Assert.assertTrue(MAX_MAP_TASK_ATTEMPTS > 0);
TTClient tClient = null;
TTClient[] ttClients = null;
JobInfo jInfo = remoteJTClient.getJobInfo(rJob.getID());
// Assert if jobInfo is null
Assert.assertNotNull(jInfo.getStatus().getRunState());
// Wait for the job to start running.
while (jInfo.getStatus().getRunState() != JobStatus.RUNNING) {
try {
Thread.sleep(10000);
} catch (InterruptedException e) {
}
;
jInfo = remoteJTClient.getJobInfo(rJob.getID());
}
// Temporarily store the jobid to use it later for comparision.
JobID jobidStore = rJob.getID();
jobidStore = JobID.downgrade(jobidStore);
LOG.info("job id is :" + jobidStore.toString());
TaskInfo[] taskInfos = null;
// After making sure that the job is running,
// the test execution has to make sure that
// at least one task has started running before continuing.
boolean runningCount = false;
int count = 0;
do {
taskInfos = cluster.getJTClient().getProxy().getTaskInfo(rJob.getID());
runningCount = false;
for (TaskInfo taskInfo : taskInfos) {
TaskStatus[] taskStatuses = taskInfo.getTaskStatus();
if (taskStatuses.length > 0) {
LOG.info("taskStatuses[0].getRunState() is :"
+ taskStatuses[0].getRunState());
if (taskStatuses[0].getRunState() == TaskStatus.State.RUNNING) {
runningCount = true;
break;
} else {
LOG.info("Sleeping 5 seconds");
Thread.sleep(5000);
}
}
}
count++;
// If the count goes beyond a point, then break; This is to avoid
// infinite loop under unforeseen circumstances. Testcase will anyway
// fail later.
if (count > 10) {
Assert.fail("Since the sleep count has reached beyond a point"
+ "failing at this point");
}
} while (!runningCount);
// This whole module is about getting the task Attempt id
// of one task and killing it MAX_MAP_TASK_ATTEMPTS times,
// whenever it re-attempts to run.
String taskIdKilled = null;
for (int i = 0; i < MAX_MAP_TASK_ATTEMPTS; i++) {
taskInfos = cluster.getJTClient().getProxy().getTaskInfo(rJob.getID());
for (TaskInfo taskInfo : taskInfos) {
TaskAttemptID taskAttemptID;
if (!taskInfo.isSetupOrCleanup()) {
// This is the task which is going to be killed continously in
// all its task attempts.The first task is getting picked up.
TaskID taskid = TaskID.downgrade(taskInfo.getTaskID());
LOG.info("taskid is :" + taskid);
if (i == 0) {
taskIdKilled = taskid.toString();
taskAttemptID = new TaskAttemptID(taskid, i);
LOG.info("taskAttemptid going to be killed is : " + taskAttemptID);
(new JobClient.NetworkedJob(jInfo.getStatus(),jobClient.cluster)).killTask(
taskAttemptID, true);
checkTaskCompletionEvent(taskAttemptID, jInfo);
break;
} else {
if (taskIdKilled.equals(taskid.toString())) {
taskAttemptID = new TaskAttemptID(taskid, i);
LOG
.info("taskAttemptid going to be killed is : "
+ taskAttemptID);
(new JobClient.NetworkedJob(jInfo.getStatus(),jobClient.cluster)).killTask(
taskAttemptID, true);
checkTaskCompletionEvent(taskAttemptID, jInfo);
break;
}
}
}
}
}
// Making sure that the job is complete.
while (jInfo != null && !jInfo.getStatus().isJobComplete()) {
Thread.sleep(10000);
jInfo = remoteJTClient.getJobInfo(rJob.getID());
}
// Making sure that the correct jobstatus is got from all the jobs