Revert "MAPREDUCE-6566. Add retry support to mapreduce CLI tool. Contributed by Varun Vasudev"
This reverts commit ba02e8f6d1
.
This commit is contained in:
parent
7b3f027ed4
commit
cab5d86738
|
@ -86,8 +86,6 @@ Release 2.7.2 - UNRELEASED
|
||||||
MAPREDUCE-6451. DistCp has incorrect chunkFilePath for multiple jobs when
|
MAPREDUCE-6451. DistCp has incorrect chunkFilePath for multiple jobs when
|
||||||
strategy is dynamic (Kuhu Shukla via kihwal)
|
strategy is dynamic (Kuhu Shukla via kihwal)
|
||||||
|
|
||||||
MAPREDUCE-6566. Add retry support to mapreduce CLI tool. (Varun Vasudev via xgong)
|
|
||||||
|
|
||||||
Release 2.7.1 - 2015-07-06
|
Release 2.7.1 - 2015-07-06
|
||||||
|
|
||||||
INCOMPATIBLE CHANGES
|
INCOMPATIBLE CHANGES
|
||||||
|
|
|
@ -26,7 +26,6 @@ import java.util.HashSet;
|
||||||
import java.util.List;
|
import java.util.List;
|
||||||
import java.util.Set;
|
import java.util.Set;
|
||||||
|
|
||||||
import com.google.common.annotations.VisibleForTesting;
|
|
||||||
import org.apache.commons.lang.StringUtils;
|
import org.apache.commons.lang.StringUtils;
|
||||||
import org.apache.commons.logging.Log;
|
import org.apache.commons.logging.Log;
|
||||||
import org.apache.commons.logging.LogFactory;
|
import org.apache.commons.logging.LogFactory;
|
||||||
|
@ -44,7 +43,6 @@ import org.apache.hadoop.mapreduce.Job;
|
||||||
import org.apache.hadoop.mapreduce.JobID;
|
import org.apache.hadoop.mapreduce.JobID;
|
||||||
import org.apache.hadoop.mapreduce.JobPriority;
|
import org.apache.hadoop.mapreduce.JobPriority;
|
||||||
import org.apache.hadoop.mapreduce.JobStatus;
|
import org.apache.hadoop.mapreduce.JobStatus;
|
||||||
import org.apache.hadoop.mapreduce.MRJobConfig;
|
|
||||||
import org.apache.hadoop.mapreduce.TaskAttemptID;
|
import org.apache.hadoop.mapreduce.TaskAttemptID;
|
||||||
import org.apache.hadoop.mapreduce.TaskCompletionEvent;
|
import org.apache.hadoop.mapreduce.TaskCompletionEvent;
|
||||||
import org.apache.hadoop.mapreduce.TaskReport;
|
import org.apache.hadoop.mapreduce.TaskReport;
|
||||||
|
@ -265,7 +263,7 @@ public class CLI extends Configured implements Tool {
|
||||||
System.out.println("Created job " + job.getJobID());
|
System.out.println("Created job " + job.getJobID());
|
||||||
exitCode = 0;
|
exitCode = 0;
|
||||||
} else if (getStatus) {
|
} else if (getStatus) {
|
||||||
Job job = getJob(JobID.forName(jobid));
|
Job job = cluster.getJob(JobID.forName(jobid));
|
||||||
if (job == null) {
|
if (job == null) {
|
||||||
System.out.println("Could not find job " + jobid);
|
System.out.println("Could not find job " + jobid);
|
||||||
} else {
|
} else {
|
||||||
|
@ -280,7 +278,7 @@ public class CLI extends Configured implements Tool {
|
||||||
exitCode = 0;
|
exitCode = 0;
|
||||||
}
|
}
|
||||||
} else if (getCounter) {
|
} else if (getCounter) {
|
||||||
Job job = getJob(JobID.forName(jobid));
|
Job job = cluster.getJob(JobID.forName(jobid));
|
||||||
if (job == null) {
|
if (job == null) {
|
||||||
System.out.println("Could not find job " + jobid);
|
System.out.println("Could not find job " + jobid);
|
||||||
} else {
|
} else {
|
||||||
|
@ -296,7 +294,7 @@ public class CLI extends Configured implements Tool {
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
} else if (killJob) {
|
} else if (killJob) {
|
||||||
Job job = getJob(JobID.forName(jobid));
|
Job job = cluster.getJob(JobID.forName(jobid));
|
||||||
if (job == null) {
|
if (job == null) {
|
||||||
System.out.println("Could not find job " + jobid);
|
System.out.println("Could not find job " + jobid);
|
||||||
} else {
|
} else {
|
||||||
|
@ -320,7 +318,7 @@ public class CLI extends Configured implements Tool {
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
} else if (setJobPriority) {
|
} else if (setJobPriority) {
|
||||||
Job job = getJob(JobID.forName(jobid));
|
Job job = cluster.getJob(JobID.forName(jobid));
|
||||||
if (job == null) {
|
if (job == null) {
|
||||||
System.out.println("Could not find job " + jobid);
|
System.out.println("Could not find job " + jobid);
|
||||||
} else {
|
} else {
|
||||||
|
@ -332,7 +330,7 @@ public class CLI extends Configured implements Tool {
|
||||||
viewHistory(historyFile, viewAllHistory);
|
viewHistory(historyFile, viewAllHistory);
|
||||||
exitCode = 0;
|
exitCode = 0;
|
||||||
} else if (listEvents) {
|
} else if (listEvents) {
|
||||||
listEvents(getJob(JobID.forName(jobid)), fromEvent, nEvents);
|
listEvents(cluster.getJob(JobID.forName(jobid)), fromEvent, nEvents);
|
||||||
exitCode = 0;
|
exitCode = 0;
|
||||||
} else if (listJobs) {
|
} else if (listJobs) {
|
||||||
listJobs(cluster);
|
listJobs(cluster);
|
||||||
|
@ -347,11 +345,11 @@ public class CLI extends Configured implements Tool {
|
||||||
listBlacklistedTrackers(cluster);
|
listBlacklistedTrackers(cluster);
|
||||||
exitCode = 0;
|
exitCode = 0;
|
||||||
} else if (displayTasks) {
|
} else if (displayTasks) {
|
||||||
displayTasks(getJob(JobID.forName(jobid)), taskType, taskState);
|
displayTasks(cluster.getJob(JobID.forName(jobid)), taskType, taskState);
|
||||||
exitCode = 0;
|
exitCode = 0;
|
||||||
} else if(killTask) {
|
} else if(killTask) {
|
||||||
TaskAttemptID taskID = TaskAttemptID.forName(taskid);
|
TaskAttemptID taskID = TaskAttemptID.forName(taskid);
|
||||||
Job job = getJob(taskID.getJobID());
|
Job job = cluster.getJob(taskID.getJobID());
|
||||||
if (job == null) {
|
if (job == null) {
|
||||||
System.out.println("Could not find job " + jobid);
|
System.out.println("Could not find job " + jobid);
|
||||||
} else if (job.killTask(taskID, false)) {
|
} else if (job.killTask(taskID, false)) {
|
||||||
|
@ -363,7 +361,7 @@ public class CLI extends Configured implements Tool {
|
||||||
}
|
}
|
||||||
} else if(failTask) {
|
} else if(failTask) {
|
||||||
TaskAttemptID taskID = TaskAttemptID.forName(taskid);
|
TaskAttemptID taskID = TaskAttemptID.forName(taskid);
|
||||||
Job job = getJob(taskID.getJobID());
|
Job job = cluster.getJob(taskID.getJobID());
|
||||||
if (job == null) {
|
if (job == null) {
|
||||||
System.out.println("Could not find job " + jobid);
|
System.out.println("Could not find job " + jobid);
|
||||||
} else if(job.killTask(taskID, true)) {
|
} else if(job.killTask(taskID, true)) {
|
||||||
|
@ -519,29 +517,6 @@ public class CLI extends Configured implements Tool {
|
||||||
return (baseUrl + "/tasklog?plaintext=true&attemptid=" + taskId);
|
return (baseUrl + "/tasklog?plaintext=true&attemptid=" + taskId);
|
||||||
}
|
}
|
||||||
|
|
||||||
@VisibleForTesting
|
|
||||||
Job getJob(JobID jobid) throws IOException, InterruptedException {
|
|
||||||
|
|
||||||
int maxRetry = getConf().getInt(MRJobConfig.MR_CLIENT_JOB_MAX_RETRIES,
|
|
||||||
MRJobConfig.DEFAULT_MR_CLIENT_JOB_MAX_RETRIES);
|
|
||||||
long retryInterval = getConf()
|
|
||||||
.getLong(MRJobConfig.MR_CLIENT_JOB_RETRY_INTERVAL,
|
|
||||||
MRJobConfig.DEFAULT_MR_CLIENT_JOB_RETRY_INTERVAL);
|
|
||||||
Job job = cluster.getJob(jobid);
|
|
||||||
|
|
||||||
for (int i = 0; i < maxRetry; ++i) {
|
|
||||||
if (job != null) {
|
|
||||||
return job;
|
|
||||||
}
|
|
||||||
LOG.info("Could not obtain job info after " + String.valueOf(i + 1)
|
|
||||||
+ " attempt(s). Sleeping for " + String.valueOf(retryInterval / 1000)
|
|
||||||
+ " seconds and retrying.");
|
|
||||||
Thread.sleep(retryInterval);
|
|
||||||
job = cluster.getJob(jobid);
|
|
||||||
}
|
|
||||||
return job;
|
|
||||||
}
|
|
||||||
|
|
||||||
|
|
||||||
/**
|
/**
|
||||||
* Dump a list of currently running jobs
|
* Dump a list of currently running jobs
|
||||||
|
|
|
@ -20,19 +20,14 @@ package org.apache.hadoop.mapreduce.tools;
|
||||||
import static org.junit.Assert.*;
|
import static org.junit.Assert.*;
|
||||||
|
|
||||||
import java.io.IOException;
|
import java.io.IOException;
|
||||||
|
|
||||||
import org.apache.hadoop.conf.Configuration;
|
|
||||||
import org.apache.hadoop.mapreduce.Cluster;
|
import org.apache.hadoop.mapreduce.Cluster;
|
||||||
import org.apache.hadoop.mapreduce.Job;
|
import org.apache.hadoop.mapreduce.Job;
|
||||||
import org.apache.hadoop.mapreduce.JobID;
|
import org.apache.hadoop.mapreduce.JobID;
|
||||||
import org.apache.hadoop.mapreduce.MRJobConfig;
|
|
||||||
import org.apache.hadoop.mapreduce.TaskReport;
|
import org.apache.hadoop.mapreduce.TaskReport;
|
||||||
import org.apache.hadoop.mapreduce.TaskType;
|
import org.apache.hadoop.mapreduce.TaskType;
|
||||||
import org.apache.hadoop.mapreduce.JobPriority;
|
import org.apache.hadoop.mapreduce.JobPriority;
|
||||||
import org.apache.hadoop.mapreduce.JobStatus;
|
import org.apache.hadoop.mapreduce.JobStatus;
|
||||||
import org.apache.hadoop.mapreduce.JobStatus.State;
|
import org.apache.hadoop.mapreduce.JobStatus.State;
|
||||||
import org.apache.hadoop.util.Time;
|
|
||||||
import org.junit.Assert;
|
|
||||||
import org.junit.Test;
|
import org.junit.Test;
|
||||||
import static org.mockito.Mockito.mock;
|
import static org.mockito.Mockito.mock;
|
||||||
import static org.mockito.Mockito.when;
|
import static org.mockito.Mockito.when;
|
||||||
|
@ -49,7 +44,7 @@ public class TestCLI {
|
||||||
JobID jobId = JobID.forName(jobIdStr);
|
JobID jobId = JobID.forName(jobIdStr);
|
||||||
Cluster mockCluster = mock(Cluster.class);
|
Cluster mockCluster = mock(Cluster.class);
|
||||||
Job job = mock(Job.class);
|
Job job = mock(Job.class);
|
||||||
CLI cli = spy(new CLI(new Configuration()));
|
CLI cli = spy(new CLI());
|
||||||
|
|
||||||
doReturn(mockCluster).when(cli).createCluster();
|
doReturn(mockCluster).when(cli).createCluster();
|
||||||
when(job.getTaskReports(TaskType.MAP)).thenReturn(
|
when(job.getTaskReports(TaskType.MAP)).thenReturn(
|
||||||
|
@ -117,7 +112,7 @@ public class TestCLI {
|
||||||
@Test
|
@Test
|
||||||
public void testJobKIll() throws Exception {
|
public void testJobKIll() throws Exception {
|
||||||
Cluster mockCluster = mock(Cluster.class);
|
Cluster mockCluster = mock(Cluster.class);
|
||||||
CLI cli = spy(new CLI(new Configuration()));
|
CLI cli = spy(new CLI());
|
||||||
doReturn(mockCluster).when(cli).createCluster();
|
doReturn(mockCluster).when(cli).createCluster();
|
||||||
String jobId1 = "job_1234654654_001";
|
String jobId1 = "job_1234654654_001";
|
||||||
String jobId2 = "job_1234654654_002";
|
String jobId2 = "job_1234654654_002";
|
||||||
|
@ -154,26 +149,4 @@ public class TestCLI {
|
||||||
when(mockJob.getStatus()).thenReturn(status);
|
when(mockJob.getStatus()).thenReturn(status);
|
||||||
return mockJob;
|
return mockJob;
|
||||||
}
|
}
|
||||||
|
|
||||||
@Test
|
|
||||||
public void testGetJob() throws Exception {
|
|
||||||
Configuration conf = new Configuration();
|
|
||||||
long sleepTime = 100;
|
|
||||||
conf.setLong(MRJobConfig.MR_CLIENT_JOB_RETRY_INTERVAL, sleepTime);
|
|
||||||
Cluster mockCluster = mock(Cluster.class);
|
|
||||||
JobID jobId1 = JobID.forName("job_1234654654_001");
|
|
||||||
when(mockCluster.getJob(jobId1)).thenReturn(null);
|
|
||||||
|
|
||||||
for (int i = 0; i < 2; ++i) {
|
|
||||||
conf.setInt(MRJobConfig.MR_CLIENT_JOB_MAX_RETRIES, i);
|
|
||||||
CLI cli = spy(new CLI(conf));
|
|
||||||
cli.cluster = mockCluster;
|
|
||||||
doReturn(mockCluster).when(cli).createCluster();
|
|
||||||
long start = Time.monotonicNow();
|
|
||||||
cli.getJob(jobId1);
|
|
||||||
long end = Time.monotonicNow();
|
|
||||||
Assert.assertTrue(end - start > (i * sleepTime));
|
|
||||||
Assert.assertTrue(end - start < ((i + 1) * sleepTime));
|
|
||||||
}
|
|
||||||
}
|
|
||||||
}
|
}
|
||||||
|
|
Loading…
Reference in New Issue