Monitor dataset refresh

Hi sandlox​,

Since you asked for more detailed steps and your semantic models are in a Pro workspace (so no Fabric notebooks or workspace monitoring), here is a complete setup that works with Pro only.

Step 1: collect the refresh history with PowerShell. The Get Refresh History API returns the recent refreshes for each semantic model. This script loops over every workspace you can access and appends the results to a CSV:

# one-time: Install-Module MicrosoftPowerBIMgmt -Scope CurrentUser
Connect-PowerBIServiceAccount

$rows = @()
foreach ($ws in Get-PowerBIWorkspace -All) {
    $datasets = (Invoke-PowerBIRestMethod -Method Get `
        -Url "groups/$($ws.Id)/datasets" | ConvertFrom-Json).value

    foreach ($ds in $datasets | Where-Object { $_.isRefreshable }) {
        $refreshes = (Invoke-PowerBIRestMethod -Method Get `
            -Url "groups/$($ws.Id)/datasets/$($ds.id)/refreshes?`$top=50" |
            ConvertFrom-Json).value

        foreach ($r in $refreshes) {
            $rows += [pscustomobject]@{
                RequestId   = $r.requestId
                Workspace   = $ws.Name
                Dataset     = $ds.name
                RefreshType = $r.refreshType
                Status      = $r.status
                StartTime   = $r.startTime
                EndTime     = $r.endTime
                Error       = $r.serviceExceptionJson
            }
        }
    }
}

$rows | Export-Csv "C:\PBI\refresh_history.csv" -NoTypeInformation -Append

Step 2: schedule it. The API only keeps recent history, so run the script daily (Windows Task Scheduler is enough) and save the CSV to a OneDrive or SharePoint folder so the service can refresh from it.

Step 3: build the report. In Power BI Desktop, load the CSV, then in Power Query remove duplicates on RequestId (the append will repeat runs that are still in the API window) and add a duration column. Useful measures:

Refreshes = COUNTROWS ( Refresh )

Failed Refreshes =
CALCULATE ( [Refreshes], Refresh[Status] = "Failed" )

Success Rate =
DIVIDE ( [Refreshes] - [Failed Refreshes], [Refreshes] )

Avg Duration (min) =
AVERAGEX (
    FILTER ( Refresh, NOT ISBLANK ( Refresh[EndTime] ) ),
    DATEDIFF ( Refresh[StartTime], Refresh[EndTime], SECOND ) / 60
)

Last Status =
VAR LastStart = MAX ( Refresh[StartTime] )
RETURN
    CALCULATE ( MAX ( Refresh[Status] ), Refresh[StartTime] = LastStart )

Step 4: the dashboard layout I would use:

cards        -> refreshes today, failed today, success rate
table        -> dataset | last status | last end time | avg duration
line chart   -> duration per dataset over time (spots slow growth)
table        -> failed refreshes with the Error text

Finally, add a data alert on the "failed today" card, or a subscription, so someone hears about failures without opening the report.

So my overall approach would be:

PowerShell + Refresh History API, appended daily
-> CSV in OneDrive / SharePoint
-> dedupe on RequestId, add duration
-> status, duration and error views + alert on failures

That gives you a working monitor today on Pro, and you can later move the same logic to a notebook and lakehouse if you get Fabric capacity.

AI-assisted drafting: AI was used to help structure and phrase this response. I reviewed and validated the technical content before posting.