Skip to content

Common Monitoring Scenarios

Step-by-step routes through the Logging Report for the questions that come up most often in day-to-day operation.

Each scenario assumes you have the report open at Observability > Logging in the NCC Portal, and that the environment switch in the header is set to the environment you want to inspect.

Did last night's load succeed?

  1. Open the Overview page.
  2. Check Last event time in the header. If it is older than your schedule, the load did not run — check the schedule on the Load Planner Group rather than the data.
  3. Scan the Latest Entity Load Details per Layer table. All ✅ means the platform is healthy.
  4. Confirm the run itself in Processes started at the bottom-right: the start time you expect should be listed.

Takes about ten seconds once you are used to it.

An entity failed to load — why?

  1. Open the Failed Entities page.
  2. Set Period to the window that covers the load, usually 24H.
  3. Find the failed entity. Note the Type column: a copy activity usually points at the source system or the connection, a notebook at the data or the transformation. It also decides how you reach the error in the next step.
  4. Select the link in the Monitoring URL column. Microsoft Fabric opens the exact run.
  5. Read the error message. For a Landing Zone failure, open the failed copy activity. For a notebook failure you land on the parallel main notebook, so open the cell that executes the DAGs and select the name of the failed notebook — see From a failure to its cause.
  6. If several entities failed at once, start with the earliest one — later failures are often consequences of it.

Warning

Do this within 30 days of the failure. Fabric keeps the run behind the Monitoring URL for 30 days; after that the report still shows the ❌ but the error message can no longer be retrieved.

See Failed Entities for the full walkthrough and Common Errors and Solutions for frequent messages.

Which entities were not loaded?

A failure is visible. An entity that was silently skipped is not — it has no status at all. The Entity Load History page is built for exactly this.

  1. Open the Entity Load History page.
  2. Set Period to 7D or 30D.
  3. Read across the rows. Where the ✅ icons stop and the cells go empty, that entity stopped being loaded.
  4. Note the last date it did load, then check whether something changed on that date — was the entity removed from its Load Planner Group, or did its promotion status change? The Audit report is a good place to check: filter it to the entity's Name and see what was changed around that date, and by whom.

Tip

Filter Load Groups to one group first. An entity that disappears from every group is a different problem from one that disappeared from a single schedule.

How do I filter on domain, data source or period?

All filters sit in the bar directly below the header, and work the same way on every page.

  • Domain Name — select your domain to limit the page to the workspaces in it. Useful when several domains share one platform.
  • Datasource Name — limit to a single source system. The quickest way to confirm whether a problem is one source or the whole platform. Not available on the layer pages.
  • Source Schema and Source Name — narrow to a group of entities or one specific entity.
  • Period — choose 24H, 48H, 7D, 30D or YTD. This is a rolling window ending now, not a calendar period. Note that beyond 30 days you can still see what happened but no longer why, because Fabric has discarded the run detail.
  • Load Groups — limit to one scheduled load.
  • Process Start Time — limit to one specific run.

Filters apply to the current page only. On the layer pages, Reset to 24H Default clears them — and also puts the environment back to PROD, so check the environment switch afterwards.

Note

The filters only offer values that actually occur in the retained history. A missing value in a filter list is itself information: nothing was recorded for it.

Is a load still running?

  1. Open the Overview page.
  2. Look for ⏳ icons. They mean an activity started but has not reported an end.
  3. Confirm against Last event time: if events are still arriving, the load is progressing.
  4. A ⏳ that is hours old on a load that normally takes minutes is not progress. Open the layer page for that layer, find the entity in the detail table and follow its Monitoring URL into Fabric to see whether the run is genuinely still active.

Started does not always mean running

An activity stays on ⏳ until it reports a result. If it broke off before it could report, or if the run was cancelled from the portal, that result never comes and the ⏳ never clears — even though nothing is running any more.

So a stale ⏳ has three possible explanations: a genuinely long load, a run that died without reporting, or a cancelled run. Only Fabric can tell you which, through the Monitoring URL. See What Started really tells you.

How much data was loaded, and is that normal?

  1. Open the layer page for the layer you are interested in.
  2. The Rows processed card shows the total for the selected period.
  3. The Rows processed line chart shows the last 30 days per entity — use it to judge whether today is normal.
  4. For a specific entity, use the detail table: Rows Processed, and in Bronze and Silver also Rows New, Rows Updated, Rows Deleted and Rows Unchanged.

A green status with an unusual row count is one of the most valuable signals in the report: the load worked, but it did not do what it usually does.

Why is a load getting slower?

  1. Open the layer page and look at the Duration (m) chart, which covers the last 30 days per entity.
  2. A steadily climbing line means growth; a step change means something changed on that date.
  3. For exact figures per run, use the Duration column in the detail table, and Processes Run for the total run time of the whole process.

Someone says the data in a report is wrong

  1. Open the Overview page and find the entity in Latest Entity Load Details per Layer.
  2. Hover over its status to see when it last loaded and how many rows changed.
  3. If the last load is older than expected, the data is stale rather than wrong — continue with Which entities were not loaded? above.
  4. If it loaded normally, right-click the entity and choose Drill through, selecting the layer of the row you clicked, to open its layer page. Compare the row counts of the last few runs.

I do not see my most recent load

It can take a while for the result of a step to reach the report — see The data can be up to an hour behind.

  1. Check Last event time in the header.
  2. Confirm the environment switch is on the right environment.
  3. Confirm the Period filter covers the moment the load ran.
  4. Refresh the page after a while.

If the load still does not appear and Fabric shows it ran, contact service@inspark.nl. Include the log message from the pipeline or notebook run in Fabric if you have it — it speeds up the investigation.

Next steps