View job information with LogView

This section describes the components of the LogView web UI.

The LogView home page is divided into two sections:

• Instance information

• Task information

The top section of the LogView page displays information about the MaxCompute instance for the submitted SQL task. This information includes the URL, Project, InstanceID, Owner, StartTime, EndTime, and Status.

• In the Instance information section, click the status to view details about the current queue. The status can be one of the following:

  • Waiting: The job is being processed by MaxCompute but has not yet been submitted to the Fuxi job scheduler.

  • Waiting List: n: The job is in the Fuxi job scheduler queue. 'n' indicates its position in the queue.

  • Running: The job is currently running in the Fuxi job scheduler.

  Note

  For a terminated instance, queue information is unavailable because the job has finished.

• Click the status to view the following detailed queue information:

  • Sub Status: The current sub-status.

  • WaitPos: The position in the queue. A value of 0 indicates that the task is running, and a value of - indicates that the task has not yet been submitted to Fuxi.

  • QueueLength: The total length of the queue in Fuxi.

  • Total Priority: The runtime priority assigned to the job.

  • SubStatus History: Click to view a detailed history of the job's execution statuses. This includes the status code, description, start time, and duration. This information may not be available in all versions.

After a job completes, you can view its execution result. For example, the following figure shows the result of a SELECT SQL statement.

Click Detail to view the operational details of a task, both during and after its execution. In the MaxCompute Task details dialog box, you can find the following information:

• A MaxCompute task can consist of one or more Fuxi jobs. For example, if you submit a complex SQL task, MaxCompute automatically submits multiple Fuxi jobs to the Fuxi job scheduler.

• Each Fuxi job can consist of one or more Fuxi tasks. For example, a simple MapReduce job typically generates two Fuxi tasks, such as a Map Task (M1) and a Reduce Task (R2), while a more complex SQL task can generate several.

• A Fuxi task name is typically composed of letters and numbers. The letter indicates the task type (for example, M for Map Task), and the numbers indicate the task ID and its dependencies. For example, R5_4 indicates that this Reduce Task starts only after task J4 has completed. J4_1_2_3 indicates that this Join Task starts only after tasks M1, M2, and M3 have all completed.

  Note

  I/O Records indicates the number of input and output records for the Fuxi task.

To view the instance information for a Fuxi task, click the button in the view column or double-click the task.

In the Fuxi instance information section at the bottom of the page, LogView groups instances by their stage. For example, you can select the Failed group to view nodes that encountered errors. You can also click the StdOut and StdErr buttons to view the corresponding standard output and standard error.

• Tasks with errors

  When a task fails, the error message appears in the Result section. Alternatively, go to the Detail page and check the StdErr of a specific Fuxi instance to find detailed error information.

• Data skew

  Slow execution can sometimes be caused by a "long tail," where a few Fuxi instances in a Fuxi task take much longer to complete than others. This long-tail phenomenon is caused by uneven workload distribution among instances of the same task. After the task completes, you can check the run results on the Summary tab. For each task, you will see output similar to the following:

  output records:
  R2_1_Stg1: 199998999  (min: 22552459, max: 177446540, avg: 99999499)

  A large difference between the min and max values, especially an order-of-magnitude difference, indicates data skew at this stage. For example, when performing a Join operation on a field, if one value appears with a much higher frequency than others, performing the Join on this field results in data skew.