Task Management

Task Management is a core module of the FCP platform. It provides end-to-end lifecycle management for tasks, from submission and monitoring to managing results. With a unified console, users can efficiently manage all compute tasks running on the platform and ensure orderly execution and proper resource utilization.

Permissions

For details, see Permissions.

Task Modes Overview

The platform supports two core execution modes: Direct tasks and CWL tasks, to cover different stacks and scenarios. For more details, see Application Center Overview.

Submit a Task

1. Prerequisites

Before submitting a task, make sure you have:

• Permission to use the selected task template.
• Permission to use the target compute cluster.
• If you need to create a new cluster, you also need the Create Cluster permission.

2. Select a Task Template

1. On the task submission page, browse the task templates you are authorized to use.
2. Select a template based on your task mode (Direct or CWL).

3. Provide Parameters

Fill in runtime parameters

1. On the task parameters page, fill in the parameters required to run the task.
   • Required parameters: Must be provided, otherwise the task cannot be submitted.
   • Optional parameters: Provide as needed.
2. Some parameters may have default values or drop-down options. Fill in based on your actual requirements.

Upload input data files

If the template includes a file picker, click the data file selection area to choose the input files required by the task.

4. Select Resource Configuration

Based on your task requirements and permissions, select the cluster resources to run this task. You can select an existing cluster, a template-associated cluster, or create a new cluster in hybrid cloud scenarios. For details, see Select Resource Configuration below.

5. Task Options (CWL only)

• Log level: Controls verbosity of workflow logs. For example, "Errors/Warnings only" or "Debug details".
• Remove intermediate results: Controls whether intermediate outputs of subtasks are retained. Off: keep outputs. On: do not keep outputs.

6. Submit

1. After confirming all parameters are correct, click Submit.
2. The system will confirm submission and automatically navigate to the task list.

Task List

Field descriptions

The following fields may appear in the task list:

• Task ID: Auto-generated unique identifier used to locate a task precisely.
• Task Name: Human-friendly name assigned by the user.
• Cluster: Unique identifier of the compute cluster running the task. Click to open Cluster Management.
• Task Template: Name of the template used to submit the task.
• Status: Current stage, such as Queued, Running, Completed, etc.
• Priority: Scheduling priority. Higher value means higher priority.
• Subtasks/Total: Shows total subtasks and progress (count, percent, progress bar). (CWL only)
• Partition: The partition allocated for the task in the cluster. (Direct only)
• Execution Node: The compute node that actually runs the task. (Direct only)
• Working Directory: Directory on the node where task files are stored. (Direct only)
• Created At: Submission time.
• Started At: Actual start time.
• Finished At: Completion time.
• Queue Time: Time spent waiting in the queue (Started At - Created At). (Direct only)
• Total Duration: Overall duration (Finished At - Created At). (Direct only)
• Total CPU Cores: Total CPU cores used. (Direct only)
• GPU Count: Total GPUs used. (Direct only)
• Username: User who submitted and runs the task.

About time display: All time fields (Created At, Started At, Finished At) are automatically converted and displayed based on your local computer timezone.

Search and quick filters

To help you quickly locate tasks, the task list provides:

• Quick status tabs (Direct only): located at the top-right of the list.
  • Incomplete: shows tasks in non-terminal states such as Queued and Running.
  • Completed: shows tasks in terminal states such as Completed, Failed, or Canceled.
  • All: shows tasks in all states. This is the default view.
• Search: use the search box at the top of the list. You can search by Task ID, Task Name, Username, etc.

Remote connection methods

The platform supports multiple remote access protocols to help you quickly access compute nodes, confirm task execution status, inspect the runtime environment, and troubleshoot issues:

• VNC (Virtual Network Computing): Graphical remote desktop access. Useful for interactive and GUI-based workflows.
• SSH (Secure Shell): Secure command-line access. Useful for checking processes, viewing logs, managing files, and running diagnostics.
• RDP (Remote Desktop Protocol): Graphical remote desktop access for Windows environments.

Why it matters: When a task behaves unexpectedly, outputs are not as expected, or interactive debugging is required, you can connect directly to the execution node using the methods above to quickly locate and confirm the issue.

Task operations

On the Task Management page, you can perform the following operations:

Single-task operations:

• Cancel: Terminate a queued or running task and release allocated compute resources.
• Delete: Remove task records that are Completed, Failed, or Canceled from the list.
• Move to front: Move a queued task to the front of the queue so it can be scheduled first.
• Suspend: Pause a running task, temporarily releasing compute resources while keeping task state. (Direct only)
• Resume: Restart a suspended task and continue execution. (Direct only)
• Clone: Create a new task based on an existing task configuration to reuse parameters quickly.
• View details: Click the task name to view parameters, logs, resource usage, and more.

Bulk operations:

• Bulk cancel: Terminate multiple queued or running tasks.
• Bulk suspend: Pause multiple running tasks and temporarily release compute resources. (Direct only)
• Bulk delete: Clean up multiple Completed/Failed/Canceled task records.
• Bulk resume: Restart multiple suspended tasks. (Direct only)

Customize columns

• Customize columns: Select which columns to show in the list, so you can focus on the information you care about.

Task Details

Click any task in the list to open the task details page. The layout and focus differ between Direct tasks and CWL tasks.

Direct task details

Direct tasks typically include:

• Output files:
  • Result files generated by the task. Preview and download are supported.
  • The platform-generated sbatch script used to submit the task.
• Cluster monitoring: Resource usage charts during execution (CPU, memory, GPU, etc.). For details, see Cluster Monitoring.
• Runtime environment: All nodes in the cluster where the task runs. Fields include host name, role, instance type (cloud nodes only), hyper-threading (cloud nodes only), system volume (GiB), IP address, and node status.
• Input parameters: Configuration parameters used for execution, including input files and commands.
• Logs: Execution logs. Preview and download are supported.

CWL task details

CWL tasks typically include:

• Subtasks: All subtasks of the workflow.
  • Fields include task name, status, created time, start time, end time; supports searching by task name, status, start time, and end time.
  • The top of the page shows subtask counts by status such as Success, Failed, Canceled, and Submitted.
  • Click a subtask name to view its output files and logs.
• Subtask data: Output files and logs for subtasks.
• Runtime environment: All nodes in the cluster where the task runs, including host name, role, instance type (cloud nodes only), hyper-threading (cloud nodes only), system volume (GiB), IP address, and node status.
• Input parameters: Configuration parameters used for execution, including input files and commands.
• Logs: Execution logs. Preview and download are supported.

Notes:

1. For running tasks, logs and monitoring data are updated in real time.
2. For data isolation and privacy, each user can only view and download input/output files and logs generated by tasks they submitted and executed.

Select Resource Configuration

Select the cluster resources to run the task based on your requirements and permissions. The platform provides two options:

Feature	Select existing cluster	Create new cluster (hybrid cloud only)
How it works	Select from clusters you have permission to use.	Create a brand-new dedicated cloud cluster on demand.
Configuration	After selecting a cluster, adjust node count, partitions, and other parameters as needed.	Configure full parameters including instance type, hyper-threading, system volume (GiB), node count, and more.
Startup speed	Fast, no cluster creation needed.	Slower, typically requires several minutes to create the cluster.
Cost model	Fixed: uses existing resources, cost is predictable.	Pay-as-you-go: billed by actual usage time. After the task completes, cloud instances are released automatically to avoid idle cost.
Best for	Routine tasks, shared team resources, cost-sensitive workloads.	Resource shortage, burst workloads, special hardware/environment needs, strict isolation requirements.
Environment	Shared environment, potential contention.	Dedicated environment, better isolation and stability.
Permissions required	Requires Use permission for the target cluster.	Requires Create Cluster permission and hybrid cloud must be enabled.

Existing cluster notes

• Data source: Cluster and partition data are fetched in real time from the Fsched scheduler, ensuring accuracy and freshness.
• When selecting a cluster, you can view key metrics such as:
  • Default partition: Whether this partition is the default scheduling partition for the cluster.
  • Free CPU: CPU cores currently available for new tasks.
  • Running CPU: CPU cores currently used by running tasks.
  • Total CPU: Total CPU cores in the cluster/partition.
  • Free memory (GiB): Available memory for new tasks.
  • Allocated memory (GiB): Memory allocated to running tasks.
  • Total memory (GiB): Total memory in the cluster/partition.
  • Running compute nodes: Number of compute nodes currently running.
  • Idle compute nodes: Number of compute nodes currently idle and ready to accept tasks.

Recommendations

• Prefer existing clusters: For routine tasks, choose an existing cluster to start quickly and control costs.
• Consider a new cluster: Create a new cluster when existing resources are insufficient, special hardware is needed, or you want a dedicated isolated environment.
• Default preselection: If you have access to multiple clusters, the system randomly preselects a default partition from one cluster and shows its utilization. You can confirm or change it.
• Cost reminder: Creating a new cluster incurs additional cloud costs. Choose carefully based on estimates and budget.

Steps

1. In Select Resource Configuration, review the preselected default cluster (if any) and its resource status.
2. Decide whether to use an existing cluster or create a new cluster based on your needs.
3. If using an existing cluster, select the target cluster and partition.
4. If creating a new cluster, configure instance type, node count, storage, and other settings as guided.
5. Confirm the selection and proceed.