public static JobStatus FromYarn(ApplicationReport application, string jobFile)
{
string trackingUrl = application.GetTrackingUrl();
trackingUrl = trackingUrl == null ? string.Empty : trackingUrl;
JobStatus jobStatus = new JobStatus(TypeConverter.FromYarn(application.GetApplicationId
()), 0.0f, 0.0f, 0.0f, 0.0f, TypeConverter.FromYarn(application.GetYarnApplicationState
(), application.GetFinalApplicationStatus()), JobPriority.Normal, application.GetUser
(), application.GetName(), application.GetQueue(), jobFile, trackingUrl, false);
jobStatus.SetSchedulingInfo(trackingUrl);
// Set AM tracking url
jobStatus.SetStartTime(application.GetStartTime());
jobStatus.SetFinishTime(application.GetFinishTime());
jobStatus.SetFailureInfo(application.GetDiagnostics());
ApplicationResourceUsageReport resourceUsageReport = application.GetApplicationResourceUsageReport
();
if (resourceUsageReport != null)
{
jobStatus.SetNeededMem(resourceUsageReport.GetNeededResources().GetMemory());
jobStatus.SetNumReservedSlots(resourceUsageReport.GetNumReservedContainers());
jobStatus.SetNumUsedSlots(resourceUsageReport.GetNumUsedContainers());
jobStatus.SetReservedMem(resourceUsageReport.GetReservedResources().GetMemory());
jobStatus.SetUsedMem(resourceUsageReport.GetUsedResources().GetMemory());
}
return(jobStatus);
}
public AppInfo(ApplicationReport app)
{
// JAXB needs this
appId = app.GetApplicationId().ToString();
if (app.GetCurrentApplicationAttemptId() != null)
{
currentAppAttemptId = app.GetCurrentApplicationAttemptId().ToString();
}
user = app.GetUser();
queue = app.GetQueue();
name = app.GetName();
type = app.GetApplicationType();
host = app.GetHost();
rpcPort = app.GetRpcPort();
appState = app.GetYarnApplicationState();
diagnosticsInfo = app.GetDiagnostics();
trackingUrl = app.GetTrackingUrl();
originalTrackingUrl = app.GetOriginalTrackingUrl();
submittedTime = app.GetStartTime();
startedTime = app.GetStartTime();
finishedTime = app.GetFinishTime();
elapsedTime = Times.Elapsed(startedTime, finishedTime);
finalAppStatus = app.GetFinalApplicationStatus();
progress = app.GetProgress() * 100;
// in percent
if (app.GetApplicationTags() != null && !app.GetApplicationTags().IsEmpty())
{
this.applicationTags = StringHelper.CsvJoiner.Join(app.GetApplicationTags());
}
}
private void VerifyEnemyAppReport(ApplicationReport appReport)
{
NUnit.Framework.Assert.AreEqual("Enemy should not see app host!", Unavailable, appReport
.GetHost());
NUnit.Framework.Assert.AreEqual("Enemy should not see app rpc port!", -1, appReport
.GetRpcPort());
NUnit.Framework.Assert.AreEqual("Enemy should not see app client token!", null, appReport
.GetClientToAMToken());
NUnit.Framework.Assert.AreEqual("Enemy should not see app diagnostics!", Unavailable
, appReport.GetDiagnostics());
NUnit.Framework.Assert.AreEqual("Enemy should not see app tracking url!", Unavailable
, appReport.GetTrackingUrl());
NUnit.Framework.Assert.AreEqual("Enemy should not see app original tracking url!"
, Unavailable, appReport.GetOriginalTrackingUrl());
ApplicationResourceUsageReport usageReport = appReport.GetApplicationResourceUsageReport
();
NUnit.Framework.Assert.AreEqual("Enemy should not see app used containers", -1, usageReport
.GetNumUsedContainers());
NUnit.Framework.Assert.AreEqual("Enemy should not see app reserved containers", -
1, usageReport.GetNumReservedContainers());
NUnit.Framework.Assert.AreEqual("Enemy should not see app used resources", -1, usageReport
.GetUsedResources().GetMemory());
NUnit.Framework.Assert.AreEqual("Enemy should not see app reserved resources", -1
, usageReport.GetReservedResources().GetMemory());
NUnit.Framework.Assert.AreEqual("Enemy should not see app needed resources", -1,
usageReport.GetNeededResources().GetMemory());
}
/// <exception cref="System.IO.IOException"/>
public virtual GetJobReportResponse GetJobReport(GetJobReportRequest request)
{
JobReport jobReport = recordFactory.NewRecordInstance <JobReport>();
jobReport.SetJobId(request.GetJobId());
jobReport.SetJobState(jobState);
jobReport.SetUser(applicationReport.GetUser());
jobReport.SetStartTime(applicationReport.GetStartTime());
jobReport.SetDiagnostics(applicationReport.GetDiagnostics());
jobReport.SetJobName(applicationReport.GetName());
jobReport.SetTrackingUrl(applicationReport.GetTrackingUrl());
jobReport.SetFinishTime(applicationReport.GetFinishTime());
GetJobReportResponse resp = recordFactory.NewRecordInstance <GetJobReportResponse>
();
resp.SetJobReport(jobReport);
return(resp);
}
/// <summary>Monitor the submitted application for completion.</summary>
/// <remarks>
/// Monitor the submitted application for completion. Kill application if time
/// expires.
/// </remarks>
/// <param name="appId">Application Id of application to be monitored</param>
/// <returns>true if application completed successfully</returns>
/// <exception cref="Org.Apache.Hadoop.Yarn.Exceptions.YarnException"/>
/// <exception cref="System.IO.IOException"/>
private ApplicationReport MonitorApplication(ApplicationId appId, ICollection <YarnApplicationState
> finalState)
{
long foundAMCompletedTime = 0;
StringBuilder expectedFinalState = new StringBuilder();
bool first = true;
foreach (YarnApplicationState state in finalState)
{
if (first)
{
first = false;
expectedFinalState.Append(state.ToString());
}
else
{
expectedFinalState.Append("," + state.ToString());
}
}
while (true)
{
// Check app status every 1 second.
try
{
Sharpen.Thread.Sleep(1000);
}
catch (Exception)
{
Log.Debug("Thread sleep in monitoring loop interrupted");
}
// Get application report for the appId we are interested in
ApplicationReport report = rmClient.GetApplicationReport(appId);
Log.Info("Got application report from ASM for" + ", appId=" + appId.GetId() + ", appAttemptId="
+ report.GetCurrentApplicationAttemptId() + ", clientToAMToken=" + report.GetClientToAMToken
() + ", appDiagnostics=" + report.GetDiagnostics() + ", appMasterHost=" + report
.GetHost() + ", appQueue=" + report.GetQueue() + ", appMasterRpcPort=" + report.
GetRpcPort() + ", appStartTime=" + report.GetStartTime() + ", yarnAppState=" + report
.GetYarnApplicationState().ToString() + ", distributedFinalState=" + report.GetFinalApplicationStatus
().ToString() + ", appTrackingUrl=" + report.GetTrackingUrl() + ", appUser="******"Waited " + AmStateWaitTimeoutMs / 1000 + " seconds after process completed for AppReport"
+ " to reach desired final state. Not waiting anymore." + "CurrentState = " + state_1
+ ", ExpectedStates = " + expectedFinalState.ToString());
throw new RuntimeException("Failed to receive final expected state" + " in ApplicationReport"
+ ", CurrentState=" + state_1 + ", ExpectedStates=" + expectedFinalState.ToString
());
}
}
}
}
}
public virtual void TestFromYarnApplicationReport()
{
ApplicationId mockAppId = Org.Mockito.Mockito.Mock <ApplicationId>();
Org.Mockito.Mockito.When(mockAppId.GetClusterTimestamp()).ThenReturn(12345L);
Org.Mockito.Mockito.When(mockAppId.GetId()).ThenReturn(6789);
ApplicationReport mockReport = Org.Mockito.Mockito.Mock <ApplicationReport>();
Org.Mockito.Mockito.When(mockReport.GetTrackingUrl()).ThenReturn("dummy-tracking-url"
);
Org.Mockito.Mockito.When(mockReport.GetApplicationId()).ThenReturn(mockAppId);
Org.Mockito.Mockito.When(mockReport.GetYarnApplicationState()).ThenReturn(YarnApplicationState
.Killed);
Org.Mockito.Mockito.When(mockReport.GetUser()).ThenReturn("dummy-user");
Org.Mockito.Mockito.When(mockReport.GetQueue()).ThenReturn("dummy-queue");
string jobFile = "dummy-path/job.xml";
try
{
JobStatus status = TypeConverter.FromYarn(mockReport, jobFile);
}
catch (ArgumentNullException)
{
NUnit.Framework.Assert.Fail("Type converstion from YARN fails for jobs without "
+ "ApplicationUsageReport");
}
ApplicationResourceUsageReport appUsageRpt = Org.Apache.Hadoop.Yarn.Util.Records.
NewRecord <ApplicationResourceUsageReport>();
Resource r = Org.Apache.Hadoop.Yarn.Util.Records.NewRecord <Resource>();
r.SetMemory(2048);
appUsageRpt.SetNeededResources(r);
appUsageRpt.SetNumReservedContainers(1);
appUsageRpt.SetNumUsedContainers(3);
appUsageRpt.SetReservedResources(r);
appUsageRpt.SetUsedResources(r);
Org.Mockito.Mockito.When(mockReport.GetApplicationResourceUsageReport()).ThenReturn
(appUsageRpt);
JobStatus status_1 = TypeConverter.FromYarn(mockReport, jobFile);
NUnit.Framework.Assert.IsNotNull("fromYarn returned null status", status_1);
NUnit.Framework.Assert.AreEqual("jobFile set incorrectly", "dummy-path/job.xml",
status_1.GetJobFile());
NUnit.Framework.Assert.AreEqual("queue set incorrectly", "dummy-queue", status_1.
GetQueue());
NUnit.Framework.Assert.AreEqual("trackingUrl set incorrectly", "dummy-tracking-url"
, status_1.GetTrackingUrl());
NUnit.Framework.Assert.AreEqual("user set incorrectly", "dummy-user", status_1.GetUsername
());
NUnit.Framework.Assert.AreEqual("schedulingInfo set incorrectly", "dummy-tracking-url"
, status_1.GetSchedulingInfo());
NUnit.Framework.Assert.AreEqual("jobId set incorrectly", 6789, status_1.GetJobID(
).GetId());
NUnit.Framework.Assert.AreEqual("state set incorrectly", JobStatus.State.Killed,
status_1.GetState());
NUnit.Framework.Assert.AreEqual("needed mem info set incorrectly", 2048, status_1
.GetNeededMem());
NUnit.Framework.Assert.AreEqual("num rsvd slots info set incorrectly", 1, status_1
.GetNumReservedSlots());
NUnit.Framework.Assert.AreEqual("num used slots info set incorrectly", 3, status_1
.GetNumUsedSlots());
NUnit.Framework.Assert.AreEqual("rsvd mem info set incorrectly", 2048, status_1.GetReservedMem
());
NUnit.Framework.Assert.AreEqual("used mem info set incorrectly", 2048, status_1.GetUsedMem
());
}
/// <exception cref="System.IO.IOException"/>
private MRClientProtocol GetProxy()
{
if (realProxy != null)
{
return(realProxy);
}
// Possibly allow nulls through the PB tunnel, otherwise deal with an exception
// and redirect to the history server.
ApplicationReport application = null;
try
{
application = rm.GetApplicationReport(appId);
}
catch (ApplicationNotFoundException)
{
application = null;
}
catch (YarnException e2)
{
throw new IOException(e2);
}
if (application != null)
{
trackingUrl = application.GetTrackingUrl();
}
IPEndPoint serviceAddr = null;
while (application == null || YarnApplicationState.Running == application.GetYarnApplicationState
())
{
if (application == null)
{
Log.Info("Could not get Job info from RM for job " + jobId + ". Redirecting to job history server."
);
return(CheckAndGetHSProxy(null, JobState.New));
}
try
{
if (application.GetHost() == null || string.Empty.Equals(application.GetHost()))
{
Log.Debug("AM not assigned to Job. Waiting to get the AM ...");
Sharpen.Thread.Sleep(2000);
Log.Debug("Application state is " + application.GetYarnApplicationState());
application = rm.GetApplicationReport(appId);
continue;
}
else
{
if (Unavailable.Equals(application.GetHost()))
{
if (!amAclDisabledStatusLogged)
{
Log.Info("Job " + jobId + " is running, but the host is unknown." + " Verify user has VIEW_JOB access."
);
amAclDisabledStatusLogged = true;
}
return(GetNotRunningJob(application, JobState.Running));
}
}
if (!conf.GetBoolean(MRJobConfig.JobAmAccessDisabled, false))
{
UserGroupInformation newUgi = UserGroupInformation.CreateRemoteUser(UserGroupInformation
.GetCurrentUser().GetUserName());
serviceAddr = NetUtils.CreateSocketAddrForHost(application.GetHost(), application
.GetRpcPort());
if (UserGroupInformation.IsSecurityEnabled())
{
Token clientToAMToken = application.GetClientToAMToken();
Org.Apache.Hadoop.Security.Token.Token <ClientToAMTokenIdentifier> token = ConverterUtils
.ConvertFromYarn(clientToAMToken, serviceAddr);
newUgi.AddToken(token);
}
Log.Debug("Connecting to " + serviceAddr);
IPEndPoint finalServiceAddr = serviceAddr;
realProxy = newUgi.DoAs(new _PrivilegedExceptionAction_202(this, finalServiceAddr
));
}
else
{
if (!amAclDisabledStatusLogged)
{
Log.Info("Network ACL closed to AM for job " + jobId + ". Not going to try to reach the AM."
);
amAclDisabledStatusLogged = true;
}
return(GetNotRunningJob(null, JobState.Running));
}
return(realProxy);
}
catch (IOException)
{
//possibly the AM has crashed
//there may be some time before AM is restarted
//keep retrying by getting the address from RM
Log.Info("Could not connect to " + serviceAddr + ". Waiting for getting the latest AM address..."
);
try
{
Sharpen.Thread.Sleep(2000);
}
catch (Exception e1)
{
Log.Warn("getProxy() call interruped", e1);
throw new YarnRuntimeException(e1);
}
try
{
application = rm.GetApplicationReport(appId);
}
catch (YarnException e1)
{
throw new IOException(e1);
}
if (application == null)
{
Log.Info("Could not get Job info from RM for job " + jobId + ". Redirecting to job history server."
);
return(CheckAndGetHSProxy(null, JobState.Running));
}
}
catch (Exception e)
{
Log.Warn("getProxy() call interruped", e);
throw new YarnRuntimeException(e);
}
catch (YarnException e)
{
throw new IOException(e);
}
}
string user = application.GetUser();
if (user == null)
{
throw new IOException("User is not set in the application report");
}
if (application.GetYarnApplicationState() == YarnApplicationState.New || application
.GetYarnApplicationState() == YarnApplicationState.NewSaving || application.GetYarnApplicationState
() == YarnApplicationState.Submitted || application.GetYarnApplicationState() ==
YarnApplicationState.Accepted)
{
realProxy = null;
return(GetNotRunningJob(application, JobState.New));
}
if (application.GetYarnApplicationState() == YarnApplicationState.Failed)
{
realProxy = null;
return(GetNotRunningJob(application, JobState.Failed));
}
if (application.GetYarnApplicationState() == YarnApplicationState.Killed)
{
realProxy = null;
return(GetNotRunningJob(application, JobState.Killed));
}
//History server can serve a job only if application
//succeeded.
if (application.GetYarnApplicationState() == YarnApplicationState.Finished)
{
Log.Info("Application state is completed. FinalApplicationStatus=" + application.
GetFinalApplicationStatus().ToString() + ". Redirecting to job history server");
realProxy = CheckAndGetHSProxy(application, JobState.Succeeded);
}
return(realProxy);
}
/// <summary>Monitor the submitted application for completion.</summary>
/// <remarks>
/// Monitor the submitted application for completion.
/// Kill application if time expires.
/// </remarks>
/// <param name="appId">Application Id of application to be monitored</param>
/// <returns>true if application completed successfully</returns>
/// <exception cref="Org.Apache.Hadoop.Yarn.Exceptions.YarnException"/>
/// <exception cref="System.IO.IOException"/>
private bool MonitorApplication(ApplicationId appId)
{
while (true)
{
// Check app status every 1 second.
try
{
Sharpen.Thread.Sleep(1000);
}
catch (Exception)
{
Log.Debug("Thread sleep in monitoring loop interrupted");
}
// Get application report for the appId we are interested in
ApplicationReport report = yarnClient.GetApplicationReport(appId);
Log.Info("Got application report from ASM for" + ", appId=" + appId.GetId() + ", clientToAMToken="
+ report.GetClientToAMToken() + ", appDiagnostics=" + report.GetDiagnostics() +
", appMasterHost=" + report.GetHost() + ", appQueue=" + report.GetQueue() + ", appMasterRpcPort="
+ report.GetRpcPort() + ", appStartTime=" + report.GetStartTime() + ", yarnAppState="
+ report.GetYarnApplicationState().ToString() + ", distributedFinalState=" + report
.GetFinalApplicationStatus().ToString() + ", appTrackingUrl=" + report.GetTrackingUrl
() + ", appUser="******"Application has completed successfully. Breaking monitoring loop");
return(true);
}
else
{
Log.Info("Application did finished unsuccessfully." + " YarnState=" + state.ToString
() + ", DSFinalStatus=" + dsStatus.ToString() + ". Breaking monitoring loop");
return(false);
}
}
else
{
if (YarnApplicationState.Killed == state || YarnApplicationState.Failed == state)
{
Log.Info("Application did not finish." + " YarnState=" + state.ToString() + ", DSFinalStatus="
+ dsStatus.ToString() + ". Breaking monitoring loop");
return(false);
}
}
if (Runtime.CurrentTimeMillis() > (clientStartTime + clientTimeout))
{
Log.Info("Reached client specified timeout for application. Killing application");
ForceKillApplication(appId);
return(false);
}
}
}
