Home / Data Platform / Fabric & Power BI

Data PlatformTroubleshootingIntermediate

Sev1: Power BI refreshes failing everywhere — on-premises data gateway offline or credentials expired

Morning dashboards show stale data because scheduled refreshes failed. Check the gateway cluster, the service account and data source credentials, fix them, and re-run refreshes in priority order.

Deepak Bhardwaj3 min read

Applies toPower BI service / Microsoft Fabric, on-premises data gateway (standard mode), PowerShell MicrosoftPowerBIMgmt

Symptoms

  • Refresh history shows failures with messages about the gateway being unreachable or offline.
  • Or: The credentials provided for the SQL source are invalid, or OAuth token expired errors.
  • Executives see yesterday's numbers.

1. Check the gateway status

In the Power BI service: Settings > Manage connections and gateways > On-premises data gateways. Each member of the cluster shows its status.

On the gateway server:

Get-Service PBIEgwService
Restart-Service PBIEgwService
Get-Content "C:\Users\PBIEgwService\AppData\Local\Microsoft\On-premises data gateway\GatewayInfo*.log" -Tail 50

(Log path depends on the gateway service account; the gateway app's Diagnostics > Export logs collects them.)

Finding Cause Fix
Service stopped Server reboot, crash, patching Start service, set to automatic
Service running, gateway offline in service Outbound connectivity to Azure Relay / Service Bus blocked Step 2
Gateway "needs update" Version too old; unsupported versions can stop working Update the gateway
Service account password expired Service can't log on Step 3

2. Network test

Run the network ports test in the gateway app (Diagnostics > Start new test). The gateway needs outbound HTTPS (443) to Azure endpoints and, unless forced to HTTPS, outbound TCP 5671-5672 and 9350-9354. A proxy or firewall change is a common cause.

Test-NetConnection -ComputerName login.microsoftonline.com -Port 443

3. Gateway service account

If the domain account running the gateway had its password changed or expired, the service fails to start. Either update the service logon in the gateway app (Service settings > Change account) or switch back to the default NT SERVICE\PBIEgwService if your data sources use explicit credentials.

4. Data source credentials

Failures that name a data source (not the gateway) are credential problems:

  • SQL login password rotated → update the connection in Manage connections and gateways.
  • OAuth sources (SharePoint, Dataverse, Azure SQL with Entra) → re-enter credentials in the semantic model settings; tokens expire if the owner left or their account changed.
  • Service principal secrets used for sources expired → rotate and update.

Find all failed refreshes quickly with PowerShell:

Connect-PowerBIServiceAccount
$ws = Get-PowerBIWorkspace -Scope Organization -All -Filter "state eq 'Active'"
foreach ($w in $ws) {
  $ds = Get-PowerBIDataset -WorkspaceId $w.Id -Scope Organization -ErrorAction SilentlyContinue
  foreach ($d in $ds) {
    $h = Invoke-PowerBIRestMethod -Method Get -Url "groups/$($w.Id)/datasets/$($d.Id)/refreshes?`$top=1" -ErrorAction SilentlyContinue | ConvertFrom-Json
    if ($h.value[0].status -eq 'Failed') { "{0} / {1}" -f $w.Name, $d.Name }
  }
}

5. Re-run refreshes in order

Trigger the critical models first, so the gateway and sources aren't flooded:

Invoke-PowerBIRestMethod -Method Post -Url "groups/<workspace-id>/datasets/<dataset-id>/refreshes" -Body '{"notifyOption":"MailOnFailure"}'

Tip

Tell users when data is current again; put a banner on key reports during the incident so nobody makes decisions on stale numbers.

Prevent it next time

  • Run a gateway cluster with at least two members on separate servers.
  • Keep gateways within supported versions; patch monthly.
  • Use a service account or managed credentials owned by the platform team, not personal accounts.
  • Alert on refresh failures for critical models (refresh failure emails to a team mailbox, or monitoring through the REST API).

Related fixes