX_BATCH_NUM_APP).forEach(this::handleWaitingWorkerReceiveInstance);
        } catch (Exception e) {
            log.error("[InstanceStatusChecker] WaitingWorkerReceiveInstance status check failed.", e);
        }
        log.info("[InstanceStatusChecker] WaitingWorkerReceiveInstance status check using {}.", stopwatch.stop());
    }

    private void handleWaitingWorkerReceiveInstance(List partAppIds) {
        // 2. 检查 WAITING_WORKER_RECEIVE 状态的任务
        long threshold = System.currentTimeMillis() - RECEIVE_TIMEOUT_MS;
        List waitingWorkerReceiveInstances = instanceInfoRepository.selectBriefInfoByAppIdInAndStatusAndActualTriggerTimeLessThan(partAppIds, InstanceStatus.WAITING_WORKER_RECEIVE.getV(), threshold, PageRequest.of(0, MAX_BATCH_NUM_INSTANCE));
        while (!waitingWorkerReceiveInstances.isEmpty()) {
            log.warn("[InstanceStatusChecker] find some instance didn't receive any reply from worker, try to redispatch: {}", waitingWorkerReceiveInstances.stream().map(BriefInstanceInfo::getInstanceId).collect(Collectors.toList()));
            final List> partitions = Lists.partition(waitingWorkerReceiveInstances, MAX_BATCH_UPDATE_NUM);
            for (List partition : partitions) {
                dispatchService.redispatchBatchAsyncLockFree(partition.stream().map(BriefInstanceInfo::getInstanceId).collect(Collectors.toList()), InstanceStatus.WAITING_WORKER_RECEIVE.getV());
            }
            // 重新查询
            threshold = System.currentTimeMillis() - RECEIVE_TIMEOUT_MS;
            waitingWorkerReceiveInstances = instanceInfoRepository.selectBriefInfoByAppIdInAndStatusAndActualTriggerTimeLessThan(partAppIds, InstanceStatus.WAITING_WORKER_RECEIVE.getV(), threshold, PageRequest.of(0, MAX_BATCH_NUM_INSTANCE));
        }
    }    

checkWaitingWorkerReceiveInstance检查等待worker接收的实例，挨个执行dispatchService.redispatchBatchAsyncLockFree

checkWorkflowInstance

    public void checkWorkflowInstance() {
        Stopwatch stopwatch = Stopwatch.createStarted();
        // 查询 DB 获取该 Server 需要负责的 AppGroup
        List allAppIds = appInfoRepository.listAppIdByCurrentServer(transportService.defaultProtocol().getAddress());
        if (CollectionUtils.isEmpty(allAppIds)) {
            log.info("[InstanceStatusChecker] current server has no app's job to check");
            return;
        }
        try {
            checkWorkflowInstance(allAppIds);
        } catch (Exception e) {
            log.error("[InstanceStatusChecker] WorkflowInstance status check failed.", e);
        }
        log.info("[InstanceStatusChecker] WorkflowInstance status check using {}.", stopwatch.stop());
    }

    private void checkWorkflowInstance(List allAppIds) {

        // 重试长时间处于 WAITING 状态的工作流实例
        long threshold = System.currentTimeMillis() - WORKFLOW_WAITING_TIMEOUT_MS;
        Lists.partition(allAppIds, MAX_BATCH_NUM_APP).forEach(partAppIds -> {
            List waitingWfInstanceList = workflowInstanceInfoRepository.findByAppIdInAndStatusAndExpectedTriggerTimeLessThan(partAppIds, WorkflowInstanceStatus.WAITING.getV(), threshold);
            if (!CollectionUtils.isEmpty(waitingWfInstanceList)) {

                List wfInstanceIds = waitingWfInstanceList.stream().map(WorkflowInstanceInfoDO::getWfInstanceId).collect(Collectors.toList());
                log.warn("[WorkflowInstanceChecker] wfInstance({}) is not started as expected, oms try to restart these workflowInstance.", wfInstanceIds);

                waitingWfInstanceList.forEach(wfInstance -> {
                    Optional workflowOpt = workflowInfoRepository.findById(wfInstance.getWorkflowId());
                    workflowOpt.ifPresent(workflowInfo -> {
                        workflowInstanceManager.start(workflowInfo, wfInstance.getWfInstanceId());
                        log.info("[Workflow-{}|{}] restart workflowInstance successfully~", workflowInfo.getId(), wfInstance.getWfInstanceId());
                    });
                });
            }
        });
    }    

checkWorkflowInstance定期检测工作流实例的状态，针对WAITING的挨个执行workflowInstanceManager.start

小结

InstanceStatusCheckService提供了checkRunningInstance、checkWaitingDispatchInstance、checkWaitingWorkerReceiveInstance、checkWorkflowInstance方法，他们分别用于检查状态是运行中但是上报超时的任务实例、等待worker接收处理的任务实例、等待调度的工作流实例。