Exemple #1
0
void main_shutdown_rescue( int exitVal, Dag::dag_status dagStatus ) {
		// Avoid possible infinite recursion if you hit a fatal error
		// while writing a rescue DAG.
	static bool inShutdownRescue = false;
	if ( inShutdownRescue ) {
		return;
	}
	inShutdownRescue = true;

	dagman.dag->_dagStatus = dagStatus;
	debug_printf( DEBUG_QUIET, "Aborting DAG...\n" );
		// Avoid writing two different rescue DAGs if the "main" DAG and
		// the final node (if any) both fail.
	static bool wroteRescue = false;
	if( dagman.dag ) {
			// We write the rescue DAG *before* removing jobs because
			// otherwise if we crashed, failed, or were killed while
			// removing them, we would leave the DAG in an
			// unrecoverable state...
		if( exitVal != 0 ) {
			if ( dagman.maxRescueDagNum > 0 ) {
				dagman.dag->Rescue( dagman.primaryDagFile.Value(),
							dagman.multiDags, dagman.maxRescueDagNum,
							wroteRescue, false,
							dagman._writePartialRescueDag );
				wroteRescue = true;
			} else {
				debug_printf( DEBUG_QUIET, "No rescue DAG written because "
							"DAGMAN_MAX_RESCUE_NUM is 0\n" );
			}
		}

		debug_printf( DEBUG_DEBUG_1, "We have %d running jobs to remove\n",
					dagman.dag->NumJobsSubmitted() );
		if( dagman.dag->NumJobsSubmitted() > 0 ) {
			debug_printf( DEBUG_NORMAL, "Removing submitted jobs...\n" );
			dagman.dag->RemoveRunningJobs(dagman);
		}
		if ( dagman.dag->NumScriptsRunning() > 0 ) {
			debug_printf( DEBUG_NORMAL, "Removing running scripts...\n" );
			dagman.dag->RemoveRunningScripts();
		}
		dagman.dag->PrintDeferrals( DEBUG_NORMAL, true );

			// Start the final node if we have one.
		if ( dagman.dag->StartFinalNode() ) {
				// We started a final node; return here so we wait for the
				// final node to finish, instead of exiting immediately.
			inShutdownRescue = false;
			return;
		}
		dagman.dag->DumpNodeStatus( false, true );
		dagman.dag->GetJobstateLog().WriteDagmanFinished( exitVal );
	}
	MSC_SUPPRESS_WARNING_FIXME(6031) // return falue of unlink ignored.
	unlink( lockFileName ); 
	dagman.CleanUp();
	inShutdownRescue = false;
	DC_Exit( exitVal );
}
Exemple #2
0
// this can be called by other functions, or by DC when the schedd is
// shutdown gracefully
void main_shutdown_graceful() {
	print_status();
	dagman.dag->DumpNodeStatus( true, false );
	dagman.dag->GetJobstateLog().WriteDagmanFinished( EXIT_RESTART );
	dagman.CleanUp();
	DC_Exit( EXIT_RESTART );
}
Exemple #3
0
void ExitSuccess() {
	dagman.dag->DumpNodeStatus( false, false );
	dagman.dag->GetJobstateLog().WriteDagmanFinished( EXIT_OKAY );
	MSC_SUPPRESS_WARNING_FIXME(6031) // return falue of unlink ignored.
	unlink( lockFileName ); 
	dagman.CleanUp();
	DC_Exit( EXIT_OKAY );
}
Exemple #4
0
void ExitSuccess() {
	print_status();
	dagman.dag->DumpNodeStatus( false, false );
	dagman.dag->GetJobstateLog().WriteDagmanFinished( EXIT_OKAY );
	tolerant_unlink( lockFileName ); 
	dagman.CleanUp();
	DC_Exit( EXIT_OKAY );
}
Exemple #5
0
//---------------------------------------------------------------------------
void main_init (int argc, char ** const argv) {

	printf ("Executing condor dagman ... \n");

		// flag used if DAGMan is invoked with -WaitForDebug so we
		// wait for a developer to attach with a debugger...
	volatile int wait_for_debug = 0;

		// process any config vars -- this happens before we process
		// argv[], since arguments should override config settings
	dagman.Config();

	// The DCpermission (last parm) should probably be PARENT, if it existed
    daemonCore->Register_Signal( SIGUSR1, "SIGUSR1",
                                 (SignalHandler) main_shutdown_remove,
                                 "main_shutdown_remove", NULL);

/****** FOR TESTING *******
    daemonCore->Register_Signal( SIGUSR2, "SIGUSR2",
                                 (SignalHandler) main_testing_stub,
                                 "main_testing_stub", NULL);
****** FOR TESTING ********/
    debug_progname = condor_basename(argv[0]);

		// condor_submit_dag version from .condor.sub
	bool allowVerMismatch = false;
	const char *csdVersion = "undefined";

	int i;
    for (i = 0 ; i < argc ; i++) {
        debug_printf( DEBUG_NORMAL, "argv[%d] == \"%s\"\n", i, argv[i] );
    }

    if (argc < 2) Usage();  //  Make sure an input file was specified

		// get dagman job id from environment, if it's there
		// (otherwise it will be set to "-1.-1.-1")
	dagman.DAGManJobId.SetFromString( getenv( EnvGetName( ENV_ID ) ) );

	//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
		// Minimum legal version for a .condor.sub file to be compatible
		// with this condor_dagman binary.

		// !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
		// Be sure to change this if the arguments or environment
		// passed to condor_dagman change in an incompatible way!!
		// !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

	struct DagVersionData {
		int majorVer;
		int minorVer;
		int subMinorVer;
	};
	const DagVersionData MIN_SUBMIT_FILE_VERSION = { 7, 1, 2 };

		// Construct a string of the minimum submit file version.
	MyString minSubmitVersionStr;
	minSubmitVersionStr.formatstr( "%d.%d.%d",
				MIN_SUBMIT_FILE_VERSION.majorVer,
				MIN_SUBMIT_FILE_VERSION.minorVer,
				MIN_SUBMIT_FILE_VERSION.subMinorVer );
	//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

    //
    // Process command-line arguments
    //
    for (i = 1; i < argc; i++) {
        if( !strcasecmp( "-Debug", argv[i] ) ) {
            i++;
            if( argc <= i || strcmp( argv[i], "" ) == 0 ) {
                debug_printf( DEBUG_SILENT, "No debug level specified\n" );
                Usage();
            }
            debug_level = (debug_level_t) atoi (argv[i]);
        } else if( !strcasecmp( "-Lockfile", argv[i] ) ) {
            i++;
            if( argc <= i || strcmp( argv[i], "" ) == 0 ) {
                debug_printf( DEBUG_SILENT, "No DagMan lockfile specified\n" );
                Usage();
            }
            lockFileName = argv[i];
        } else if( !strcasecmp( "-Help", argv[i] ) ) {
            Usage();
        } else if (!strcasecmp( "-Dag", argv[i] ) ) {
            i++;
            if( argc <= i || strcmp( argv[i], "" ) == 0 ) {
                debug_printf( DEBUG_SILENT, "No DAG specified\n" );
                Usage();
            }
			dagman.dagFiles.append( argv[i] );
        } else if( !strcasecmp( "-MaxIdle", argv[i] ) ) {
            i++;
            if( argc <= i || strcmp( argv[i], "" ) == 0 ) {
                debug_printf( DEBUG_SILENT,
							  "Integer missing after -MaxIdle\n" );
                Usage();
            }
            dagman.maxIdle = atoi( argv[i] );
        } else if( !strcasecmp( "-MaxJobs", argv[i] ) ) {
            i++;
            if( argc <= i || strcmp( argv[i], "" ) == 0 ) {
                debug_printf( DEBUG_SILENT,
							  "Integer missing after -MaxJobs\n" );
                Usage();
            }
            dagman.maxJobs = atoi( argv[i] );
        } else if( !strcasecmp( "-MaxScripts", argv[i] ) ) {
			debug_printf( DEBUG_SILENT, "-MaxScripts has been replaced with "
						   "-MaxPre and -MaxPost arguments\n" );
			Usage();
        } else if( !strcasecmp( "-MaxPre", argv[i] ) ) {
            i++;
            if( argc <= i || strcmp( argv[i], "" ) == 0 ) {
                debug_printf( DEBUG_SILENT,
							  "Integer missing after -MaxPre\n" );
                Usage();
            }
            dagman.maxPreScripts = atoi( argv[i] );
        } else if( !strcasecmp( "-MaxPost", argv[i] ) ) {
            i++;
            if( argc <= i || strcmp( argv[i], "" ) == 0 ) {
                debug_printf( DEBUG_SILENT,
							  "Integer missing after -MaxPost\n" );
                Usage();
            }
            dagman.maxPostScripts = atoi( argv[i] );
        } else if( !strcasecmp( "-NoEventChecks", argv[i] ) ) {
			debug_printf( DEBUG_QUIET, "Warning: -NoEventChecks is "
						"ignored; please use the DAGMAN_ALLOW_EVENTS "
						"config parameter instead\n");
			check_warning_strictness( DAG_STRICT_1 );

        } else if( !strcasecmp( "-AllowLogError", argv[i] ) ) {
			dagman.allowLogError = true;

        } else if( !strcasecmp( "-DontAlwaysRunPost",argv[i] ) ) {
			dagman._runPost = false;

        } else if( !strcasecmp( "-WaitForDebug", argv[i] ) ) {
			wait_for_debug = 1;

        } else if( !strcasecmp( "-UseDagDir", argv[i] ) ) {
			dagman.useDagDir = true;

        } else if( !strcasecmp( "-AutoRescue", argv[i] ) ) {
            i++;
            if( argc <= i || strcmp( argv[i], "" ) == 0 ) {
                debug_printf( DEBUG_SILENT, "No AutoRescue value specified\n" );
                Usage();
            }
            dagman.autoRescue = (atoi( argv[i] ) != 0);

        } else if( !strcasecmp( "-DoRescueFrom", argv[i] ) ) {
            i++;
            if( argc <= i || strcmp( argv[i], "" ) == 0 ) {
                debug_printf( DEBUG_SILENT, "No rescue DAG number specified\n" );
                Usage();
            }
            dagman.doRescueFrom = atoi (argv[i]);

        } else if( !strcasecmp( "-CsdVersion", argv[i] ) ) {
            i++;
            if( argc <= i || strcmp( argv[i], "" ) == 0 ) {
                debug_printf( DEBUG_SILENT, "No CsdVersion value specified\n" );
                Usage();
            }
			csdVersion = argv[i];

        } else if( !strcasecmp( "-AllowVersionMismatch", argv[i] ) ) {
			allowVerMismatch = true;

        } else if( !strcasecmp( "-DumpRescue", argv[i] ) ) {
			dagman.dumpRescueDag = true;

        } else if( !strcasecmp( "-verbose", argv[i] ) ) {
			dagman._submitDagDeepOpts.bVerbose = true;

        } else if( !strcasecmp( "-force", argv[i] ) ) {
			dagman._submitDagDeepOpts.bForce = true;
		
        } else if( !strcasecmp( "-notification", argv[i] ) ) {
            i++;
            if( argc <= i || strcmp( argv[i], "" ) == 0 ) {
                debug_printf( DEBUG_SILENT, "No notification value specified\n" );
                Usage();
            }
			dagman._submitDagDeepOpts.strNotification = argv[i];

        } else if( !strcasecmp( "-dagman", argv[i] ) ) {
            i++;
            if( argc <= i || strcmp( argv[i], "" ) == 0 ) {
                debug_printf( DEBUG_SILENT, "No dagman value specified\n" );
                Usage();
            }
			dagman._submitDagDeepOpts.strDagmanPath = argv[i];

        } else if( !strcasecmp( "-outfile_dir", argv[i] ) ) {
            i++;
            if( argc <= i || strcmp( argv[i], "" ) == 0 ) {
                debug_printf( DEBUG_SILENT, "No outfile_dir value specified\n" );
                Usage();
            }
			dagman._submitDagDeepOpts.strOutfileDir = argv[i];

        } else if( !strcasecmp( "-update_submit", argv[i] ) ) {
			dagman._submitDagDeepOpts.updateSubmit = true;

        } else if( !strcasecmp( "-import_env", argv[i] ) ) {
			dagman._submitDagDeepOpts.importEnv = true;

        } else if( !strcasecmp( "-priority", argv[i] ) ) {
		++i;
		if( i >= argc || strcmp( argv[i], "" ) == 0 ) {
			debug_printf( DEBUG_NORMAL, "No priority value specified\n");
			Usage();
		}
		dagman._submitDagDeepOpts.priority = atoi(argv[i]);
		} else if( !strcasecmp( "-dont_use_default_node_log", argv[i] ) ) {
			dagman._submitDagDeepOpts.always_use_node_log = false;
        } else {
    		debug_printf( DEBUG_SILENT, "\nUnrecognized argument: %s\n",
						argv[i] );
			Usage();
		}
    }

	dagman.dagFiles.rewind();
	dagman.primaryDagFile = dagman.dagFiles.next();
	dagman.multiDags = (dagman.dagFiles.number() > 1);

	MyString tmpDefaultLog;
	if ( dagman._defaultNodeLog != NULL ) {
		tmpDefaultLog = dagman._defaultNodeLog;
		free( dagman._defaultNodeLog );
	} else {
		tmpDefaultLog = dagman.primaryDagFile + ".nodes.log";
	}

		// Force default log file path to be absolute so it works
		// with -usedagdir and DIR nodes.
	CondorError errstack;
	if ( !MultiLogFiles::makePathAbsolute( tmpDefaultLog, errstack) ) {
       	debug_printf( DEBUG_QUIET, "Unable to convert default log "
					"file name to absolute path: %s\n",
					errstack.getFullText().c_str() );
		dagman.dag->GetJobstateLog().WriteDagmanFinished( EXIT_ERROR );
		DC_Exit( EXIT_ERROR );
	}
	dagman._defaultNodeLog = strdup( tmpDefaultLog.Value() );
	debug_printf( DEBUG_NORMAL, "Default node log file is: <%s>\n",
				dagman._defaultNodeLog);

    //
    // Check the arguments
    //

	//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
	// Checking for version compatibility between the .condor.sub
	// file and this condor_dagman binary...

	// Note: if we're in recovery mode and the submit file version
	// causes us to quit, we leave any existing node jobs still
	// running -- may want to change that eventually.  wenger 2009-10-13.

		// Version of the condor_submit_dag that created our submit file.
	CondorVersionInfo submitFileVersion( csdVersion );

		// Version of this condor_dagman binary.
	CondorVersionInfo dagmanVersion;

		// Just generate this message fragment in one place.
	MyString versionMsg;
	versionMsg.formatstr("the version (%s) of this DAG's Condor submit "
				"file (created by condor_submit_dag)", csdVersion );

		// Make sure version in submit file is valid.
	if( !submitFileVersion.is_valid() ) {
		if ( !allowVerMismatch ) {
        	debug_printf( DEBUG_QUIET, "Error: %s is invalid!\n",
						versionMsg.Value() );
			DC_Exit( EXIT_ERROR );
		} else {
        	debug_printf( DEBUG_NORMAL, "Warning: %s is invalid; "
						"continuing because of -AllowVersionMismatch flag\n",
						versionMsg.Value() );
		}

		// Make sure .condor.sub file is recent enough.
	} else if ( submitFileVersion.compare_versions(
				CondorVersion() ) != 0 ) {

		if( !submitFileVersion.built_since_version(
					MIN_SUBMIT_FILE_VERSION.majorVer,
					MIN_SUBMIT_FILE_VERSION.minorVer,
					MIN_SUBMIT_FILE_VERSION.subMinorVer ) ) {
			if ( !allowVerMismatch ) {
        		debug_printf( DEBUG_QUIET, "Error: %s is older than "
							"oldest permissible version (%s)\n",
							versionMsg.Value(), minSubmitVersionStr.Value() );
				DC_Exit( EXIT_ERROR );
			} else {
        		debug_printf( DEBUG_NORMAL, "Warning: %s is older than "
							"oldest permissible version (%s); continuing "
							"because of -AllowVersionMismatch flag\n",
							versionMsg.Value(), minSubmitVersionStr.Value() );
			}

			// Warn if .condor.sub file is a newer version than this binary.
		} else if (dagmanVersion.compare_versions( csdVersion ) > 0 ) {
        	debug_printf( DEBUG_NORMAL, "Warning: %s is newer than "
						"condor_dagman version (%s)\n", versionMsg.Value(),
						CondorVersion() );
			check_warning_strictness( DAG_STRICT_3 );
		} else {
        	debug_printf( DEBUG_NORMAL, "Note: %s differs from "
						"condor_dagman version (%s), but the "
						"difference is permissible\n", 
						versionMsg.Value(), CondorVersion() );
		}
	}
	//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

    if( dagman.primaryDagFile == "" ) {
        debug_printf( DEBUG_SILENT, "No DAG file was specified\n" );
        Usage();
    }
    if (lockFileName == NULL) {
        debug_printf( DEBUG_SILENT, "No DAG lock file was specified\n" );
        Usage();
    }
    if( dagman.maxJobs < 0 ) {
        debug_printf( DEBUG_SILENT, "-MaxJobs must be non-negative\n");
        Usage();
    }
    if( dagman.maxPreScripts < 0 ) {
        debug_printf( DEBUG_SILENT, "-MaxPre must be non-negative\n" );
        Usage();
    }
    if( dagman.maxPostScripts < 0 ) {
        debug_printf( DEBUG_SILENT, "-MaxPost must be non-negative\n" );
        Usage();
    }
    if( dagman.doRescueFrom < 0 ) {
        debug_printf( DEBUG_SILENT, "-DoRescueFrom must be non-negative\n" );
        Usage();
    }

    debug_printf( DEBUG_VERBOSE, "DAG Lockfile will be written to %s\n",
                   lockFileName );
	if ( dagman.dagFiles.number() == 1 ) {
    	debug_printf( DEBUG_VERBOSE, "DAG Input file is %s\n",
				  	dagman.primaryDagFile.Value() );
	} else {
		MyString msg = "DAG Input files are ";
		dagman.dagFiles.rewind();
		const char *dagFile;
		while ( (dagFile = dagman.dagFiles.next()) != NULL ) {
			msg += dagFile;
			msg += " ";
		}
		msg += "\n";
    	debug_printf( DEBUG_VERBOSE, "%s", msg.Value() );
	}

		// if requested, wait for someone to attach with a debugger...
	while( wait_for_debug ) { }

    {
		MyString cwd;
		if( !condor_getcwd(cwd) ) {
			cwd = "<null>";
		}
        debug_printf( DEBUG_DEBUG_1, "Current path is %s\n",cwd.Value());

		char *temp = my_username();
		debug_printf( DEBUG_DEBUG_1, "Current user is %s\n",
					   temp ? temp : "<null>" );
		if( temp ) {
			free( temp );
		}
    }

		//
		// Figure out the rescue DAG to run, if any (this is with "new-
		// style" rescue DAGs).
		//
	int rescueDagNum = 0;
	MyString rescueDagMsg;

	if ( dagman.doRescueFrom != 0 ) {
		rescueDagNum = dagman.doRescueFrom;
		rescueDagMsg.formatstr( "Rescue DAG number %d specified", rescueDagNum );
		RenameRescueDagsAfter( dagman.primaryDagFile.Value(),
					dagman.multiDags, rescueDagNum, dagman.maxRescueDagNum );

	} else if ( dagman.autoRescue ) {
		rescueDagNum = FindLastRescueDagNum(
					dagman.primaryDagFile.Value(),
					dagman.multiDags, dagman.maxRescueDagNum );
		rescueDagMsg.formatstr( "Found rescue DAG number %d", rescueDagNum );
	}

		//
		// Fill in values in the deep submit options that we haven't
		// already set.
		//
	dagman._submitDagDeepOpts.bAllowLogError = dagman.allowLogError;
	dagman._submitDagDeepOpts.useDagDir = dagman.useDagDir;
	dagman._submitDagDeepOpts.autoRescue = dagman.autoRescue;
	dagman._submitDagDeepOpts.doRescueFrom = dagman.doRescueFrom;
	dagman._submitDagDeepOpts.allowVerMismatch = allowVerMismatch;
	dagman._submitDagDeepOpts.recurse = false;

    //
    // Create the DAG
    //

	// Note: a bunch of the parameters we pass here duplicate things
	// in submitDagOpts, but I'm keeping them separate so we don't have to
	// bother to construct a new SubmitDagOtions object for splices.
	// wenger 2010-03-25
    dagman.dag = new Dag( dagman.dagFiles, dagman.maxJobs,
						  dagman.maxPreScripts, dagman.maxPostScripts,
						  dagman.allowLogError, dagman.useDagDir,
						  dagman.maxIdle, dagman.retrySubmitFirst,
						  dagman.retryNodeFirst, dagman.condorRmExe,
						  dagman.storkRmExe, &dagman.DAGManJobId,
						  dagman.prohibitMultiJobs, dagman.submitDepthFirst,
						  dagman._defaultNodeLog,
						  dagman._generateSubdagSubmits,
						  &dagman._submitDagDeepOpts,
						  false ); /* toplevel dag! */

    if( dagman.dag == NULL ) {
        EXCEPT( "ERROR: out of memory!\n");
    }

	dagman.dag->SetAbortOnScarySubmit( dagman.abortOnScarySubmit );
	dagman.dag->SetAllowEvents( dagman.allow_events );
	dagman.dag->SetConfigFile( dagman._dagmanConfigFile );
	dagman.dag->SetMaxJobHolds( dagman._maxJobHolds );
	dagman.dag->SetPostRun(dagman._runPost);
	if( dagman._submitDagDeepOpts.priority != 0 ) { // From command line
		dagman.dag->SetDefaultPriority(dagman._submitDagDeepOpts.priority);
	} else if( dagman._defaultPriority != 0 ) { // From config file
		dagman.dag->SetDefaultPriority(dagman._defaultPriority);
		dagman._submitDagDeepOpts.priority = dagman._defaultPriority;
	}

    //
    // Parse the input files.  The parse() routine
    // takes care of adding jobs and dependencies to the DagMan
    //
	dagman.mungeNodeNames = (dagman.dagFiles.number() > 1);
	parseSetDoNameMunge( dagman.mungeNodeNames );
   	debug_printf( DEBUG_VERBOSE, "Parsing %d dagfiles\n", 
		dagman.dagFiles.number() );
	dagman.dagFiles.rewind();
	char *dagFile;

	// Here we make a copy of the dagFiles for iteration purposes. Deep inside
	// of the parsing, copies of the dagman.dagFile string list happen which
	// mess up the iteration of this list.
	StringList sl( dagman.dagFiles );
	sl.rewind();
	while ( (dagFile = sl.next()) != NULL ) {
    	debug_printf( DEBUG_VERBOSE, "Parsing %s ...\n", dagFile );

    	if( !parse( dagman.dag, dagFile, dagman.useDagDir ) ) {
			if ( dagman.dumpRescueDag ) {
					// Dump the rescue DAG so we can see what we got
					// in the failed parse attempt.
    			debug_printf( DEBUG_QUIET, "Dumping rescue DAG "
							"because of -DumpRescue flag\n" );
				dagman.dag->Rescue( dagman.primaryDagFile.Value(),
							dagman.multiDags, dagman.maxRescueDagNum,
							false, true, false );
			}
			
			dagman.dag->RemoveRunningJobs(dagman, true);
			MSC_SUPPRESS_WARNING_FIXME(6031) // return falue of unlink ignored.
			unlink( lockFileName );
			dagman.CleanUp();
			
				// Note: debug_error calls DC_Exit().
        	debug_error( 1, DEBUG_QUIET, "Failed to parse %s\n",
					 	dagFile );
    	}
	}
	if( dagman.dag->GetDefaultPriority() != 0 ) {
		dagman.dag->SetDefaultPriorities(); // Applies to the nodes of the dag
	}
	dagman.dag->GetJobstateLog().WriteDagmanStarted( dagman.DAGManJobId );
	if ( rescueDagNum > 0 ) {
			// Get our Pegasus sequence numbers set correctly.
		dagman.dag->GetJobstateLog().InitializeRescue();
	}

	// lift the final set of splices into the main dag.
	dagman.dag->LiftSplices(SELF);

		//
		// Actually parse the "new-new" style (partial DAG info only)
		// rescue DAG here.  Note: this *must* be done after splices
		// are lifted!
		//
	if ( rescueDagNum > 0 ) {
		dagman.rescueFileToRun = RescueDagName(
					dagman.primaryDagFile.Value(),
					dagman.multiDags, rescueDagNum );
		debug_printf ( DEBUG_QUIET, "%s; running %s in combination with "
					"normal DAG file%s\n", rescueDagMsg.Value(),
					dagman.rescueFileToRun.Value(),
					dagman.multiDags ? "s" : "");
		debug_printf ( DEBUG_QUIET,
					"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n");

		debug_printf ( DEBUG_QUIET, "USING RESCUE DAG %s\n",
					dagman.rescueFileToRun.Value() );

			// Turn off node name munging for the rescue DAG, because
			// it will already have munged node names.
		parseSetDoNameMunge( false );

    	if( !parse( dagman.dag, dagman.rescueFileToRun.Value(),
					dagman.useDagDir ) ) {
			if ( dagman.dumpRescueDag ) {
					// Dump the rescue DAG so we can see what we got
					// in the failed parse attempt.
    			debug_printf( DEBUG_QUIET, "Dumping rescue DAG "
							"because of -DumpRescue flag\n" );
				dagman.dag->Rescue( dagman.primaryDagFile.Value(),
							dagman.multiDags, dagman.maxRescueDagNum,
							true, false );
			}
			
			dagman.dag->RemoveRunningJobs(dagman, true);
			MSC_SUPPRESS_WARNING_FIXME(6031) // return falue of unlink ignored.
			unlink( lockFileName );
			dagman.CleanUp();
			
				// Note: debug_error calls DC_Exit().
        	debug_error( 1, DEBUG_QUIET, "Failed to parse %s\n",
					 	dagFile );
    	}
	}

	dagman.dag->CheckThrottleCats();

	// fix up any use of $(JOB) in the vars values for any node
	dagman.dag->ResolveVarsInterpolations();

/*	debug_printf(DEBUG_QUIET, "COMPLETED DAG!\n");*/
/*	dagman.dag->PrintJobList();*/

#ifndef NOT_DETECT_CYCLE
	if( dagman.startup_cycle_detect && dagman.dag->isCycle() )
	{
		// Note: maybe we should run the final node here, if there is one.
		// wenger 2011-12-19.
		debug_error (1, DEBUG_QUIET, "ERROR: a cycle exists in the dag, please check input\n");
	}
#endif
    debug_printf( DEBUG_VERBOSE, "Dag contains %d total jobs\n",
				  dagman.dag->NumNodes( true ) );

	MyString firstLocation;
	if ( dagman.dag->GetReject( firstLocation ) ) {
    	debug_printf( DEBUG_QUIET, "Exiting because of REJECT "
					"specification in %s.  This most likely means "
					"that the DAG file was produced with the -DumpRescue "
					"flag when parsing the original DAG failed.\n",
					firstLocation.Value() );
		DC_Exit( EXIT_ERROR );
		return;
	}

	dagman.dag->DumpDotFile();

	if ( dagman.dumpRescueDag ) {
    	debug_printf( DEBUG_QUIET, "Dumping rescue DAG and exiting "
					"because of -DumpRescue flag\n" );
		dagman.dag->Rescue( dagman.primaryDagFile.Value(),
					dagman.multiDags, dagman.maxRescueDagNum, false,
					false, false );
		ExitSuccess();
		return;
	}

    //------------------------------------------------------------------------
    // Bootstrap and Recovery
    //
    // If the Lockfile exists, this indicates a premature termination
    // of a previous run of Dagman. If condor log is also present,
    // we run in recovery mode
  
    // If the Daglog is not present, then we do not run in recovery
    // mode
  
    {
      bool recovery = access(lockFileName,  F_OK) == 0;
      
        if (recovery) {
            debug_printf( DEBUG_VERBOSE, "Lock file %s detected, \n",
                           lockFileName);
			if (dagman.abortDuplicates) {
				if (util_check_lock_file(lockFileName) == 1) {
        			debug_printf( DEBUG_QUIET, "Aborting because it "
							"looks like another instance of DAGMan is "
							"currently running on this DAG; if that is "
							"not the case, delete the lock file (%s) "
							"and re-submit the DAG.\n", lockFileName );
					dagman.dag->GetJobstateLog().
								WriteDagmanFinished( EXIT_RESTART );
    				dagman.CleanUp();
					DC_Exit( EXIT_ERROR );
					// We should never get to here!
				}
			}
        }

			//
			// If this DAGMan continues, it should overwrite the lock
			// file if it exists.
			//
		util_create_lock_file(lockFileName, dagman.abortDuplicates);

        debug_printf( DEBUG_VERBOSE, "Bootstrapping...\n");
        if( !dagman.dag->Bootstrap( recovery ) ) {
            dagman.dag->PrintReadyQ( DEBUG_DEBUG_1 );
            debug_error( 1, DEBUG_QUIET, "ERROR while bootstrapping\n");
        }
    }

    debug_printf( DEBUG_VERBOSE, "Registering condor_event_timer...\n" );
    daemonCore->Register_Timer( 1, dagman.m_user_log_scan_interval, 
				condor_event_timer, "condor_event_timer" );

	dagman.dag->SetPendingNodeReportInterval(
				dagman.pendingReportInterval );
}