Load tasks¶

A load task is a background job that continues to run after data loading is created. It records the data source, destination Catalog location, synchronization policy, runtime resources, and execution history. Load tasks let you view progress, adjust editable settings, backfill historical data, and pause, resume, or stop synchronization.

List and filters¶

Go to Data Connection > Data Loading to view the load-task list. Click Load data to create a task. The list generally supports searching, filtering, sorting, and pagination.

Column	Description
Task name	Identifies the task and its purpose
Data source	Source connector, local upload, web collection, or another source
Load type	Structured or unstructured
Load mode	One-time, periodic, or real-time
Destination	Directory, database, table, or volume in Catalog
Status	Current scheduling or execution state
Creation time	Can be sorted by time
Actions	Details, edit, delete, and task runtime controls

Search by task name, data source, or destination. Filters can combine load type, mode, and status. If the UI retains the original search and filter conditions when you return to the list after editing a task, continue verifying the result in that context.

Status and details¶

Status	Meaning	Recommendation
Waiting	Created and awaiting scheduling or a trigger	Check the schedule, resources, and dependencies
Loading	Reading, processing, or writing data	Review runtime, progress, and source-system load
Backfilling	Filling in historical data	Review progress percentage, batches, and execution window
Paused	Manually paused with progress and watermark retained	Confirm the reason, then resume or adjust settings
Load failed	The current execution encountered an error	Open details to review the error and troubleshoot by source

Click Details to view task configuration and runtime information, commonly including:

Basic information: connector, data type, load mode, destination, schedule, creation time, latest run, and cumulative statistics.
Synchronization policy: incremental field, lookback window, or full-overwrite method.
Data backfill: start time, progress, batches, and execution window.
Connection details: the connector actually referenced by the task and its source type.
Destination data: table schema for structured tasks; recent files, sizes, modification times, and processing status for unstructured tasks.
Error information: failure cause and troubleshooting clues.

Editable scope¶

Click Edit in the list or details to open the reused load-configuration page. After task creation, choices that affect data structure and identity are locked to prevent an existing task from becoming a different type of task.

Usually editable	Usually locked
Load mode and frequency	Data type (structured / unstructured)
Synchronization policy, incremental field, and lookback window	Load method (connector / local upload / web collection)
Backfill parameters and preprocessing scripts	Connector, source database table, or source object
Decompression, deduplication, and scope filters	Destination Catalog location and existing-table / new-table mode
Compute resources and runtime priority	—

Changing the incremental field resets the watermark, so the next load may reread historical ranges. Before saving, confirm that values in the new field are stable and sortable and that the source permits rereading the corresponding data.

Scheduling and runtime controls¶

Action	Behavior
Pause	Suspend the task while retaining the watermark and completed backfill progress
Resume	Resume from the paused position; a backfill task generally resumes from the last completed batch
Stop	Abort the current execution; whether it is scheduled again depends on the current task configuration
Delete	Delete the task and its schedule; first confirm that no downstream process still depends on updates to this data

A one-time load task writes once; a periodic task runs at the configured interval; a real-time task reads source changes continuously or near real time. To reduce source pressure while a task is running, lower its priority, narrow its scope, increase the interval, or run backfills during off-peak windows.

Data backfill¶

Real-time or incremental synchronization generally begins after the current watermark. Data backfill fills in earlier historical data: the system reads in batches from a specified starting point and automatically returns to normal synchronization after catching up to current data.

Setting	Purpose
Backfill start time	Start from the earliest available data or from a business-required historical point
Batch size	Time span read each time; use smaller batches when source pressure is high
Batch interval	Wait between consecutive batches to avoid continuously occupying the production system
Execution window	Limit execution to nighttime or other off-peak periods to reduce business impact

Backfill can be configured when creating a task or triggered from its details. After a pause or failure, check source data, networking, and the incremental field before resuming from retained progress. A full-overwrite policy rereads the entire table on every run, so it generally does not require a separate backfill.

Troubleshooting failures¶

Review the error, latest run time, and destination in the task details.
Return to Connectors and run a connection test to confirm that the network, credentials, and source-side authorization have not changed.
Check whether the source object still exists: whether the file path, database/table, collection, topic, queue, endpoint, or email scope has changed.
Verify destination Catalog table or volume permissions, schema, primary-key conflict policy, and available compute resources.
For incremental tasks, check the watermark, incremental field, and lookback window; redesign the backfill scope if necessary.