Known Limitations¶
The Logging Report is an operational view built on what the NCC platform reports about itself. That shapes what it can and cannot tell you. This page collects the boundaries worth knowing before you rely on the report for a decision.
What the report covers¶
Only activity on entities is recorded¶
The report is built on the work done for an entity: the copy activities that bring data into the Landing Zone and the notebooks that build the Landing Zone, Bronze and Silver. Everything you see is one of those activities.
Failures that happen at a higher level are not shown. If an orchestrating pipeline fails before it reaches the entities — or fails between them, or on a step that is not tied to a single entity — the report has no row to colour red. The entities it never reached simply show no activity for that run.
Info
Surfacing orchestration-level failures in the report is on the roadmap. Until then, check the pipeline run itself in the Fabric monitoring hub when the report shows nothing for a load you know was triggered.
The Gold layer is not covered¶
The report follows your data through the Landing Zone, Bronze and Silver. The Gold layer is not included.
Gold is built to your own specification and differs from customer to customer, so there is no standard shape for NCC Logging to report on. Monitoring of a Gold layer is arranged as part of the work that builds it.
Files are counted as files, not as rows¶
Some Landing Zone entities are files — a CSV, a Parquet file, a document — that are taken from the source and stored as they are. Nothing reads through them line by line, so there is no row count to report and Rows Processed shows 0.
A 0 on a file-based entity is normal and does not mean the load was empty or failed. Read the Status for that entity instead, and use Start Time and Duration to confirm the load actually did work.
Change counts come only from notebooks¶
Rows New, Rows Updated, Rows Deleted and Rows Unchanged are produced by notebooks, which compare incoming data against what is already stored. A copy activity does not compare, so for copy-activity rows only Rows Processed is available.
The Landingzone page leaves the change columns out of its table altogether, because that layer mixes copy activities and notebooks. Even a notebook-loaded Landing Zone entity therefore shows no change counts on that page.
What the statuses can and cannot tell you¶
Started does not always mean running¶
A status changes only when the activity reports it. An activity says that it started, and later says that it succeeded or failed. If that second message never arrives, the entity keeps showing ⏳ Started for the rest of the retention window.
That happens in two situations:
- The run hit an exception it could not report. A lost Spark session, a timeout or a capacity problem can end an activity so abruptly that it never sends a failure. The load did fail; the report was never told.
- The run was cancelled. Cancelling a load from the portal stops it without a closing status, so its entities stay on ⏳ rather than turning ❌.
So a ⏳ that has not moved for far longer than the load normally takes should be read as a problem, not as work in progress. Only Fabric, through the Monitoring URL, can tell you whether the run is genuinely still active. See What Started really tells you.
An empty cell means no result was reported¶
When an entity has no result for a run, the report leaves the cell blank. Almost always this means the entity was not part of that load — it is the signal the Entity Load History page is built around.
A blank is the absence of a result rather than a positive statement that nothing ran, so where it matters, confirm against the run itself before you conclude that an entity was definitely skipped.
Getting to the cause¶
The report does not contain error messages¶
There is no error text in the report and nothing to search or filter on. The report tells you that an entity failed; the message itself lives in Microsoft Fabric and is reached through the Monitoring URL.
That also means finding out why something failed requires access to the underlying Fabric workspace. Someone who can open the report is not automatically able to open the run behind it.
Run detail in Fabric expires after 30 days¶
The report keeps about 90 days of history. Fabric keeps the runs behind the Monitoring URL for 30 days. Beyond that boundary the report still shows that a load ran and whether it failed, but the error message, the logs and the run detail are gone and cannot be recovered.
If you need to keep something from a run, capture it while the run still exists. See How far back does it go?.
Working with the report¶
Custom notebooks may not report row counts¶
A custom notebook you write yourself appears in the report like any other load, with its status, timings and links into Fabric.
If its row counts stay empty while the load itself succeeds, check the custom notebook documentation first: the row counts depend on the notebook returning its data in the expected form. If everything looks right and the counts are still missing, contact InSpark and we will look at it with you.
It is a dashboard, not an alert¶
The Logging Report does not notify anyone. It shows the current state when you open it, so a failure is discovered when somebody looks.
For email when a scheduled load fails, use the failure notifications on the pipeline schedule itself — see Configure Failure Notifications. Those cover scheduled runs; they are not sent for runs started on demand.
The data can be up to an hour behind¶
Events travel from a running activity through internal processing and enrichment before they reach the report, so it is close to real time rather than live. This typically takes under 15 minutes, but it can take up to an hour. The Last event time in the header tells you how fresh the data on screen actually is. Do not use the report as a live console for a load you are watching.
You cannot export from the report¶
There is no download of the data behind a visual. A list of failed entities has to be read from the screen or captured another way.
Filters and context do not follow you¶
Each tab keeps its own filters, so a selection on one page does not carry to the next. Drill-through passes the entity you clicked but not the run, so the layer page opens on its own Period and you may have to widen it to find the run again.
Warning
Reset to 24H Default on the layer pages also switches the environment back to PROD. Check the environment switch in the header after a reset if you were working in a development environment.