private static async Task MainAsync()
{
string poolId = "JobPrepReleaseSamplePool";
string jobId = "JobPrepReleaseSampleJob";
var settings = Config.LoadAccountSettings();
// Location of the file that the job tasks will work with, a text file in the
// node's "shared" directory.
string taskOutputFile = "$AZ_BATCH_NODE_SHARED_DIR/job_prep_and_release.txt";
// The job prep task will write the node ID to the text file in the shared directory
string jobPrepCmdLine = $@"/bin/bash -c ""echo $AZ_BATCH_NODE_ID tasks: > {taskOutputFile}""";
// Each task then echoes its ID to the same text file
string taskCmdLine = $@"/bin/bash -c ""echo $AZ_BATCH_TASK_ID >> {taskOutputFile}""";
// The job release task will then delete the text file from the shared directory
string jobReleaseCmdLine = $@"/bin/bash -c ""rm {taskOutputFile}""";
BatchSharedKeyCredentials cred = new BatchSharedKeyCredentials(settings.BatchServiceUrl, settings.BatchAccountName, settings.BatchAccountKey);
using (BatchClient batchClient = BatchClient.Open(cred))
{
var pool = await BatchUtils.CreatePoolIfNotExistAsync(batchClient, poolId);
var prepTask = new JobPreparationTask {
CommandLine = jobPrepCmdLine
};
var releaseTask = new JobReleaseTask {
CommandLine = jobReleaseCmdLine
};
var job = await BatchUtils.CreateJobIfNotExistAsync(batchClient, pool.Id, jobId, prepTask : prepTask, releaseTask : releaseTask);
// Create the tasks that the job will execute
List <CloudTask> tasks = new List <CloudTask>();
for (int i = 1; i <= 8; i++)
{
string taskId = "task" + i.ToString().PadLeft(3, '0');
string taskCommandLine = taskCmdLine;
CloudTask task = new CloudTask(taskId, taskCommandLine);
tasks.Add(task);
}
// Add the tasks in one API call as opposed to a separate AddTask call for each. Bulk task
// submission helps to ensure efficient underlying API calls to the Batch service.
Console.WriteLine("Submitting tasks and awaiting completion...");
await batchClient.JobOperations.AddTaskAsync(job.Id, tasks);
// Wait for the tasks to complete before proceeding. The long timeout here is to allow time
// for the nodes within the pool to be created and started if the pool had not yet been created.
await batchClient.Utilities.CreateTaskStateMonitor().WhenAll(
job.ListTasks(),
TaskState.Completed,
TimeSpan.FromMinutes(30));
Console.WriteLine("All tasks completed.");
Console.WriteLine();
// Print the contents of the shared text file modified by the job preparation and other tasks.
ODATADetailLevel nodeDetail = new ODATADetailLevel(selectClause: "id, state");
IPagedEnumerable <ComputeNode> nodes = batchClient.PoolOperations.ListComputeNodes(poolId, nodeDetail);
await nodes.ForEachAsync(async (node) =>
{
// Check to ensure that the node is Idle before attempting to pull the text file.
// If the pool was just created, there is a chance that another node completed all
// of the tasks prior to the other node(s) completing their startup procedure.
if (node.State == ComputeNodeState.Idle)
{
var files = await node.ListNodeFiles().ToListAsync();
NodeFile sharedTextFile = await node.GetNodeFileAsync("shared/job_prep_and_release.txt");
Console.WriteLine("Contents of {0} on {1}:", sharedTextFile.Path, node.Id);
Console.WriteLine("-------------------------------------------");
Console.WriteLine(await sharedTextFile.ReadAsStringAsync());
}
});
// Terminate the job to mark it as Completed; this will initiate the Job Release Task on any node
// that executed job tasks. Note that the Job Release Task is also executed when a job is deleted,
// thus you need not call Terminate if you typically delete your jobs upon task completion.
await batchClient.JobOperations.TerminateJobAsync(job.Id);
// Wait for the job to reach state "Completed." Note that this wait is not typically necessary in
// production code, but is done here to enable the checking of the release tasks exit code below.
await BatchUtils.WaitForJobToReachStateAsync(batchClient, job.Id, JobState.Completed, TimeSpan.FromMinutes(2));
// Print the exit codes of the prep and release tasks by obtaining their execution info
List <JobPreparationAndReleaseTaskExecutionInformation> prepReleaseInfo = await batchClient.JobOperations.ListJobPreparationAndReleaseTaskStatus(job.Id).ToListAsync();
foreach (JobPreparationAndReleaseTaskExecutionInformation info in prepReleaseInfo)
{
Console.WriteLine();
Console.WriteLine("{0}: ", info.ComputeNodeId);
// If no tasks were scheduled to run on the node, the JobPreparationTaskExecutionInformation will be null
if (info.JobPreparationTaskExecutionInformation != null)
{
Console.WriteLine(" Prep task exit code: {0}", info.JobPreparationTaskExecutionInformation.ExitCode);
}
// If no tasks were scheduled to run on the node, the JobReleaseTaskExecutionInformation will be null
if (info.JobReleaseTaskExecutionInformation != null)
{
Console.WriteLine(" Release task exit code: {0}", info.JobReleaseTaskExecutionInformation.ExitCode);
}
}
// Clean up the resources we've created in the Batch account
Console.WriteLine();
Console.WriteLine("Delete job? [yes] no");
string response = Console.ReadLine().ToLower();
if (response != "n" && response != "no")
{
// Note that deleting the job will execute the job release task if the job was not previously terminated
await batchClient.JobOperations.DeleteJobAsync(job.Id);
}
Console.WriteLine("Delete pool? [yes] no");
response = Console.ReadLine();
if (response != "n" && response != "no")
{
await batchClient.PoolOperations.DeletePoolAsync(poolId);
}
}
}
private static async Task MainAsync()
{
//const string nodeSize = "standard_a1_v2";
const string nodeSize = "standard_a4_v2";
const int nodeCount = 4;
const int maxTasksPerNode = 4; //can be set to 4* the number of cores on each node
const int taskCount = 32;
// Ensure there are enough tasks to help avoid hitting some timeout conditions below
int minimumTaskCount = nodeCount * maxTasksPerNode * 2;
if (taskCount < minimumTaskCount)
{
Console.WriteLine("You must specify at least two tasks per node core for this sample ({0} tasks in this configuration).", minimumTaskCount);
Console.WriteLine();
// Not enough tasks, exit the application
return;
}
// In this sample, the tasks simply ping localhost on the compute nodes; adjust these
// values to simulate variable task duration
const int minPings = 30;
const int maxPings = 60;
const string poolId = "ParallelTasksPoolDemo";
const string jobId = "ParallelTasksJob";
var settings = Config.LoadAccountSettings();
BatchSharedKeyCredentials cred = new BatchSharedKeyCredentials(settings.BatchServiceUrl, settings.BatchAccountName, settings.BatchAccountKey);
using (BatchClient batchClient = BatchClient.Open(cred))
{
var pool = await BatchUtils.CreatePoolIfNotExistAsync(
batchClient,
poolId,
nodeSize : nodeSize,
maxTasksPerNode : maxTasksPerNode,
schedulingPolicy : new TaskSchedulingPolicy(ComputeNodeFillType.Pack),
lowPriorityNodes : nodeCount);
CloudJob job = await BatchUtils.CreateJobIfNotExistAsync(batchClient, poolId, jobId);
// The job's tasks ping localhost a random number of times between minPings and maxPings.
// Adjust the minPings/maxPings values above to experiment with different task durations.
Random rand = new Random();
List <CloudTask> tasks = new List <CloudTask>();
for (int i = 1; i <= taskCount; i++)
{
string taskId = "task" + i.ToString().PadLeft(3, '0');
string taskCommandLine = "ping -c " + rand.Next(minPings, maxPings + 1).ToString() + " localhost";
CloudTask task = new CloudTask(taskId, taskCommandLine);
tasks.Add(task);
}
// Pause execution until the pool is steady and its compute nodes are ready to accept jobs.
// NOTE: Such a pause is not necessary within your own code. Tasks can be added to a job at any point and will be
// scheduled to execute on a compute node as soon any node has reached Idle state. Because the focus of this sample
// is the demonstration of running tasks in parallel on multiple compute nodes, we wait for all compute nodes to
// complete initialization and reach the Idle state in order to maximize the number of compute nodes available for
// parallelization.
await BatchUtils.WaitForPoolToReachStateAsync(batchClient, poolId, AllocationState.Steady, TimeSpan.FromMinutes(30));
await BatchUtils.WaitForNodesToReachStateAsync(batchClient, poolId, ComputeNodeState.Idle, TimeSpan.FromMinutes(30));
// Add the tasks in one API call as opposed to a separate AddTask call for each. Bulk task
// submission helps to ensure efficient underlying API calls to the Batch service.
Console.WriteLine("Submitting tasks and awaiting completion...");
await batchClient.JobOperations.AddTaskAsync(job.Id, tasks);
// Pause again to wait until *all* nodes are running tasks
await BatchUtils.WaitForNodesToReachStateAsync(batchClient, poolId, ComputeNodeState.Running, TimeSpan.FromMinutes(2));
Stopwatch stopwatch = Stopwatch.StartNew();
// Print out task assignment information.
Console.WriteLine();
await BatchUtils.PrintNodeTasksAsync(batchClient, poolId);
Console.WriteLine();
// Pause execution while we wait for all of the tasks to complete
Console.WriteLine("Waiting for task completion...");
Console.WriteLine();
try
{
await batchClient.Utilities.CreateTaskStateMonitor().WhenAll(
job.ListTasks(),
TaskState.Completed,
TimeSpan.FromMinutes(30));
}
catch (TimeoutException e)
{
Console.WriteLine(e.ToString());
}
stopwatch.Stop();
// Obtain the tasks, specifying a detail level to limit the number of properties returned for each task.
// If you have a large number of tasks, specifying a DetailLevel is extremely important in reducing the
// amount of data transferred, lowering your query response times in increasing performance.
ODATADetailLevel detail = new ODATADetailLevel(selectClause: "id,commandLine,nodeInfo,state");
IPagedEnumerable <CloudTask> allTasks = batchClient.JobOperations.ListTasks(job.Id, detail);
// Get a collection of the completed tasks sorted by the compute nodes on which they executed
List <CloudTask> completedTasks = allTasks
.Where(t => t.State == TaskState.Completed)
.OrderBy(t => t.ComputeNodeInformation.ComputeNodeId)
.ToList();
// Print the completed task information
Console.WriteLine();
Console.WriteLine("Completed tasks:");
string lastNodeId = string.Empty;
foreach (CloudTask task in completedTasks)
{
if (!string.Equals(lastNodeId, task.ComputeNodeInformation.ComputeNodeId))
{
Console.WriteLine();
Console.WriteLine(task.ComputeNodeInformation.ComputeNodeId);
}
lastNodeId = task.ComputeNodeInformation.ComputeNodeId;
Console.WriteLine("\t{0}: {1}", task.Id, task.CommandLine);
}
// Get a collection of the uncompleted tasks which may exist if the TaskMonitor timeout was hit
List <CloudTask> uncompletedTasks = allTasks
.Where(t => t.State != TaskState.Completed)
.OrderBy(t => t.Id)
.ToList();
// Print a list of uncompleted tasks, if any
Console.WriteLine();
Console.WriteLine("Uncompleted tasks:");
Console.WriteLine();
if (uncompletedTasks.Any())
{
foreach (CloudTask task in uncompletedTasks)
{
Console.WriteLine("\t{0}: {1}", task.Id, task.CommandLine);
}
}
else
{
Console.WriteLine("\t<none>");
}
// Print some summary information
Console.WriteLine();
Console.WriteLine(" Nodes: " + nodeCount);
Console.WriteLine(" Node size: " + nodeSize);
Console.WriteLine("Max tasks per node: " + pool.MaxTasksPerComputeNode);
Console.WriteLine(" Tasks: " + tasks.Count);
Console.WriteLine(" Duration: " + stopwatch.Elapsed);
Console.WriteLine();
Console.WriteLine("Done!");
Console.WriteLine();
// Clean up the resources we've created in the Batch account
Console.WriteLine();
Console.WriteLine("Delete job? [yes] no");
string response = Console.ReadLine().ToLower();
if (response != "n" && response != "no")
{
// Note that deleting the job will execute the job release task if the job was not previously terminated
await batchClient.JobOperations.DeleteJobAsync(job.Id);
}
Console.WriteLine("Delete pool? [yes] no");
response = Console.ReadLine();
if (response != "n" && response != "no")
{
await batchClient.PoolOperations.DeletePoolAsync(poolId);
}
}
}
private static async Task MainAsync()
{
const string poolId = "FileHandlingPool";
const string jobId = "FileHandlingJobDemo";
var settings = Config.LoadAccountSettings();
SetupStorage(settings.StorageAccountName, settings.StorageAccountKey);
BatchSharedKeyCredentials cred = new BatchSharedKeyCredentials(settings.BatchServiceUrl, settings.BatchAccountName, settings.BatchAccountKey);
using (BatchClient batchClient = BatchClient.Open(cred))
{
var pool = await BatchUtils.CreatePoolIfNotExistAsync(batchClient, poolId);
var job = await BatchUtils.CreateJobIfNotExistAsync(batchClient, poolId, jobId);
//set up auto storage file
ResourceFile autoStorageFile = ResourceFile.FromAutoStorageContainer(AutoStorageContainerName, AutoStorageFileName);
Console.WriteLine("\n[INFO] Autostorage resource File reference: ");
Console.WriteLine("AutoStorageContainer: " + autoStorageFile.AutoStorageContainerName);
Console.WriteLine("FilePath: " + autoStorageFile.FilePath);
//upload file to external storage and add it as a resource file
string storageConnectionString =
$"DefaultEndpointsProtocol=https;AccountName={settings.StorageAccountName};AccountKey={settings.StorageAccountKey}";
CloudStorageAccount storageAccount = CloudStorageAccount.Parse(storageConnectionString);
CloudBlobClient blobClient = storageAccount.CreateCloudBlobClient();
CloudBlobContainer externalContainer = blobClient.GetContainerReference(ExternalStorageContainerName);
await externalContainer.CreateIfNotExistsAsync();
var externalFile = await UploadFileToContainer(blobClient, ExternalStorageContainerName, "resource_files/resource_file.txt", "resource_file.txt");
Console.WriteLine("\n[INFO] External storage resource File reference:");
Console.WriteLine("SAS Url: " + externalFile.HttpUrl);
Console.WriteLine("FilePath: " + externalFile.FilePath);
// using staging files API
var filesToStage = new List <IFileStagingProvider>();
StagingStorageAccount fileStagingStorageAccount = new StagingStorageAccount(
storageAccount: settings.StorageAccountName,
storageAccountKey: settings.StorageAccountKey,
blobEndpoint: storageAccount.BlobEndpoint.ToString());
FileToStage stagedFile = new FileToStage("resource_files/staged_file.txt", fileStagingStorageAccount);
Console.WriteLine("\n[INFO] Staged File added:");
Console.WriteLine("Local File: " + stagedFile.LocalFileToStage);
Console.WriteLine("Node File: " + stagedFile.NodeFileName);
filesToStage.Add(stagedFile);
// setup output files
// Generate SAS for outputcontainer
CloudBlobContainer outputContainer = blobClient.GetContainerReference(OutputContainerName);
await outputContainer.CreateIfNotExistsAsync();
string containerSas = outputContainer.GetSharedAccessSignature(new SharedAccessBlobPolicy()
{
Permissions = SharedAccessBlobPermissions.Write,
SharedAccessExpiryTime = DateTimeOffset.UtcNow.AddDays(1)
});
string containerUrl = outputContainer.Uri.AbsoluteUri + containerSas;
Console.WriteLine("\n[INFO] Output container: " + containerUrl);
Console.WriteLine("\nPress return to continue...");
Console.ReadLine();
// Create tasks
List <CloudTask> tasks = new List <CloudTask>();
for (var i = 1; i <= 10; i++)
{
var taskId = i.ToString().PadLeft(3, '0');
var commandLine = $@"/bin/bash -c ""echo 'Hello from {taskId}' && printf 'root dir:\n' > output.txt && ls -la >> output.txt && printf '\ninput dir:\n' >> output.txt && ls -la input >> output.txt""";
var task = new CloudTask(taskId, commandLine);
// add resource files to task (one autostorage, one in external storage)
task.ResourceFiles = new[] { autoStorageFile, externalFile };
// add staged files
task.FilesToStage = filesToStage;
// add output files
var outputFiles = new List <OutputFile>
{
new OutputFile(
filePattern: @"../std*.txt",
destination: new OutputFileDestination(new OutputFileBlobContainerDestination(
containerUrl: containerUrl,
path: taskId)),
uploadOptions: new OutputFileUploadOptions(
uploadCondition: OutputFileUploadCondition.TaskCompletion)),
new OutputFile(
filePattern: @"output.txt",
destination: new OutputFileDestination(new OutputFileBlobContainerDestination(
containerUrl: containerUrl,
path: taskId + @"\output.txt")),
uploadOptions: new OutputFileUploadOptions(
uploadCondition: OutputFileUploadCondition.TaskCompletion)),
};
task.OutputFiles = outputFiles;
tasks.Add(task);
}
Console.WriteLine("Submitting tasks and awaiting completion...");
// Add all tasks to the job.
batchClient.JobOperations.AddTask(job.Id, tasks);
await BatchUtils.WaitForTasksAndPrintOutputAsync(batchClient, job.ListTasks(), TimeSpan.FromMinutes(30));
// Clean up Batch resources (if the user so chooses)
Console.WriteLine();
Console.Write("Delete job? [yes] no: ");
string response = Console.ReadLine().ToLower();
if (response != "n" && response != "no")
{
batchClient.JobOperations.DeleteJob(jobId);
}
Console.Write("Delete pool? [yes] no: ");
response = Console.ReadLine().ToLower();
if (response != "n" && response != "no")
{
batchClient.PoolOperations.DeletePool(poolId);
}
}
}
private static async Task MainAsync()
{
const int nodeCount = 1;
string poolId = "TaskDependenciesSamplePool";
string jobId = "TaskDependenciesJob";
var settings = Config.LoadAccountSettings();
BatchSharedKeyCredentials cred = new BatchSharedKeyCredentials(settings.BatchServiceUrl, settings.BatchAccountName, settings.BatchAccountKey);
using (BatchClient batchClient = BatchClient.Open(cred))
{
var pool = await BatchUtils.CreatePoolIfNotExistAsync(batchClient, poolId, lowPriorityNodes : nodeCount);
var job = await BatchUtils.CreateJobIfNotExistAsync(batchClient, pool.Id, jobId, usesTaskDependencies : true);
string taskOutputFile = "$AZ_BATCH_NODE_SHARED_DIR/task_output.txt";
// Create the collection of tasks that will be added to the job.
List <CloudTask> tasks = new List <CloudTask>
{
// 'Rain' and 'Sun' don't depend on any other tasks
new CloudTask("Rain", $"/bin/bash -c \"echo Rain >> {taskOutputFile}\""),
new CloudTask("Sun", $"/bin/bash -c \"echo Sun >> {taskOutputFile}\""),
// Task 'Flowers' depends on completion of both 'Rain' and 'Sun'
// before it is run.
new CloudTask("Flowers", $"/bin/bash -c \"echo Flowers >> {taskOutputFile}\"")
{
DependsOn = TaskDependencies.OnIds("Rain", "Sun")
},
// Tasks 1, 2, and 3 don't depend on any other tasks. Because
// we will be using them for a task range dependency, we must
// specify string representations of integers as their ids.
new CloudTask("1", $"/bin/bash -c \"echo 1 >> {taskOutputFile}\""),
new CloudTask("2", $"/bin/bash -c \"echo 2 >> {taskOutputFile}\""),
new CloudTask("3", $"/bin/bash -c \"echo 3 >> {taskOutputFile}\""),
// Task dependency on ID range
new CloudTask("Final", $"/bin/bash -c \"echo Final >> {taskOutputFile}\"")
{
DependsOn = TaskDependencies.OnIdRange(1, 3)
},
// Task A is the parent task.
new CloudTask("A", $"/bin/bash -c \"echo A >> {taskOutputFile}\"")
{
// Specify exit conditions for task A and their dependency actions.
ExitConditions = new ExitConditions
{
// If task A exits with a pre-processing error, block any downstream tasks (in this example, task B).
PreProcessingError = new ExitOptions
{
DependencyAction = DependencyAction.Block
},
// If task A exits with the specified error codes, block any downstream tasks (in this example, task B).
ExitCodes = new List <ExitCodeMapping>
{
new ExitCodeMapping(10, new ExitOptions()
{
DependencyAction = DependencyAction.Block
}),
new ExitCodeMapping(20, new ExitOptions()
{
DependencyAction = DependencyAction.Block
})
},
// If task A succeeds or fails with any other error, any downstream tasks become eligible to run
// (in this example, task B).
Default = new ExitOptions
{
DependencyAction = DependencyAction.Satisfy
},
}
},
// Task B depends on task A. Whether it becomes eligible to run depends on how task A exits.
new CloudTask("B", $"/bin/bash -c \"echo B >> {taskOutputFile}\"")
{
DependsOn = TaskDependencies.OnId("A")
},
};
// Add the tasks in one API call as opposed to a separate AddTask call for each. Bulk task
// submission helps to ensure efficient underlying API calls to the Batch service.
Console.WriteLine("Submitting tasks and awaiting completion...");
await batchClient.JobOperations.AddTaskAsync(job.Id, tasks);
// Wait for the tasks to complete before proceeding. The long timeout here is to allow time
// for the nodes within the pool to be created and started if the pool had not yet been created.
await batchClient.Utilities.CreateTaskStateMonitor().WhenAll(
job.ListTasks(),
TaskState.Completed,
TimeSpan.FromMinutes(30));
Console.WriteLine("All tasks completed.");
Console.WriteLine();
// Print the contents of the shared text file modified by the job preparation and other tasks.
ODATADetailLevel nodeDetail = new ODATADetailLevel(selectClause: "id, state");
IPagedEnumerable <ComputeNode> nodes = batchClient.PoolOperations.ListComputeNodes(poolId, nodeDetail);
await nodes.ForEachAsync(async (node) =>
{
// Check to ensure that the node is Idle before attempting to pull the text file.
// If the pool was just created, there is a chance that another node completed all
// of the tasks prior to the other node(s) completing their startup procedure.
if (node.State == ComputeNodeState.Idle)
{
NodeFile sharedTextFile = await node.GetNodeFileAsync("shared/task_output.txt");
Console.WriteLine("Contents of {0} on {1}:", sharedTextFile.Path, node.Id);
Console.WriteLine("-------------------------------------------");
Console.WriteLine(await sharedTextFile.ReadAsStringAsync());
}
});
// Clean up the resources we've created in the Batch account
Console.WriteLine();
Console.WriteLine("Delete job? [yes] no");
string response = Console.ReadLine().ToLower();
if (response != "n" && response != "no")
{
// Note that deleting the job will execute the job release task if the job was not previously terminated
await batchClient.JobOperations.DeleteJobAsync(job.Id);
}
Console.WriteLine("Delete pool? [yes] no");
response = Console.ReadLine();
if (response != "n" && response != "no")
{
await batchClient.PoolOperations.DeletePoolAsync(poolId);
}
}
}
