Docs

Torch Log Service

Scheduled SQL tasks FAQ

How to handle scheduled SQL tasks that have not started scheduling

Why has the scheduled SQL analysis task not started scheduling for an extended period?

The possible reasons why a scheduled SQL analysis task has not started scheduling for a long time are as follows:

Scheduling cycle not met: A scheduled SQL analysis task will only generate a scheduling instance after the scheduled time point is reached.

Scheduling delay: If a previous instance is delayed, the newly generated scheduling instance will wait in the scheduling queue.

Excessive scheduling instances: Too many concurrent scheduling instances can cause a backlog in the execution queue, resulting in scheduling instances remaining unexecuted for a long time.

Index not built: The execution has actually completed, but the index for the result data has not been built, causing the execution result generated by scheduling to not be displayed.

Why does the scheduling instance of the scheduled SQL task remain unexecuted?

Before executing a scheduling instance, TLS checks whether the instance meets the execution conditions. If the previous instance has not run because the data within the query time range has not been fully written, and the next instance already meets the execution conditions, the next instance will be executed first.

Tip

After the instance meets the execution conditions, TLS will schedule resources and prepare the execution instance in the backend. This process may take more than one minute. Please wait patiently.

Generally, the following conditions must be met for an instance to be executed:

The scheduled time has been reached. If no delay time is set for the task, the backend service will execute the instance according to the scheduled time. If a delay time is set, the instance will be executed after a certain delay following the scheduled time.

The log data corresponding to the SQL time window of the instance has been fully written. If the log topic already contains log data later than the query time range, it is determined that all data within the query time range has been fully written. For example, if the query time range is [10:00~11:00), and a log entry with a timestamp of 11:01 is written, it is determined that all data within the query time range has been fully written, and the instance can start running.

The maximum concurrency has not been reached. At most, only one scheduling instance can be in the running state within a task.

Last updated: 2026.06.25 17:47:42