Data Observability Overview¶
Go to Data Observability > Overview to view run results and change trends. After identifying failed runs, you can go to the corresponding job to view details.
The Overview counts runs that have reached an ended state. Runs that have not yet reached an ended state can be viewed on the Jobs page.
Run Statistics¶
| Metric | Description |
|---|---|
| Run count | Total number of ended runs, including successful, failed, and terminated runs. |
| Failed runs | Number of failed and terminated runs. |
| Failed jobs | Number of jobs with failed or terminated runs. |
| Success rate | Ratio of successful runs to the total number of ended runs. The metric is unavailable when there are no ended runs. |
For example, if the selected range contains 8 successful runs, 1 failed run, 1 terminated run, and 2 runs in progress, the Overview shows a run count of 10, failed runs of 2, and a success rate of 80%.
The run count on the Jobs page also includes running and unknown states, so it may differ from the Overview. For details, see Job Run Statistics.
Run Trends¶
Run Count Trend¶
Shows changes in successful and failed runs over time, where the failed statistics include terminated runs. You can use the trend chart to identify time periods with concentrated failures.
Run Duration Trend¶
Shows the P50 and P95 duration curves, in minutes:
- P50: Median duration, used to observe typical run duration.
- P95: 95th percentile duration, used to observe changes in slower runs.
Duration curves are suitable for identifying time periods that require further investigation. For the start time, end time, and duration of specific runs, check the Run History.
Recent Failed Runs¶
The list shows recent failed runs in descending order by end time, including job, time, status, duration, and error summary.
Click a run record to go to the corresponding job and continue viewing the run history. Click View Job to open the job module, then filter by status and time to locate the target.
The error summary requires error information to be collected. When error details are not displayed, check the related collection configuration.
Troubleshooting Suggestions¶
- Observe the run trends to locate time periods with concentrated failures.
- Open the job to which the failed run belongs and select the corresponding record in the run history.
- Verify the error information, run environment, and inputs and outputs.
- To trace the source or check dependencies, go to Data Lineage. Whether downstream objects are actually affected must be confirmed with their run records.