Dataphin:Instance runtime diagnostics

Limits

• Runtime diagnostics is supported only for offline recurring instances and data backfill instances, such as script, detail and aggregate table, and extraction instances. It is not supported for real-time instances (such as real-time computing and real-time integration) or one-time instances.

• Field-level analysis is not supported for detail and aggregate table instances. Analysis is supported only at the materialization node level.

Function overview

In the Operation Center, different colors and icons indicate the status of an instance and its stage in the execution flow. You can use these visual cues to determine an instance's stage or to find out why it is not running. The instance statuses and the execution flow are as follows:

The successful execution of an instance depends on multiple factors, such as its upstream dependencies, scheduled time, resources, and its own runtime status. If an instance fails or remains in one status for a long time, you can use the runtime diagnostics feature to analyze it based on the following flow or dimensions:

Access runtime diagnostics

1. On the Dataphin home page, choose Develop > Task O&M from the top menu bar.

2. In the navigation pane on the left, choose Instance O&M > Recurring Instance/Data Backfill Instance/One-time Instance.

3. On the Recurring Instance/Data Backfill Instance/One-time Instance page, click the name of the target instance. Below the directed acyclic graph (DAG) on the right, click View Node Details.

   The following figure uses the Recurring Instance page as an example.

4. On the node details page, click the Diagnosis tab.

Upstream dependencies

The upstream dependency diagnostics shows the result of the last diagnostic run for the instance and the latest status of its upstream instances. The check proceeds to the next step only if all upstream instances have run successfully. You can view the upstream dependency diagnostics results to further investigate the cause of a failure. An instance that passed the upstream dependency diagnostics in its last run is not diagnosed again. To refresh the last run result or the status of upstream instances, click the refresh icon.

• If the result of the last run of the instance was Succeeded and it was not a forced rerun, the diagnostic result is Pass.

  Feature	Description
  Last run	Shows the run status and the time it passed. Note The current instance is scheduled only after all upstream instances run successfully.
  Current diagnostic result	Shows the diagnostic result. Schedule Type: Includes Dry Run, Normal Run, and Paused Run. If the current instance is in the Paused Run state, you must Resume Scheduling before it can start. Initial blocking node: Shows the top-level node that prevents the current node from running. Instances that pass the upstream diagnostics do not have an initial blocking node. Direct upstream list: Shows the list of direct upstream nodes. You can search by node name, node ID, or instance ID. You can also filter by Run Status and Owner.

• If the instance has not started and is not paused, the schedule type in the diagnostic result is Normal Run. Follow the prompts and focus on the initial blocking node. Adjust the execution of the blocking node to allow the current node to resume. The current instance is scheduled only after all upstream instances run successfully.

• If the instance is currently in the Paused Run state, the diagnostics stops and the result is Paused Run.

• A forced rerun instance does not check whether all upstream instances ran successfully. If the last run of the instance was a forced rerun, the diagnostic result is Skipped.

Scheduled time

The scheduled time diagnostic shows the result of the last diagnostic run for the instance. An instance is scheduled only when its scheduled time is reached. Otherwise, it remains in the Waiting for scheduled time state. To refresh the diagnostic result, click the refresh icon.

• If the instance has not reached its scheduled time and is not paused, the diagnostic result is Waiting for scheduled time. To schedule the instance earlier, you can perform a Forced Rerun. However, you must first ensure that this action will not affect the data quality of downstream nodes.

• If the instance is currently paused and its scheduling is paused, the diagnostic result is Paused. To run it, click Resume Scheduling.

• If the instance reached its scheduled time during its last run and it was not a forced rerun, the diagnostic result is Passed.

• A forced rerun of a start node does not check the scheduled time. It starts running directly and skips the diagnostics. If the last run of the instance was a forced rerun, the diagnostic result is Skipped.

Throttling rules

If you have purchased the artificial intelligence for IT operations add-on, you can configure throttling rules. For more information, see Configure throttling rules.

All running instances must be diagnosed against throttling rules. After an instance passes the upstream dependency and scheduled time diagnostics, it is dispatched to the resource scheduling system only if it meets all the throttling rules it has hit. To refresh the diagnostic result, click the refresh icon.

• If the last run of the instance met the upstream dependency and scheduled time requirements and all the throttling rules it hit, the diagnostic result is Pass.

• If the instance is currently being throttled and is waiting to be dispatched, and it is not paused, the diagnostic result is Throttling. The time it has been waiting is also displayed.

  Scenario	Description
  Blocking rule	Shows the names of the throttling rules that the current instance has hit. Click a rule name to view its details.
  Dispatched instance list	Shows the instances that have been dispatched in the queue of the throttling rule that was hit. You can search or filter the dispatched instances by name or ID.

• If the instance is currently paused and its scheduling is paused, the diagnostic result is Paused. You must resume the instance before it can be dispatched for resource scheduling.

Schedule resources

Instances that share running resources are generally less affected by the availability of schedule resources. Instances that use exclusive running resources are scheduled only if their resource group has sufficient idle resources to allocate. Otherwise, they enter the Waiting for schedule resource state. To refresh the diagnostic result, click the refresh icon.

• If the resource group of the instance had sufficient idle resources during its last run, and the instance is not paused, the diagnostic result is Passed.

• If the current instance does not have sufficient idle schedule resources, the diagnostic result is Waiting for schedule resource. The schedule resource diagnostics page shows the Waiting Time for Resources, diagnostic Suggestions, and a list of Instances Occupying Resources. You can follow the diagnostic suggestions on the page and refer to the Instances Occupying Resources list to ensure that the current instance obtains enough resources to run normally.

Instance execution

Only instances that have reached the execution stage appear on the Instance execution diagnostics page. The page shows the instance Run Result and Runtime Log. If the Run Result is Failed, you can use the Runtime Log to troubleshoot and resolve the issue. To refresh the diagnostic result, click the refresh icon.

Click Open Operational Log to go to the Runtime Log page. The operational log contains log information, exception information, performance diagnostics, and exception codes. For more information about performance diagnostics, see Performance diagnostics for offline integration nodes.