Logview is a tool that you can use to view and debug jobs after you submit jobs to MaxCompute.

You can use Logview to view the following information of a job:
  • Running status of tasks
  • Running results of tasks
  • Details of each task and the progress of each step
After a job is submitted to MaxCompute, the system generates a Logview URL. You can enter the Logview URL in the search engine and press Enter to view job information. Screenshots
Note The Logview page of each job is valid for seven days.

UI elements

This section describes elements on the Logview UI.
The Logview page consists of two sections:
  • ODPS Instance
  • ODPS Tasks

ODPS Instance

The ODPS Instance section shows the information of the instance that corresponds to the submitted SQL job. The information includes URL, Project, InstanceID, Owner, StartTime, EndTime, and Status.
  • You can click the value in the Status column to view queue information. Valid values of Status:
    • Waiting: The job is being processed in MaxCompute and has not been submitted to Job Scheduler.
    • Waiting List: n: The job has been submitted to Job Scheduler and is queued. n indicates the order number of the job in the queue.
    • Running: The job is running in Job Scheduler.
    Note If the value of Status is Terminated, the job has been terminated and no queue information is available.
  • After you click the value of Status, the following queue information is displayed:
    • Sub Status: the current sub-status information.
    • WaitPos: the position of the job in the queue. The value 0 indicates that the job is running. The value - indicates that the job has not been submitted to Job Scheduler.
    • QueueLength: the total queue length in Job Scheduler.
    • Total Priority: the running priority assigned by the system.
    • SubStatus History: You can click the icon in this column to view the status history. This includes the status code, status description, start time, and duration of a state. The information is unavailable in some versions.

ODPS Tasks

The ODPS Tasks section shows the information of the task that corresponds to the instance. The information includes Name, Type, Status, Result, Detail, StartTime, EndTime, Latency (s), and TimeLine. Like on other pages, Latency (s) indicates the running duration.

Result:

After a job is completed, you can click the icon in the Result column to view running results. The following figure shows the execution results of a SELECT statement.

Detail:

You can click the icon in the Detail column to view running details. You can view the details of both running and completed jobs.In the dialog box that shows the details of a MaxCompute task, you can view the following information:
  • A MaxCompute job consists of one or more Fuxi jobs. For example, if a complex SQL job is submitted, MaxCompute automatically submits multiple Fuxi jobs to Job Scheduler.
  • Each Fuxi job consists of one or more Fuxi tasks. For example, a simple MapReduce job generates two Fuxi tasks: Map Task (M1) and Reduce Task (R2). If an SQL statement is complex, multiple Fuxi tasks may be generated.
  • The name of a Fuxi task. In most cases, a task name consists of letters and digits. A letter represents the type of a task, such as M for a map task. The digits that follow the letter represents the ID and dependencies of a task. For example, R5_4 indicates that the reduce task can be executed only after the J4 task is completed. J4_1_2_3 indicates that the join task can be executed only after the M1, M2, and M3 tasks are completed.
    Note I/O Records indicates the numbers of the input and output records of a Fuxi task.
  • I/O Records indicates the numbers of the input and output records of this task.
To view the information of the instance that corresponds to a Fuxi task, you can click the icon in the Show Detail column of the Fuxi task or double-click the Fuxi task.
Note Each Fuxi task consists of one or more Fuxi instances. If the amount of input data for a Fuxi task increases, MaxCompute activates more nodes for the task to process the data. A node is equivalent to a Fuxi instance.
In the lower part of the dialog box, the status information about Fuxi instances at different stages is displayed in groups. For example, you can click the Failed tab to view the nodes where errors occurred. You can click the icon in the StdOut column to view standard output information. You can also click the icon in the StdErr column to view standard error information.
Note To-be-displayed information written in the submitted MaxCompute task is also displayed in the standard output information and standard error information.

Use Logview to handle issues

  • Tasks with errors

    If an error occurs during a task, find the task and click the icon in the Result column in the lower part of the Logview page to view the error information. You can also click the icon in the StdErr column of a Fuxi instance in the details dialog box to view the error information of the instance.

  • Data skew
    The long tail of one or more Fuxi instances may decelerate the execution of a Fuxi task. The long tail is caused by uneven data distribution in a task. After a task is completed, you can view the running results on the Summary tab of the details dialog box. The following output provides an example of the running results.
    output records:
    R2_1_Stg1: 199998999  (min: 22552459, max: 177446540, avg: 99999499)

    If the difference between the min and max values is large, data skew occurs. For example, if a specific value appears more often than other values in a column, data skew occurs when you execute a JOIN operation based on this column.