MaxCompute:Job instance details

Basic information

The properties bar shows the job ID, creator, status, start and end times, computing resource usage, and input and output tables. For running jobs, this information refreshes automatically on a scheduled basis.

Double-click a table name to view basic information about that table.

Execution view

The Execution view displays job dependencies as a directed acyclic graph (DAG) across three layers: Fuxi Job, Fuxi Task, and Operation. The Fuxi Task layer is shown by default.

The numbered elements in the view:

Fuxi Job layer

Click MaxCompute job to switch to the Fuxi Job layer. Each worker in this layer shows the name, start time, and end time of a Fuxi task.

Fuxi Task layer

Double-click a Fuxi job worker to go to the Fuxi Task layer. If multiple Fuxi jobs exist, the Fuxi Task layer of the last Fuxi job is shown by default.

This layer displays task dependencies, input tables, output tables, and partitions. Each Fuxi task worker shows:

• Instance count — in a/b/c format, where a is the number of running instances, b is the number of completed instances, and c is the total.

• I/O records — input and output record counts at the current point in time.

• Progress bar — an orange bar with a percentage, derived from the analysis of running instances.

• Connecting lines — each line shows the number of output records; the arrow indicates data flow direction.

After a job completes, open the Progress Chart drop-down list to overlay a heat map on the DAG. Available options:

• Input Heat Chart (Record)

• Output Heat Chart (Record)

• Input Heat Chart (Byte)

• Output Heat Chart (Byte)

• TaskTime Heat Chart

• InstanceTime Heat Chart

The heat map uses different colors to distinguish between the heat sources of different workers.

Operation layer

Double-click a Fuxi task to go to the Operation layer, which shows how the Fuxi task runs internally. Click a worker to view all Operation layer details.

For non-SQL jobs, only the Fuxi Job and Fuxi Task layers are available. The Operation layer is not supported.

Job playback

Job playback lets you replay the execution history of a completed job in 12 seconds, so you can see how subtasks progressed over time, determine their order and relative duration, and identify the critical path.

Click >

Playback estimates I/O data based on execution time — it does not reflect the actual I/O volume at each point in time. Jobs in the Running state do not support playback.

Timeline view

The Timeline view renders the distributed execution of a job as a Gantt chart. Adjust the display granularity to show all compute workers on the chart. Use the available filters to surface:

• Critical path

• Workers that processed the largest volume of data

• Workers with the longest execution time

The Gantt chart makes time bottlenecks and long-tail workers immediately visible across the job.

Analysis view

The Analysis view shows long-tail workers and workers with data skew, plotted as scatter charts and column charts. Click any worker in a chart to open its details page and inspect the underlying Fuxi instance metrics.

Details view

The Details view shows Fuxi task lists and compute worker lists for SQL data manipulation language (DML) jobs. Each job maps to one or more Fuxi jobs. Each Fuxi job is divided into Fuxi tasks (stages), and each Fuxi task contains multiple Fuxi instances (workers).

The view contains:

1. The Fuxi job tab

2. The Fuxi task list

3. The details and compute worker list for each Fuxi task

Right-click a Fuxi instance to view its standard output, standard errors, and debugging information.