// counter for job Loop
countLoop++;
TTClient tClient = null;
JobInfo jInfo = wovenClient.getJobInfo(rJob.getID());
LOG.info("jInfo is :" + jInfo);
// Assert if jobInfo is null
Assert.assertNotNull("jobInfo is null", jInfo);
// Wait for the job to start running.
count = 0;
while (jInfo.getStatus().getRunState() != JobStatus.RUNNING) {
UtilsForTests.waitFor(10000);
count++;
jInfo = wovenClient.getJobInfo(rJob.getID());
// 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("job has not reached running state for more than"
+ "100 seconds. Failing at this point");
}
}
LOG.info("job id is :" + rJob.getID().toString());
TaskInfo[] taskInfos =
cluster.getJTClient().getProxy().getTaskInfo(rJob.getID());
boolean distCacheFileIsFound;
for (TaskInfo taskInfo : taskInfos) {
distCacheFileIsFound = false;
String[] taskTrackers = taskInfo.getTaskTrackers();
for (String taskTracker : taskTrackers) {
// Formatting tasktracker to get just its FQDN
taskTracker = UtilsForTests.getFQDNofTT(taskTracker);
LOG.info("taskTracker is :" + taskTracker);
// The tasktrackerFound variable is initialized
taskTrackerFound = false;
// This will be entered from the second job onwards
if (countLoop > 1) {
if (taskTracker != null) {
continueLoop = taskTrackerCollection.contains(taskTracker);
}
if (continueLoop) {
taskTrackerFound = true;
}
}
// Collecting the tasktrackers
if (taskTracker != null)
taskTrackerCollection.add(taskTracker);
// we have loopped through two times to look for task
// getting submitted on same tasktrackers.The same tasktracker
// for subsequent jobs was not hit maybe because of many number
// of tasktrackers. So, testcase has to stop here.
if (countLoop > 1) {
continueLoop = false;
}
tClient = cluster.getTTClient(taskTracker);
// tClient maybe null because the task is already dead. Ex: setup
if (tClient == null) {
continue;
}
String[] localDirs = tClient.getMapredLocalDirs();
int distributedFileCount = 0;
// Go to every single path
for (String localDir : localDirs) {
// Public Distributed cache will always be stored under
// mapre.local.dir/tasktracker/archive
localDir =
localDir
+ Path.SEPARATOR
+ TaskTracker.getPublicDistributedCacheDir();
LOG.info("localDir is : " + localDir);
// Get file status of all the directories
// and files under that path.
FileStatus[] fileStatuses =
tClient.listStatus(localDir, true, true);
for (FileStatus fileStatus : fileStatuses) {
Path path = fileStatus.getPath();
LOG.info("path is :" + path.toString());
// Checking if the received path ends with
// the distributed filename
distCacheFileIsFound =
(path.toString()).endsWith(distributedFileName);
// If file is found, check for its permission.
// Since the file is found break out of loop
if (distCacheFileIsFound) {
LOG.info("PATH found is :" + path.toString());
distributedFileCount++;
String filename = path.getName();
FsPermission fsPerm = fileStatus.getPermission();
Assert.assertTrue("File Permission is not 777", fsPerm
.equals(new FsPermission("777")));
}
}
}
LOG.debug("The distributed FileCount is :" + distributedFileCount);
LOG.debug("The taskTrackerFound is :" + taskTrackerFound);
// If distributed cache is modified in dfs
// between two job runs, it can be present more than once
// in any of the task tracker, in which job ran.
if (distributedFileCount != 2 && taskTrackerFound) {
Assert.fail("The distributed cache file has to be two. "
+ "But found was " + distributedFileCount);
} else if (distributedFileCount > 1 && !taskTrackerFound) {
Assert.fail("The distributed cache file cannot more than one."
+ " But found was " + distributedFileCount);
} else if (distributedFileCount < 1)
Assert.fail("The distributed cache file is less than one. "
+ "But found was " + distributedFileCount);
if (!distCacheFileIsFound) {
Assert.assertEquals(
"The distributed cache file does not exist",
distCacheFileIsFound, false);
}
}
}
// Allow the job to continue through MR control job.
for (TaskInfo taskInfoRemaining : taskInfos) {
FinishTaskControlAction action =
new FinishTaskControlAction(TaskID.downgrade(taskInfoRemaining
.getTaskID()));
Collection<TTClient> tts = cluster.getTTClients();
for (TTClient cli : tts) {
cli.getProxy().sendAction(action);
}
}
// Killing the job because all the verification needed
// for this testcase is completed.
rJob.killJob();
// Waiting for 3 seconds for cleanup to start
Thread.sleep(3000);
// Getting the last cleanup task's tasktracker also, as
// distributed cache gets uploaded even during cleanup.
TaskInfo[] myTaskInfos = wovenClient.getTaskInfo(rJob.getID());
if (myTaskInfos != null) {
for (TaskInfo info : myTaskInfos) {
if (info.isSetupOrCleanup()) {
String[] taskTrackers = info.getTaskTrackers();
for (String taskTracker : taskTrackers) {
// Formatting tasktracker to get just its FQDN
taskTracker = UtilsForTests.getFQDNofTT(taskTracker);
LOG.info("taskTracker is :" + taskTracker);
// Collecting the tasktrackers
if (taskTracker != null)
taskTrackerCollection.add(taskTracker);
}
}
}
}
// Making sure that the job is complete.
while (jInfo != null && !jInfo.getStatus().isJobComplete()) {
Thread.sleep(10000);
jInfo = wovenClient.getJobInfo(rJob.getID());
}
} while (continueLoop);