Semantic Models¶
The Semantic Models dashboard shows dataset refresh volume, success, duration, waste and failure reasons for the selected range.
Open it with Select Analysis > Semantic Models. It follows the common filters; see Command Center. Refresh history comes from the refresh collector, which reads the Power BI admin refreshables list every 10 minutes and keeps every run it sees. See Collection Processors.
Image of the Semantic Models dashboard: KPI cards, Daily Refresh Volume, Status Mix, Refresh Load By Capacity and Slowest Semantic Models
../images/dashboards/semantic-models.png
TODO: confirm refresh coverage for datasets on shared capacity
The Microsoft admin refreshables list is capacity based. Confirm whether datasets in shared-capacity (Pro) workspaces appear on this page, and state it here.
KPI cards¶
| Card | What it shows | Click |
|---|---|---|
| Refreshing Datasets | Datasets with refresh history in the range (not the inventory count) | - |
| Total Refreshes | All refresh runs, with the number completed | - |
| Success Rate | Completed runs as a share of attempted runs | - |
| Failed | Failed runs, with cancelled, in-flight or unknown, disabled and other counts | Failing Datasets |
| Avg Completed Duration | Average duration of completed runs that have an end time | - |
Panels¶
| Panel | What it shows | Click |
|---|---|---|
| Daily Refresh Volume | Completed, failed and other runs per day. Days without bars had no refreshes. | - |
| Status Mix | Share of outcomes: Completed, Failed, Cancelled, Unknown, Disabled, Other | - |
| Refresh Load By Capacity | Per capacity: Capacity, SKU, Total, Completed, Failed, Fail %, Avg Dur, Max Dur. Shows who carries the load, where it fails and where it runs slow. | - |
| Slowest Semantic Models | Highest average completed duration. Columns Dataset, Workspace, Runs, Avg, Max, Fails. Models with fewer than 3 completed refreshes are left out. | A dataset opens Per-Day Refresh History; a day opens Refresh Runs |
| Most-Refreshed Datasets | Refresh count against fail rate: Dataset, Workspace, Refreshes, Fail % | A dataset opens Per-Day Refresh History |
| Refresh Waste | Refreshes on datasets whose workspace had zero report views in the window, with the share of refreshing datasets affected. Columns Dataset, Workspace, Capacity, Refreshes, Fails. | Wasted Refresh Spend |
| Failure Reasons | Failed runs grouped by the Power BI service error code | A reason opens Datasets For Reason |
Failure reason labels¶
| Label | Power BI error code |
|---|---|
| Processing error | ModelRefresh_ShortMessage_ProcessingError |
| Service error | ModelRefresh_ShortMessage_ServiceError |
| Model refresh failed | ModelRefreshFailed |
| Refresh disabled | ModelRefreshDisabled |
| Internal error | ModelRefresh_ShortMessage_InternalError |
| Dynamic datasources unsupported | DynamicDataSourcesIsNotSupportedForRefresh |
| Unbound datasources | DMTS_MonikerWithUnboundDataSources |
| Datasource moniker not found | DMTS_MonikerNotFoundError |
| OAuth token refresh failed | DMTS_OAuthTokenRefreshFailedError |
| Personal gateway offline | PersonalGateway_ShortMessage_Gateway_Offline |
| Personal gateway timeout | PersonalGateway_ProcessingTimeout |
| Enterprise gateway offline | EnterpriseGateway_ShortMessage_Gateway_Offline |
| Credentials not specified | CredentialsNotSpecified |
| Invalid credentials | InvalidConnectionCredentials |
| Processing timeout | ProcessingTimeout |
| Capacity denied operation | CapacityOperationDeniedInCurrentState |
| No error code | (no code) |
Other codes are shown as reported by Power BI.
Detail lists¶
| Drawer | Contents |
|---|---|
| Failing Datasets | Datasets with failed, cancelled or in-flight-unknown refreshes in the window |
| Datasets For Reason | Datasets whose failed refreshes carry one error code |
| Per-Day Refresh History | One dataset's refresh volume and completed durations per day |
| Refresh Runs | The individual runs behind one day |
| Wasted Refresh Spend | All refreshes on datasets whose workspace had zero report views in the covered window |
Questions this page answers¶
- Are refreshes reliable, and which datasets fail most, and why?
- Which capacities carry the refresh load and where is it slow?
- Which models refresh often but are never viewed, so their schedule can be reduced or removed?
- Which failures are credential or gateway problems that a dataset owner can fix?
Settings¶
| Property | Default | Effect |
|---|---|---|
web.apps.pbiusage.refresh.slowest.min.samples |
3 |
Completed refreshes a model needs to appear in Slowest Semantic Models |
Waste is measured at workspace and dataset level
A refresh counts as wasted when no report in the dataset's workspace was viewed in the window. MetaMiner does not attribute waste to individual reports. When the activity data does not cover the refresh window, the page says "View data does not cover this refresh window; waste verdict withheld."