Operations management is a tool provided by this product that helps users efficiently manage and maintain various tasks. This document introduces the specific features of operations management.

Please ensure that your account has operations or management permissions.

This feature is supported in the on-premises deployment mode starting from version V2.64.0 of this product.

Users can utilize features such as the task dashboard, operations support, intelligent diagnostics, and resource governance to monitor task progress and status in real time, standardize management of task processes, and quickly identify and resolve issues encountered during task operations.
Operations management currently includes three main features: task dashboard, task operations, and resource governance.

• Task dashboard: Serving as the "data cockpit" of operations management, it provides real-time visual displays of task status and insights into historical data, helping users intuitively grasp the overall situation of tasks.
• Task operations: Focused on controlling the task execution process, it supports multi-dimensional queries of task instances, custom batch reruns, termination of abnormal tasks, and other operations. It is the core tool for handling task exceptions and restoring task execution.
• Resource governance: Identifies and cleans up idle data resources (datasets, Excel/CSV files) to prevent storage resource waste and ensure efficient utilization of system resources.

1. Please log in via the BytePlus Console (Portal) login page. You need to ensure that your account has operations or management permissions.

   

2. Task dashboard, task operations, or resource governance. After logging in, you can navigate to the operations management page by clicking System Management in the upper right corner of the page.
   

Today's task overview #

Today's task overview page includes the following information:

• Statistics of task instances by status for today
• Today's task summary (hourly statistics)

Historical task insights #

The historical task insights page contains the following information:

• Details of yesterday's tasks
  • Number of tasks triggered in the past 48 hours
  • Top 20 tasks by duration yesterday: This section displays required metadata, such as dataset/visual modeling task name, dataset/visual modeling task ID, owner, project ID, duration of this task instance, output data volume of this instance, and the corresponding average task duration.
  • Top 20 tasks by output data volume yesterday: This section displays required metadata, such as dataset/visual modeling task name, dataset/visual modeling task ID, owner, project ID, output database table name, TTL of the output base table, and output data volume of this instance.
• Task insights for the past two weeks (up to one month can be selected)
  • Statistics of successful tasks
  • Statistics of task instance runtime duration
  • Statistics of data source synchronization task runtime duration

• Task operations and maintenance include multi-dimensional queries, rerunning task instances, and terminating task instances.
• Abnormal task instances are highlighted: consecutive failures are shown with a red background, and slow tasks are shown with a blue background.
• Task operations and maintenance support custom batch reruns of tasks, typically used for cluster migration or missing instances.

Explanation of search criteria:

Idle dataset management #

The idle dataset management page contains the following information:

• You can select "project name" (multi-selection is supported in privatized version V2.91.0 and above), enter "idle days", and select "recent synchronization time" to locate the target idle dataset.
• You can batch pause synchronization and batch delete. Batch deletion moves items to the recycle bin, and recovery is supported via the recycle bin.

Idle Excel/CSV file management #

The idle Excel/CSV file management page contains the following information:

• You can select "project name" (multi-selection is supported in privatized version V2.91.0 and above), enter "idle days", and select "recent synchronization time" to locate the target idle Excel/CSV file.
• Support multi-select results and batch deletion of idle Excel/CSV files that the user has permission to access.

Troubleshooting #

(1) View logs after receiving an alert
First, go to the task board page and click the failed card. The system will automatically redirect to the task operations page and search for failed instances.
Then, select the failed instance and click Log to view the detailed log of the failed task.

(2) Determine whether the alert is caused by a large task backlog
First, on the task board page, click the running card, and then check the resources used by the running instance.
If the resource usage accounts for a large proportion of the entire queue, check the execution duration of the instance. If the duration is long, it indicates that the alert is caused by this task instance, and you need to handle and adjust the task model.

(3) Determine whether the alert is caused by excessive backtracking
If there are many backtracking instances on your task board and the number of slow tasks is zero, it can be confirmed that the alert is caused by excessive backtracking.
In this case, on the task operations page, select [Backtracking] as the trigger method and [Queued] as the instance status to find the details of the specific backtracking task instances.
(4) View detailed information of running tasks
To view detailed information of running tasks, such as dataset, project, owner, and more, directly click the running card on the task board to view the detailed metadata of the running instance. You can also click the [Jump] button to go directly to the dataset or visual modeling task page, and view the task operation log by clicking the [Log] button.

Task recovery #

(1) Supplement instance records and rerun tasks
When instance records are missing or the cluster is migrated, you need to supplement the instance records and rerun the tasks. This can be done through custom rerun.

(2) Recover projects by priority
When a task fails and it is necessary to restore the tasks of project 1 with high priority before restoring project 2, you can complete this operation by following the steps below.
Step 1: Click Custom Rerun, select the business date, set the rerun type to Project, enter 1 for Id, select [high-quality] for Priority, and then submit.

Step 2: Click Custom Rerun, select the business date, set the type to Project, select project 2 for Id, select [Ordinary] for Priority, and then submit.

(3) Batch rerun
To perform a batch rerun for specific datasets or visual modeling tasks. First, click the [Custom Rerun] button, select the business date, and set the type to Dataset or Visual Modeling Task. Then, enter the batch Ids separated by English commas in the Id column, and finally click OK.

(4) Terminate backtracking tasks
When task resources are limited, you may need to prioritize routine tasks and terminate all backtracking tasks at once. You can complete this operation by following the steps below.
Step 1: For Task Maintenance, set the trigger mode to [Backtracking], set the instance status to [Queued], select all instances, and click Terminate.
Step 2: For Task Maintenance, set the trigger mode to [Backtracking], set the instance status to [Running], select all instances, and click Terminate.

(5) View the time for queued consumption tasks
When task queuing is normal, you can view the [Queued] card on the Task Dashboard page, which displays the estimated time.

(6) Rerun all failed tasks
First, click the Run Failed button on the Task Dashboard page. The system will redirect to the Task Maintenance page and display all failed tasks for today.
Next, select all failed tasks, and then choose Batch Rerun.

Clean up tasks #

Scenario: Clean up datasets that have been idle for more than 90 days when disk space is limited.
First, on the Idle Datasets page, enter 90 for Idle Days, and then filter the results. Second, after selecting all, you can choose to pause synchronization.
Finally, after refreshing the page, the scheduling type of all selected tasks on the previous page has already been changed to manual.