Monitor and Optimize an Analytics Solution · Topic 3 of 5
Resolving Eventhouse, Eventstream, T-SQL and OneLake Shortcut Errors
DP-700 troubleshooting: Eventhouse ingestion failures and query limits, Eventstream runtime logs, warehouse T-SQL errors, and broken OneLake shortcuts.
- 5 min
.show ingestion failureslists failed ingestions with a transient or permanent kind.- KQL results are truncated at 500,000 records or 64 MB by default.
- Eventstream: data insights for counts, runtime logs for errors, and publish after edits.
What it is
Real-time items, SQL engines and shortcuts fail differently from batch jobs. A stream can run while silently dropping events, a T-SQL statement can hit an unsupported feature, and a shortcut breaks when someone renames its target folder.
Find where data stops, read the message, and fix the cause: format, schema, permissions, credentials or unsupported syntax.
How it works
Eventhouse errors
Failed ingestions are listed by a management command:
.show ingestion failures
| where FailedOn > ago(1d)
| project FailedOn, Table, ErrorCode, FailureKind, Details
| Symptom | Likely cause | Fix |
|---|---|---|
| Ingestion failures, or rows with empty columns | Ingestion mapping does not match the incoming JSON or CSV fields | Correct the ingestion mapping or the data format |
| Data appears with a delay | Ingestion batching waits to collect data | Adjust the batching policy, or use streaming ingestion for low latency |
| Query result is cut off | The default result limit of 500,000 records or 64 MB | Filter or aggregate; avoid set notruncation for large results |
| Query fails on memory or runs very long | Query scans too much data or joins large tables badly | Filter by time first, project fewer columns, summarize before joining |
| Old data missing | It aged out under the retention policy | Extend retention, or keep history in OneLake |
FailureKind shows whether a failure is Transient (a retry may succeed) or Permanent (the data or mapping must change).
Eventstream errors
| Where | What it shows |
|---|---|
| Data insights | Incoming and outgoing event counts per node |
| Runtime logs | Errors, warnings and information for sources, operators and destinations |
| Node status | Whether a source or destination is active or has failed |
| Symptom | Likely cause | Fix |
|---|---|---|
| No incoming events | Wrong connection string or key, or another reader using the same Event Hubs consumer group | Fix the connection; give the eventstream its own consumer group |
| Events come in but none reach the destination | Format mismatch (JSON, CSV or Avro), or an operator filters everything out | Set the correct data format; test operators in the editor |
| Destination errors | Eventhouse table or lakehouse table schema does not match the events | Align the schema or mapping |
| Changes have no effect | The eventstream was edited but not published | Publish the changes |
T-SQL errors
| Symptom | Likely cause | Fix |
|---|---|---|
INSERT or UPDATE fails on a lakehouse | The SQL analytics endpoint is read-only | Write with Spark, or load into a warehouse |
CREATE TABLE fails on a data type | Unsupported type in Fabric Warehouse, such as nvarchar, datetime or money | Use varchar, datetime2(6) and decimal |
| A new lakehouse table is not visible in the endpoint | Metadata sync has not caught up, or the folder is not a valid Delta table | Wait or refresh the endpoint's metadata; check the table format |
| Permission denied | Missing ReadData, or missing GRANT | Grant the item permission or object permission |
| Update conflict error | Two transactions modified the same table at once under snapshot isolation | Retry the losing transaction; avoid concurrent writes to one table |
| Cross-database query fails | Three-part names work only for items in the same workspace | Use a shortcut to bring the other item's table into this workspace |
-- Fails in Fabric Warehouse: nvarchar and datetime are not supported
-- CREATE TABLE dbo.Customer (Name NVARCHAR(100), Created DATETIME);
CREATE TABLE dbo.Customer
(
Name VARCHAR(100) NOT NULL,
Created DATETIME2(6) NOT NULL
);
OneLake shortcut errors
| Symptom | Likely cause | Fix |
|---|---|---|
| Shortcut shows an error or no data | Target folder deleted, moved or renamed | Recreate the shortcut to the new path |
| Works for some users, fails for others | Internal shortcut: those users lack permission on the target | Grant them access to the target item |
| Fails for everyone after working before | External shortcut: connection credentials expired or were rotated | Update the connection |
| Access denied from cloud storage | The storage account's role assignments or firewall do not allow the identity | Grant the identity a storage data role, or allow access |
| Shortcut sits under an Unidentified folder in Tables | Target is not a Delta table | Point it to a Delta folder, or create it under Files |
A worked example
Analysts report that a KQL dashboard has shown no new data since the morning, but the IoT devices are sending normally.
In the eventstream, data insights shows incoming events but none outgoing, and the runtime logs show deserialization errors. A firmware update had switched the devices from JSON to Avro, so setting the source's data format to Avro and publishing the eventstream restores the flow.
How the exam asks about this
- "Find why rows are not being ingested into a KQL table."
.show ingestion failures. - "An eventstream receives events but the destination gets none." Check runtime logs and data insights for format or schema errors.
- "A shortcut works for the lakehouse owner but not for a Viewer." The Viewer lacks permission on the shortcut's target.
What to remember for the exam
.show ingestion failureslists failed ingestions with a transient or permanent kind.- KQL results are truncated at 500,000 records or 64 MB by default.
- Eventstream: data insights for counts, runtime logs for errors, and publish after edits.
- Fabric Warehouse does not support
nvarchar,datetimeormoney; the SQL analytics endpoint is read-only. - Internal shortcuts fail per user; external shortcuts fail for everyone when the connection breaks.
Rotating a storage key or SAS token breaks every external shortcut that uses that connection at the same moment. Update the connection as part of the rotation.
Common mistakes
Assuming no errors means no data loss in a stream. Mapping or format mismatches can drop or empty fields; compare incoming and outgoing counts.
Running set notruncation to "fix" a truncated result. Filter or aggregate instead.
Granting workspace Contributor to fix a shortcut access error. Grant the user read access on the target item instead.
Check yourself
Which command lists failed ingestions in an Eventhouse?
.show ingestion failures, which includes the error code and whether the failure is transient or permanent.Why does an
UPDATEagainst a lakehouse table through the SQL analytics endpoint fail?The endpoint is read-only; lakehouse tables are changed with Spark or loaded into a warehouse instead.
What does an Unidentified folder in a lakehouse's Tables area indicate for a shortcut?
The shortcut target is not a Delta table, so it cannot be registered as a table.
Topic 3 of 5 in Monitor and Optimize an Analytics Solution
Test yourself on Monitor and Optimize an Analytics Solution
The full Microsoft DP-700 question bank has 645 questions, with timed practice exams and an explanation for every answer.
This study guide is independent and is not affiliated with, endorsed by, or sponsored by Microsoft. Vendor names, certification names and exam codes are trademarks of their respective owners and are used for identification only.