v2.2.0

Author: jmcave

cancelJobFcn.m

@@ -4,75 +4,8 @@
 % Set your cluster's PluginScriptsLocation to the parent folder of this
 % function to run it when you cancel a job.
 
-% Copyright 2010-2022 The MathWorks, Inc.
+% Copyright 2010-2023 The MathWorks, Inc.
 
-% Store the current filename for the errors, warnings and
-% dctSchedulerMessages
-currFilename = mfilename;
-if ~isa(cluster, 'parallel.Cluster')
-    error('parallelexamples:GenericGridEngine:SubmitFcnError', ...
-        'The function %s is for use with clusters created using the parcluster command.', currFilename)
-end
-
-% Get the information about the actual cluster used
-data = cluster.getJobClusterData(job);
-if isempty(data)
-    % This indicates that the job has not been submitted, so return true
-    dctSchedulerMessage(1, '%s: Job cluster data was empty for job with ID %d.', currFilename, job.ID);
-    OK = true;
-    return
-end
-
-% Get a simplified list of schedulerIDs to reduce the number of calls to
-% the scheduler.
-schedulerIDs = getSimplifiedSchedulerIDsForJob(job);
-erroredJobAndCauseStrings = cell(size(schedulerIDs));
-% Get the cluster to delete the job
-for ii = 1:length(schedulerIDs)
-    schedulerID = schedulerIDs{ii};
-    commandToRun = sprintf('qdel "%s"', schedulerID);
-    dctSchedulerMessage(4, '%s: Canceling job on cluster using command:\n\t%s.', currFilename, commandToRun);
-    try
-        [cmdFailed, cmdOut] = runSchedulerCommand(cluster, commandToRun);
-    catch err
-        cmdFailed = true;
-        cmdOut = err.message;
-    end
-    % If a job is already in a terminal state, qdel will return a failed
-    % failed error code and cmdOut will be of the form:
-    % 'denied: job "2979" does not exist'
-    % If this happens we do not consider the command to have failed.
-    if cmdFailed && ~contains(cmdOut, 'does not exist')
-        % Keep track of all jobs that errored when being cancelled, either
-        % through a bad exit code or if an error was thrown. We'll report
-        % these later on.
-        erroredJobAndCauseStrings{ii} = sprintf('Job ID: %s\tReason: %s', schedulerID, strtrim(cmdOut));
-        dctSchedulerMessage(1, '%s: Failed to cancel job %s on cluster.  Reason:\n\t%s', currFilename, schedulerID, cmdOut);
-    end
-end
-
-if ~cluster.HasSharedFilesystem
-    % Only stop mirroring if we are actually mirroring
-    remoteConnection = getRemoteConnection(cluster);
-    if remoteConnection.isJobUsingConnection(job.ID)
-        dctSchedulerMessage(5, '%s: Stopping the mirror for job %d.', currFilename, job.ID);
-        try
-            remoteConnection.stopMirrorForJob(job);
-        catch err
-            warning('parallelexamples:GenericGridEngine:FailedToStopMirrorForJob', ...
-                'Failed to stop the file mirroring for job %d.\nReason: %s', ...
-                job.ID, err.getReport);
-        end
-    end
-end
-
-% Now warn about those jobs that we failed to cancel.
-erroredJobAndCauseStrings = erroredJobAndCauseStrings(~cellfun(@isempty, erroredJobAndCauseStrings));
-if ~isempty(erroredJobAndCauseStrings)
-    warning('parallelexamples:GenericGridEngine:FailedToCancelJob', ...
-        'Failed to cancel the following jobs on the cluster:\n%s', ...
-        sprintf('  %s\n', erroredJobAndCauseStrings{:}));
-end
-OK = isempty(erroredJobAndCauseStrings);
+OK = cancelJobOnCluster(cluster, job);
 
 end

cancelTaskFcn.m

@@ -4,66 +4,8 @@
 % Set your cluster's PluginScriptsLocation to the parent folder of this
 % function to run it when you cancel a task.
 
-% Copyright 2020-2022 The MathWorks, Inc.
+% Copyright 2020-2023 The MathWorks, Inc.
 
-% Store the current filename for the errors, warnings and
-% dctSchedulerMessages
-currFilename = mfilename;
-if ~isa(cluster, 'parallel.Cluster')
-    error('parallelexamples:GenericGridEngine:SubmitFcnError', ...
-        'The function %s is for use with clusters created using the parcluster command.', currFilename)
-end
-
-% Get the information about the actual cluster used
-data = cluster.getJobClusterData(task.Parent);
-if isempty(data)
-    % This indicates that the parent job has not been submitted, so return true
-    dctSchedulerMessage(1, '%s: Job cluster data was empty for the parent job with ID %d.', currFilename, task.Parent.ID);
-    OK = true;
-    return
-end
-% We can't cancel a single task of a communicating job on the scheduler
-% without cancelling the entire job, so warn and return in this case
-if ~strcmpi(task.Parent.Type, 'independent')
-    OK = false;
-    warning('parallelexamples:GenericGridEngine:FailedToCancelTask', ...
-        'Unable to cancel a single task of a communicating job. If you want to cancel the entire job, use the cancel function on the job object instead.');
-    return
-end
-
-% Get the cluster to delete the task
-if verLessThan('matlab', '9.7') % schedulerID stored in job data
-    schedulerIDs = data.ClusterJobIDs;
-    schedulerID = schedulerIDs{task.ID};
-else % schedulerID on task since 19b
-    schedulerID = task.SchedulerID;
-end
-erroredTaskAndCauseString = '';
-commandToRun = sprintf('qdel "%s"', schedulerID);
-dctSchedulerMessage(4, '%s: Canceling task on cluster using command:\n\t%s.', currFilename, commandToRun);
-try
-    [cmdFailed, cmdOut] = runSchedulerCommand(cluster, commandToRun);
-catch err
-    cmdFailed = true;
-    cmdOut = err.message;
-end
-% If a job is already in a terminal state, qdel will return a failed
-% failed error code and cmdOut will be of the form:
-% 'denied: job "2979" does not exist'
-% If this happens we do not consider the command to have failed.
-if cmdFailed && ~contains(cmdOut, 'does not exist')
-    % Record if the task errored when being cancelled, either through a bad
-    % exit code or if an error was thrown. We'll report this as a warning.
-    erroredTaskAndCauseString = sprintf('Job ID: %s\tReason: %s', schedulerID, strtrim(cmdOut));
-    dctSchedulerMessage(1, '%s: Failed to cancel task %s on cluster.  Reason:\n\t%s', currFilename, schedulerID, cmdOut);
-end
-
-% Warn if task cancellation failed.
-OK = isempty(erroredTaskAndCauseString);
-if ~OK
-    warning('parallelexamples:GenericGridEngine:FailedToCancelTask', ...
-        'Failed to cancel the task on the cluster:\n  %s\n', ...
-        erroredTaskAndCauseString);
-end
+OK = cancelTaskOnCluster(cluster, task);
 
 end

communicatingSubmitFcn.m

@@ -25,14 +25,18 @@ function communicatingSubmitFcn(cluster, job, environmentProperties)
 
 % Get the correct quote and file separator for the Cluster OS.
 % This check is unnecessary in this file because we explicitly
-% checked that the ClusterOsType is unix. This code is an example
+% checked that the clusterOS is unix. This code is an example
 % of how to deal with clusters that can be unix or pc.
 if strcmpi(clusterOS, 'unix')
     quote = '''';
     fileSeparator = '/';
+    scriptExt = '.sh';
+    shellCmd = 'sh';
 else
     quote = '"';
     fileSeparator = '\';
+    scriptExt = '.bat';
+    shellCmd = 'cmd /c';
 end
 
 if isprop(cluster.AdditionalProperties, 'ClusterHost')
@@ -131,7 +135,7 @@ function communicatingSubmitFcn(cluster, job, environmentProperties)
 quotedLogFile = sprintf('%s%s%s', quote, logFile, quote);
 dctSchedulerMessage(5, '%s: Using %s as log file', currFilename, quotedLogFile);
 
-jobName = sprintf('Job%d', job.ID);
+jobName = sprintf('MATLAB_R%s_Job%d', version('-release'), job.ID);
 
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 %% CUSTOMIZATION MAY BE REQUIRED %%
@@ -144,29 +148,32 @@ function communicatingSubmitFcn(cluster, job, environmentProperties)
 end
 commonSubmitArgs = getCommonSubmitArgs(cluster);
 additionalSubmitArgs = strtrim(sprintf('%s %s', additionalSubmitArgs, commonSubmitArgs));
-
-% Extension to use for scripts
-scriptExt = '.sh';
+if validatedPropValue(cluster.AdditionalProperties, 'DisplaySubmitArgs', 'logical', false)
+    fprintf('Submit arguments: %s\n', additionalSubmitArgs);
+end
 
 % Path to the submit script, to submit the Grid Engine job using qsub
-localSubmitScriptPath = [tempname(localJobDirectory) scriptExt];
-[~, submitScriptName, submitScriptExt] = fileparts(localSubmitScriptPath);
-submitScriptPathOnCluster = sprintf('%s%s%s%s', jobDirectoryOnCluster, fileSeparator, submitScriptName, submitScriptExt);
+submitScriptName = sprintf('submitScript%s', scriptExt);
+localSubmitScriptPath = sprintf('%s%s%s', localJobDirectory, fileSeparator, submitScriptName);
+submitScriptPathOnCluster = sprintf('%s%s%s', jobDirectoryOnCluster, fileSeparator, submitScriptName);
 quotedSubmitScriptPathOnCluster = sprintf('%s%s%s', quote, submitScriptPathOnCluster, quote);
 
 % Path to the environment wrapper, which will set the environment variables
 % for the job then execute the job wrapper
-localEnvScriptPath = [tempname(localJobDirectory) scriptExt];
-[~, envScriptName, envScriptExt] = fileparts(localEnvScriptPath);
-envScriptPathOnCluster = sprintf('%s%s%s%s', jobDirectoryOnCluster, fileSeparator, envScriptName, envScriptExt);
+envScriptName = sprintf('environmentWrapper%s', scriptExt);
+localEnvScriptPath = sprintf('%s%s%s', localJobDirectory, fileSeparator, envScriptName);
+envScriptPathOnCluster = sprintf('%s%s%s', jobDirectoryOnCluster, fileSeparator, envScriptName);
 quotedEnvScriptPathOnCluster = sprintf('%s%s%s', quote, envScriptPathOnCluster, quote);
 
+% Create the scripts to submit a Grid Engine job.
+% These will be created in the job directory.
+dctSchedulerMessage(5, '%s: Generating scripts for job %d', currFilename, job.ID);
 createEnvironmentWrapper(localEnvScriptPath, quotedWrapperPath, variables);
 createSubmitScript(localSubmitScriptPath, jobName, quotedLogFile, ...
     quotedEnvScriptPathOnCluster, additionalSubmitArgs);
 
 % Create the command to run on the cluster
-commandToRun = sprintf('sh %s', quotedSubmitScriptPathOnCluster);
+commandToRun = sprintf('%s %s', shellCmd, quotedSubmitScriptPathOnCluster);
 
 if ~cluster.HasSharedFilesystem
     % Start the mirror to copy all the job files over to the cluster
@@ -239,7 +246,7 @@ function communicatingSubmitFcn(cluster, job, environmentProperties)
 if verLessThan('matlab', '9.7') % schedulerID stored in job data
     jobData.ClusterJobIDs = jobIDs;
 else % schedulerID on task since 19b
-    if numel(job.Tasks) == 1
+    if isscalar(job.Tasks)
         schedulerIDs = jobIDs{1};
     else
         schedulerIDs = repmat(jobIDs, size(job.Tasks));

deleteJobFcn.m

@@ -4,9 +4,9 @@ function deleteJobFcn(cluster, job)
 % Set your cluster's PluginScriptsLocation to the parent folder of this
 % function to run it when you delete a job.
 
-% Copyright 2017-2022 The MathWorks, Inc.
+% Copyright 2017-2023 The MathWorks, Inc.
 
-cancelJobFcn(cluster, job);
+cancelJobOnCluster(cluster, job);
 
 if cluster.HasSharedFilesystem
     % If we delete the job files before Grid Engine has actually finished
@@ -36,7 +36,9 @@ function deleteJobFcn(cluster, job)
         dctSchedulerMessage(4, '%s: Checking job does not exist on scheduler using command:\n\t%s.', currFilename, commandToRun);
         try
             [cmdFailed, ~] = runSchedulerCommand(cluster, commandToRun);
-        catch err %#ok<NASGU>
+        catch err
+            dctSchedulerMessage(5, '%s: Command ''%s'' failed:\n%s.', ...
+                currFilename, commandToRun, err.message);
             cmdFailed = true;
         end
 

deleteTaskFcn.m

@@ -1,11 +1,11 @@
 function deleteTaskFcn(cluster, task)
-%DELETEJOBFCN Deletes a job on cluster
+%DELETETASKFCN Deletes a job on Grid Engine
 %
 % Set your cluster's PluginScriptsLocation to the parent folder of this
 % function to run it when you delete a job.
 
-% Copyright 2020-2022 The MathWorks, Inc.
+% Copyright 2020-2023 The MathWorks, Inc.
 
-cancelTaskFcn(cluster, task);
+cancelTaskOnCluster(cluster, task);
 
 end