/**
* 创建任务JOb
*
* @return
*/
private JobDetail createJobDetail() {
JobDataMap jobDataMap = new JobDataMap();
jobDataMap.put("jobConfiguration", jobConfiguration);
JobDetail result =
JobBuilder.newJob(jobConfiguration.getJobClass())
.setJobData(jobDataMap)
.withIdentity(jobConfiguration.getJobName())
.build();
return result;
}
private JobDetail createJobDetail() {
JobDetail result =
JobBuilder.newJob(jobConfiguration.getJobClass())
.withIdentity(jobConfiguration.getJobName())
.build();
result.getJobDataMap().put("configService", configService);
result.getJobDataMap().put("shardingService", shardingService);
result.getJobDataMap().put("executionContextService", executionContextService);
result.getJobDataMap().put("executionService", executionService);
result.getJobDataMap().put("failoverService", failoverService);
result.getJobDataMap().put("offsetService", offsetService);
return result;
}
/** 初始化作业. */
public void init() {
log.debug("Elastic job: job controller init, job name is: {}.", jobConfiguration.getJobName());
coordinatorRegistryCenter.addCacheData("/" + jobConfiguration.getJobName());
registerElasticEnv();
jobDetail = createJobDetail();
try {
scheduler = initializeScheduler(jobDetail.getKey().toString());
scheduleJob(createTrigger(configService.getCron()));
} catch (final SchedulerException ex) {
throw new JobException(ex);
}
JobRegistry.getInstance().addJob(jobConfiguration.getJobName(), this);
}
/**
* fsCleanup: Cleanup intermediate files and reducer files directory for a job
*
* @param jobConf: job's configuration
*/
public void fsCleanup(JobConfiguration jobConf) {
Path path = Paths.get(jobConf.getJobDir() + "-intermediate");
if (Files.exists(path)) {
File dir = new File(jobConf.getJobDir() + "-intermediate");
deleteDir(dir);
}
path = Paths.get(jobConf.getJobDir() + "-reducerfiles");
if (Files.exists(path)) {
File dir = new File(jobConf.getJobDir() + "-reducerfiles");
deleteDir(dir);
}
}
/**
* 创建触发器
*
* @param cronExpression
* @return
*/
private CronTrigger createTrigger(final String cronExpression) {
CronScheduleBuilder cronScheduleBuilder = CronScheduleBuilder.cronSchedule(cronExpression);
return TriggerBuilder.newTrigger()
.withIdentity(jobConfiguration.getJobName())
.withSchedule(cronScheduleBuilder)
.build();
}
/** 停止作业. */
public void stopJob() {
try {
JobRegistry.getInstance().getJobInstance(jobConfiguration.getJobName()).stop();
scheduler.pauseAll();
} catch (final SchedulerException ex) {
throw new JobException(ex);
}
}
/** 重新调度作业. */
public void rescheduleJob(final String cronExpression) {
try {
scheduler.rescheduleJob(
TriggerKey.triggerKey(
Joiner.on("_").join(jobConfiguration.getJobName(), CRON_TRIGGER_INDENTITY_SUFFIX)),
createTrigger(cronExpression));
} catch (final SchedulerException ex) {
throw new JobException(ex);
}
}
/**
* performReduce: Add the reduceTask to the list of current running reduceTasks, Allocate a new
* ReduceRunner and an executor service and start the reduceTask
*
* @param reduceTask: reduce task to be added
*/
public void performReduce(ReduceTask reduceTask) {
JobConfiguration jobConf = reduceTask.getJobConf();
Path path = Paths.get(jobConf.getJobDir() + "/" + jobConf.getOutputDirectoryPath());
// create an output directory for the final output if one doesn't exist
if (!Files.exists(path)) {
File dir = new File(jobConf.getJobDir() + "/" + jobConf.getOutputDirectoryPath());
dir.mkdir();
}
reduceLock.lock();
reduceTasks.add(reduceTask);
reduceLock.unlock();
ReduceRunner runner = new ReduceRunner(reduceTask);
// add the DataNode as a listener for this task
runner.addListener(this);
ExecutorService executor = Executors.newSingleThreadExecutor();
addExecutorToMap(reduceTask.getJobConf().getJobID(), executor);
executor.execute(runner);
}
/** 恢复手工停止的作业. */
public void resumeManualStopedJob() {
try {
if (scheduler.isShutdown()) {
return;
}
JobRegistry.getInstance().getJobInstance(jobConfiguration.getJobName()).resume();
scheduler.resumeAll();
} catch (final SchedulerException ex) {
throw new JobException(ex);
}
serverService.clearJobStopedStatus();
}
private CronTrigger createTrigger(final String cronExpression) {
CronScheduleBuilder cronScheduleBuilder = CronScheduleBuilder.cronSchedule(cronExpression);
if (configService.isMisfire()) {
cronScheduleBuilder = cronScheduleBuilder.withMisfireHandlingInstructionFireAndProceed();
} else {
cronScheduleBuilder = cronScheduleBuilder.withMisfireHandlingInstructionDoNothing();
}
return TriggerBuilder.newTrigger()
.withIdentity(
Joiner.on("_").join(jobConfiguration.getJobName(), CRON_TRIGGER_INDENTITY_SUFFIX))
.withSchedule(cronScheduleBuilder)
.build();
}
/**
* 恢复因服务器崩溃而停止的作业.
*
* <p>不会恢复手工设置停止运行的作业.
*/
public void resumeCrashedJob() {
serverService.persistServerOnline();
executionService.clearRunningInfo(shardingService.getLocalHostShardingItems());
if (serverService.isJobStopedManually()) {
return;
}
JobRegistry.getInstance().getJobInstance(jobConfiguration.getJobName()).resume();
try {
scheduler.resumeAll();
} catch (final SchedulerException ex) {
throw new JobException(ex);
}
}
/**
* notifyOfMapperThreadCompletion: We come in here when a mapper task is successfully completed.
* At this point the mapper has generated intermediate files. In this function we run the
* partition function on the keys and decide which reducers we need to send the intermediate files
* to.
*
* <p>The sending of intermediate files as soon as a map task completes is better than sending all
* at the end, especially in situations where the network bandwidth can be a bottleneck.
*
* @param mapRunner
*/
@Override
public synchronized void notifyOfMapperThreadCompletion(MapRunner mapRunner) {
MapTask finished = mapRunner.getMapTask();
JobConfiguration jobConf = finished.getJobConf();
final int chunkID = finished.getChunkID();
File folder = new File(jobConf.getJobDir() + "-intermediate/");
File[] intermediateFiles =
folder.listFiles(
new FilenameFilter() {
@Override
public boolean accept(File dir, String name) {
return name.endsWith(Integer.toString(chunkID));
}
});
/** Iterate through all the intermediate files created for a particular chunkID */
for (File file : intermediateFiles) {
try {
/** The intermediate filename is key-chunkID. Extract the key from the filename */
String[] arr = file.getName().split("-");
String key = "";
for (int i = 0; i < arr.length - 1; i++) {
key += arr[i];
}
// create new partitioner object
Partitioner partitioner = new Partitioner(jobConf.getNumberOfReducers(), key);
/**
* run a partition function which returns a reducer to which this intermediate file should
* be sent to
*/
int dataNodeIndex = partitioner.partition();
/**
* get the list of all the data nodes. Sort them. Use the dataNodeIndex returned by the
* partition function to get the actual reducer node
*/
communicator.acquireDataNodesLock();
List<String> allDataNodes = communicator.getAllDataNodes();
Collections.sort(allDataNodes);
String reducerNode = allDataNodes.get(dataNodeIndex);
/**
* Get the communicator object of the reducer node, Read the intermediate file into the
* memory and call receiveIntermediateFile() on the communicator of the reducer node.
*/
String intermediateFilePath =
jobConf.getJobDir() + "-intermediate/" + key.toString() + "-" + chunkID;
Registry registry;
registry = LocateRegistry.getRegistry(reducerNode, communicator.getREGISTRY_PORT());
CommunicatorInterface communicator =
(CommunicatorInterface) registry.lookup("communicator_" + reducerNode);
communicator.receiveIntermediateFile(
jobConf, Files.readAllBytes(Paths.get(intermediateFilePath)), file.getName());
/**
* At the end of the task completion we send to master a list of reducer nodes to which
* intermediate files were sent to. Hence store this reducer node to the list
*/
finished.addReducer(reducerNode);
} catch (RemoteException e) {
e.printStackTrace();
} catch (IOException e) {
e.printStackTrace();
} catch (NotBoundException e) {
e.printStackTrace();
} catch (IndexOutOfBoundsException e) {
System.out.println("This job is about to be terminated");
}
communicator.releaseDataNodesLock();
}
// notify the master that map task has successfully completed
notifyMasterOfMapTaskCompletion(finished);
// remove this task from local data structure
mapLock.lock();
mapTasks.remove(finished);
mapLock.unlock();
}
/**
* 初始化作业.
*
* @throws SchedulerException
*/
public void init() throws SchedulerException {
jobDetail = createJobDetail();
scheduler = initializeScheduler(jobDetail.getKey().toString());
scheduleJob(createTrigger(jobConfiguration.getCron()));
}
/**
* 重新调度作业.
*
* @throws SchedulerException
*/
public void rescheduleJob(final String cronExpression) throws SchedulerException {
scheduler.rescheduleJob(
TriggerKey.triggerKey(jobConfiguration.getJobName()), createTrigger(cronExpression));
}
