use of org.apache.hadoop.mapreduce.v2.api.records.Counter in project hadoop by apache.
the class TypeConverter method toYarn.
public static Counters toYarn(org.apache.hadoop.mapred.Counters counters) {
if (counters == null) {
return null;
}
Counters yCntrs = recordFactory.newRecordInstance(Counters.class);
yCntrs.addAllCounterGroups(new HashMap<String, CounterGroup>());
for (org.apache.hadoop.mapred.Counters.Group grp : counters) {
CounterGroup yGrp = recordFactory.newRecordInstance(CounterGroup.class);
yGrp.setName(grp.getName());
yGrp.setDisplayName(grp.getDisplayName());
yGrp.addAllCounters(new HashMap<String, Counter>());
for (org.apache.hadoop.mapred.Counters.Counter cntr : grp) {
Counter yCntr = recordFactory.newRecordInstance(Counter.class);
yCntr.setName(cntr.getName());
yCntr.setDisplayName(cntr.getDisplayName());
yCntr.setValue(cntr.getValue());
yGrp.setCounter(yCntr.getName(), yCntr);
}
yCntrs.setCounterGroup(yGrp.getName(), yGrp);
}
return yCntrs;
}
use of org.apache.hadoop.mapreduce.v2.api.records.Counter in project hadoop by apache.
the class SingleCounterBlock method populateMembers.
private void populateMembers(AppContext ctx) {
JobId jobID = null;
TaskId taskID = null;
String tid = $(TASK_ID);
if ($(TITLE).contains("MAPS")) {
counterType = TaskType.MAP;
} else if ($(TITLE).contains("REDUCES")) {
counterType = TaskType.REDUCE;
} else {
counterType = null;
}
if (!tid.isEmpty()) {
taskID = MRApps.toTaskID(tid);
jobID = taskID.getJobId();
} else {
String jid = $(JOB_ID);
if (!jid.isEmpty()) {
jobID = MRApps.toJobID(jid);
}
}
if (jobID == null) {
return;
}
job = ctx.getJob(jobID);
if (job == null) {
return;
}
if (taskID != null) {
task = job.getTask(taskID);
if (task == null) {
return;
}
for (Map.Entry<TaskAttemptId, TaskAttempt> entry : task.getAttempts().entrySet()) {
long value = 0;
Counters counters = entry.getValue().getCounters();
CounterGroup group = (counters != null) ? counters.getGroup($(COUNTER_GROUP)) : null;
if (group != null) {
Counter c = group.findCounter($(COUNTER_NAME));
if (c != null) {
value = c.getValue();
}
}
values.put(MRApps.toString(entry.getKey()), value);
}
return;
}
// Get all types of counters
Map<TaskId, Task> tasks = job.getTasks();
for (Map.Entry<TaskId, Task> entry : tasks.entrySet()) {
long value = 0;
Counters counters = entry.getValue().getCounters();
CounterGroup group = (counters != null) ? counters.getGroup($(COUNTER_GROUP)) : null;
if (group != null) {
Counter c = group.findCounter($(COUNTER_NAME));
if (c != null) {
value = c.getValue();
}
}
if (counterType == null || counterType == entry.getValue().getType()) {
values.put(MRApps.toString(entry.getKey()), value);
}
}
}
use of org.apache.hadoop.mapreduce.v2.api.records.Counter in project hadoop by apache.
the class TestTaskAttemptListenerImpl method testCheckpointIDTracking.
@Test
public void testCheckpointIDTracking() throws IOException, InterruptedException {
SystemClock clock = SystemClock.getInstance();
org.apache.hadoop.mapreduce.v2.app.job.Task mockTask = mock(org.apache.hadoop.mapreduce.v2.app.job.Task.class);
when(mockTask.canCommit(any(TaskAttemptId.class))).thenReturn(true);
Job mockJob = mock(Job.class);
when(mockJob.getTask(any(TaskId.class))).thenReturn(mockTask);
Dispatcher dispatcher = mock(Dispatcher.class);
@SuppressWarnings("unchecked") EventHandler<Event> ea = mock(EventHandler.class);
when(dispatcher.getEventHandler()).thenReturn(ea);
RMHeartbeatHandler rmHeartbeatHandler = mock(RMHeartbeatHandler.class);
AppContext appCtx = mock(AppContext.class);
when(appCtx.getJob(any(JobId.class))).thenReturn(mockJob);
when(appCtx.getClock()).thenReturn(clock);
when(appCtx.getEventHandler()).thenReturn(ea);
JobTokenSecretManager secret = mock(JobTokenSecretManager.class);
final TaskHeartbeatHandler hbHandler = mock(TaskHeartbeatHandler.class);
when(appCtx.getEventHandler()).thenReturn(ea);
CheckpointAMPreemptionPolicy policy = new CheckpointAMPreemptionPolicy();
policy.init(appCtx);
TaskAttemptListenerImpl listener = new MockTaskAttemptListenerImpl(appCtx, secret, rmHeartbeatHandler, policy) {
@Override
protected void registerHeartbeatHandler(Configuration conf) {
taskHeartbeatHandler = hbHandler;
}
};
Configuration conf = new Configuration();
conf.setBoolean(MRJobConfig.TASK_PREEMPTION, true);
//conf.setBoolean("preemption.reduce", true);
listener.init(conf);
listener.start();
TaskAttemptID tid = new TaskAttemptID("12345", 1, TaskType.REDUCE, 1, 0);
List<Path> partialOut = new ArrayList<Path>();
partialOut.add(new Path("/prev1"));
partialOut.add(new Path("/prev2"));
Counters counters = mock(Counters.class);
final long CBYTES = 64L * 1024 * 1024;
final long CTIME = 4344L;
final Path CLOC = new Path("/test/1");
Counter cbytes = mock(Counter.class);
when(cbytes.getValue()).thenReturn(CBYTES);
Counter ctime = mock(Counter.class);
when(ctime.getValue()).thenReturn(CTIME);
when(counters.findCounter(eq(EnumCounter.CHECKPOINT_BYTES))).thenReturn(cbytes);
when(counters.findCounter(eq(EnumCounter.CHECKPOINT_MS))).thenReturn(ctime);
// propagating a taskstatus that contains a checkpoint id
TaskCheckpointID incid = new TaskCheckpointID(new FSCheckpointID(CLOC), partialOut, counters);
listener.setCheckpointID(org.apache.hadoop.mapred.TaskID.downgrade(tid.getTaskID()), incid);
// and try to get it back
CheckpointID outcid = listener.getCheckpointID(tid.getTaskID());
TaskCheckpointID tcid = (TaskCheckpointID) outcid;
assertEquals(CBYTES, tcid.getCheckpointBytes());
assertEquals(CTIME, tcid.getCheckpointTime());
assertTrue(partialOut.containsAll(tcid.getPartialCommittedOutput()));
assertTrue(tcid.getPartialCommittedOutput().containsAll(partialOut));
//assert it worked
assert outcid == incid;
listener.stop();
}
use of org.apache.hadoop.mapreduce.v2.api.records.Counter in project hadoop by apache.
the class CounterGroupPBImpl method initCounters.
private void initCounters() {
if (this.counters != null) {
return;
}
CounterGroupProtoOrBuilder p = viaProto ? proto : builder;
List<StringCounterMapProto> list = p.getCountersList();
this.counters = new HashMap<String, Counter>();
for (StringCounterMapProto c : list) {
this.counters.put(c.getKey(), convertFromProtoFormat(c.getValue()));
}
}
use of org.apache.hadoop.mapreduce.v2.api.records.Counter in project hadoop by apache.
the class CLI method run.
public int run(String[] argv) throws Exception {
int exitCode = -1;
if (argv.length < 1) {
displayUsage("");
return exitCode;
}
// process arguments
String cmd = argv[0];
String submitJobFile = null;
String jobid = null;
String taskid = null;
String historyFileOrJobId = null;
String historyOutFile = null;
String historyOutFormat = HistoryViewer.HUMAN_FORMAT;
String counterGroupName = null;
String counterName = null;
JobPriority jp = null;
String taskType = null;
String taskState = null;
int fromEvent = 0;
int nEvents = 0;
int jpvalue = 0;
String configOutFile = null;
boolean getStatus = false;
boolean getCounter = false;
boolean killJob = false;
boolean listEvents = false;
boolean viewHistory = false;
boolean viewAllHistory = false;
boolean listJobs = false;
boolean listAllJobs = false;
boolean listActiveTrackers = false;
boolean listBlacklistedTrackers = false;
boolean displayTasks = false;
boolean killTask = false;
boolean failTask = false;
boolean setJobPriority = false;
boolean logs = false;
boolean downloadConfig = false;
if ("-submit".equals(cmd)) {
if (argv.length != 2) {
displayUsage(cmd);
return exitCode;
}
submitJobFile = argv[1];
} else if ("-status".equals(cmd)) {
if (argv.length != 2) {
displayUsage(cmd);
return exitCode;
}
jobid = argv[1];
getStatus = true;
} else if ("-counter".equals(cmd)) {
if (argv.length != 4) {
displayUsage(cmd);
return exitCode;
}
getCounter = true;
jobid = argv[1];
counterGroupName = argv[2];
counterName = argv[3];
} else if ("-kill".equals(cmd)) {
if (argv.length != 2) {
displayUsage(cmd);
return exitCode;
}
jobid = argv[1];
killJob = true;
} else if ("-set-priority".equals(cmd)) {
if (argv.length != 3) {
displayUsage(cmd);
return exitCode;
}
jobid = argv[1];
try {
jp = JobPriority.valueOf(argv[2]);
} catch (IllegalArgumentException iae) {
try {
jpvalue = Integer.parseInt(argv[2]);
} catch (NumberFormatException ne) {
LOG.info(ne);
displayUsage(cmd);
return exitCode;
}
}
setJobPriority = true;
} else if ("-events".equals(cmd)) {
if (argv.length != 4) {
displayUsage(cmd);
return exitCode;
}
jobid = argv[1];
fromEvent = Integer.parseInt(argv[2]);
nEvents = Integer.parseInt(argv[3]);
listEvents = true;
} else if ("-history".equals(cmd)) {
viewHistory = true;
if (argv.length < 2 || argv.length > 7) {
displayUsage(cmd);
return exitCode;
}
// Some arguments are optional while others are not, and some require
// second arguments. Due to this, the indexing can vary depending on
// what's specified and what's left out, as summarized in the below table:
// [all] <jobHistoryFile|jobId> [-outfile <file>] [-format <human|json>]
// 1 2 3 4 5 6
// 1 2 3 4
// 1 2 3 4
// 1 2
// 1 2 3 4 5
// 1 2 3
// 1 2 3
// 1
// "all" is optional, but comes first if specified
int index = 1;
if ("all".equals(argv[index])) {
index++;
viewAllHistory = true;
if (argv.length == 2) {
displayUsage(cmd);
return exitCode;
}
}
// Get the job history file or job id argument
historyFileOrJobId = argv[index++];
// "-outfile" is optional, but if specified requires a second argument
if (argv.length > index + 1 && "-outfile".equals(argv[index])) {
index++;
historyOutFile = argv[index++];
}
// "-format" is optional, but if specified required a second argument
if (argv.length > index + 1 && "-format".equals(argv[index])) {
index++;
historyOutFormat = argv[index++];
}
// Check for any extra arguments that don't belong here
if (argv.length > index) {
displayUsage(cmd);
return exitCode;
}
} else if ("-list".equals(cmd)) {
if (argv.length != 1 && !(argv.length == 2 && "all".equals(argv[1]))) {
displayUsage(cmd);
return exitCode;
}
if (argv.length == 2 && "all".equals(argv[1])) {
listAllJobs = true;
} else {
listJobs = true;
}
} else if ("-kill-task".equals(cmd)) {
if (argv.length != 2) {
displayUsage(cmd);
return exitCode;
}
killTask = true;
taskid = argv[1];
} else if ("-fail-task".equals(cmd)) {
if (argv.length != 2) {
displayUsage(cmd);
return exitCode;
}
failTask = true;
taskid = argv[1];
} else if ("-list-active-trackers".equals(cmd)) {
if (argv.length != 1) {
displayUsage(cmd);
return exitCode;
}
listActiveTrackers = true;
} else if ("-list-blacklisted-trackers".equals(cmd)) {
if (argv.length != 1) {
displayUsage(cmd);
return exitCode;
}
listBlacklistedTrackers = true;
} else if ("-list-attempt-ids".equals(cmd)) {
if (argv.length != 4) {
displayUsage(cmd);
return exitCode;
}
jobid = argv[1];
taskType = argv[2];
taskState = argv[3];
displayTasks = true;
if (!taskTypes.contains(org.apache.hadoop.util.StringUtils.toUpperCase(taskType))) {
System.out.println("Error: Invalid task-type: " + taskType);
displayUsage(cmd);
return exitCode;
}
if (!taskStates.contains(org.apache.hadoop.util.StringUtils.toLowerCase(taskState))) {
System.out.println("Error: Invalid task-state: " + taskState);
displayUsage(cmd);
return exitCode;
}
} else if ("-logs".equals(cmd)) {
if (argv.length == 2 || argv.length == 3) {
logs = true;
jobid = argv[1];
if (argv.length == 3) {
taskid = argv[2];
} else {
taskid = null;
}
} else {
displayUsage(cmd);
return exitCode;
}
} else if ("-config".equals(cmd)) {
downloadConfig = true;
if (argv.length != 3) {
displayUsage(cmd);
return exitCode;
}
jobid = argv[1];
configOutFile = argv[2];
} else {
displayUsage(cmd);
return exitCode;
}
// initialize cluster
cluster = createCluster();
// Submit the request
try {
if (submitJobFile != null) {
Job job = Job.getInstance(new JobConf(submitJobFile));
job.submit();
System.out.println("Created job " + job.getJobID());
exitCode = 0;
} else if (getStatus) {
Job job = getJob(JobID.forName(jobid));
if (job == null) {
System.out.println("Could not find job " + jobid);
} else {
Counters counters = job.getCounters();
System.out.println();
System.out.println(job);
if (counters != null) {
System.out.println(counters);
} else {
System.out.println("Counters not available. Job is retired.");
}
exitCode = 0;
}
} else if (getCounter) {
Job job = getJob(JobID.forName(jobid));
if (job == null) {
System.out.println("Could not find job " + jobid);
} else {
Counters counters = job.getCounters();
if (counters == null) {
System.out.println("Counters not available for retired job " + jobid);
exitCode = -1;
} else {
System.out.println(getCounter(counters, counterGroupName, counterName));
exitCode = 0;
}
}
} else if (killJob) {
Job job = getJob(JobID.forName(jobid));
if (job == null) {
System.out.println("Could not find job " + jobid);
} else {
JobStatus jobStatus = job.getStatus();
if (jobStatus.getState() == JobStatus.State.FAILED) {
System.out.println("Could not mark the job " + jobid + " as killed, as it has already failed.");
exitCode = -1;
} else if (jobStatus.getState() == JobStatus.State.KILLED) {
System.out.println("The job " + jobid + " has already been killed.");
exitCode = -1;
} else if (jobStatus.getState() == JobStatus.State.SUCCEEDED) {
System.out.println("Could not kill the job " + jobid + ", as it has already succeeded.");
exitCode = -1;
} else {
job.killJob();
System.out.println("Killed job " + jobid);
exitCode = 0;
}
}
} else if (setJobPriority) {
Job job = getJob(JobID.forName(jobid));
if (job == null) {
System.out.println("Could not find job " + jobid);
} else {
if (jp != null) {
job.setPriority(jp);
} else {
job.setPriorityAsInteger(jpvalue);
}
System.out.println("Changed job priority.");
exitCode = 0;
}
} else if (viewHistory) {
// it's a Job ID
if (historyFileOrJobId.endsWith(".jhist")) {
viewHistory(historyFileOrJobId, viewAllHistory, historyOutFile, historyOutFormat);
exitCode = 0;
} else {
Job job = getJob(JobID.forName(historyFileOrJobId));
if (job == null) {
System.out.println("Could not find job " + jobid);
} else {
String historyUrl = job.getHistoryUrl();
if (historyUrl == null || historyUrl.isEmpty()) {
System.out.println("History file for job " + historyFileOrJobId + " is currently unavailable.");
} else {
viewHistory(historyUrl, viewAllHistory, historyOutFile, historyOutFormat);
exitCode = 0;
}
}
}
} else if (listEvents) {
Job job = getJob(JobID.forName(jobid));
if (job == null) {
System.out.println("Could not find job " + jobid);
} else {
listEvents(job, fromEvent, nEvents);
exitCode = 0;
}
} else if (listJobs) {
listJobs(cluster);
exitCode = 0;
} else if (listAllJobs) {
listAllJobs(cluster);
exitCode = 0;
} else if (listActiveTrackers) {
listActiveTrackers(cluster);
exitCode = 0;
} else if (listBlacklistedTrackers) {
listBlacklistedTrackers(cluster);
exitCode = 0;
} else if (displayTasks) {
Job job = getJob(JobID.forName(jobid));
if (job == null) {
System.out.println("Could not find job " + jobid);
} else {
displayTasks(getJob(JobID.forName(jobid)), taskType, taskState);
exitCode = 0;
}
} else if (killTask) {
TaskAttemptID taskID = TaskAttemptID.forName(taskid);
Job job = getJob(taskID.getJobID());
if (job == null) {
System.out.println("Could not find job " + jobid);
} else if (job.killTask(taskID, false)) {
System.out.println("Killed task " + taskid);
exitCode = 0;
} else {
System.out.println("Could not kill task " + taskid);
exitCode = -1;
}
} else if (failTask) {
TaskAttemptID taskID = TaskAttemptID.forName(taskid);
Job job = getJob(taskID.getJobID());
if (job == null) {
System.out.println("Could not find job " + jobid);
} else if (job.killTask(taskID, true)) {
System.out.println("Killed task " + taskID + " by failing it");
exitCode = 0;
} else {
System.out.println("Could not fail task " + taskid);
exitCode = -1;
}
} else if (logs) {
JobID jobID = JobID.forName(jobid);
if (getJob(jobID) == null) {
System.out.println("Could not find job " + jobid);
} else {
try {
TaskAttemptID taskAttemptID = TaskAttemptID.forName(taskid);
LogParams logParams = cluster.getLogParams(jobID, taskAttemptID);
LogCLIHelpers logDumper = new LogCLIHelpers();
logDumper.setConf(getConf());
exitCode = logDumper.dumpAContainersLogs(logParams.getApplicationId(), logParams.getContainerId(), logParams.getNodeId(), logParams.getOwner());
} catch (IOException e) {
if (e instanceof RemoteException) {
throw e;
}
System.out.println(e.getMessage());
}
}
} else if (downloadConfig) {
Job job = getJob(JobID.forName(jobid));
if (job == null) {
System.out.println("Could not find job " + jobid);
} else {
String jobFile = job.getJobFile();
if (jobFile == null || jobFile.isEmpty()) {
System.out.println("Config file for job " + jobFile + " could not be found.");
} else {
Path configPath = new Path(jobFile);
FileSystem fs = FileSystem.get(getConf());
fs.copyToLocalFile(configPath, new Path(configOutFile));
exitCode = 0;
}
}
}
} catch (RemoteException re) {
IOException unwrappedException = re.unwrapRemoteException();
if (unwrappedException instanceof AccessControlException) {
System.out.println(unwrappedException.getMessage());
} else {
throw re;
}
} finally {
cluster.close();
}
return exitCode;
}
Aggregations