Download logs on the virtual network (VNet) data gateway

You can download diagnostic logs for your virtual network data gateway from the Manage connections and gateways page. The following sections describe each log type, its retention period and file format, and how to interpret its fields.

Note

  • You must be an admin on the virtual network data gateway to download logs. Check the manage users option to determine if you're an admin.
  • Log retention periods are fixed and can't be changed.
  • The gateway must be online and running to generate logs. Previously collected logs can be downloaded while the gateway is offline, within their retention period. Check the status of the virtual network by selecting the icon in the Status column.
  • The download isn't available if the Fabric private link is enabled in your tenant.
  • If you don’t see the download, check to make sure pop-ups aren't disabled on your browser.
  • This feature isn't available in certain regions and US Gov Virginia regions.

How to download logs

  1. From the page header in the Power BI service, select the Settings icon and then select Manage connections and gateways.

  2. Select the Virtual network data gateways tab.

  3. Find your virtual network data gateway and, in the same row, select the download icon.

    Screenshot of the manage connections and gateways page showing the diagnostics download icon

  4. In the pop-up window, you can choose the specific log category you need or select all available logs. Additionally, you can pick a predefined time range or customize the time range to suit your needs.

    Screenshot of the download diagnostics dialog where you choose the log category and time range

  5. Select Download diagnostics button, and view logs in the downloaded folder.

SystemCounters

System counters provide CPU and memory measurements for individual gateway instances.

  • Retention period: 29 days
  • Format: Parquet
Attribute Description
Timestamp The time the counter was emitted.
ID Unique identifier of the gateway instance (node) that reported the measurement.
CommittedMemoryinMB Memory used out of the total 8-GB RAM reserved per VNet.
CPUUsagePercentage CPU used.
MemoryAvailableinMB Memory available.

MashupLogs

Mashup logs contain detailed information about actions performed by the Power Query engine. Use these logs to investigate query execution and troubleshoot errors.

  • Retention period: 48 hours
  • Format: JSONL
Attribute Description
Start Start of action.
Action Action name.
ProductVersion Version of Mashup.
ActivityId Activity identifier.
CorrelationId Correlation identifier.
Process Process name.
Pid Process identifier.
Tid Thread identifier.
Duration Duration of action.

QueryExecutionReport

Query execution reports provide information about individual queries processed by the gateway, including their status and execution timings. Use these reports to investigate slow or failed queries and identify the gateway instance that handled them.

  • Retention period: 29 days
  • Format: Parquet
Attribute Description
ID Log identifier.
Timestamp Start of the query execution.
QueryId Query identifier.
InstanceId Unique identifier of the gateway instance (node) that reported the diagnostic record.
RequestId Request identifier.
QueryType Query type, such as DirectQuery or Data refresh operations including semantic model and dataflow refreshes.
ConnectionPath Data source kinds and paths.
Status Indicates whether the query failed or succeeded.
StartTime Start of the query execution.
DurationMS Duration of the query in milliseconds.
ExecutionStartTime Start of the query execution.
ExecutionDurationMS Duration of the query in milliseconds.
DataReadingAndSerializationStartTime Start of data serialization.
DataReadingAndSerializationDurationMS Duration of the data serialization.
StreamingDurationMS Duration of the data streaming.

ResourceConsumption

Resource consumption logs record capacity consumption attributed to the gateway. Use these logs to examine consumption over time and identify the associated Fabric capacity.

  • Retention period: 29 days
  • Format: Parquet
Attribute Description
CapacityId ID of the capacity that's charged.
CapacityUnitSeconds Capacity consumption emitted, as CapacityUnitSeconds.
GatewayId Gateway cluster identifier.
GatewayName Name of the gateway cluster.
Id Log identifier.
Timestamp The time consumption was emitted.

PipelineExecutionReport and CopyJobExecutionReport

These reports provide information about pipeline activities and copy job tasks that the gateway processes. These reports describe gateway-side activity execution. They don't necessarily describe the entire pipeline or copy job's end-to-end duration. Use them to investigate queueing, execution duration, status, and errors.

  • Retention period: 29 days
  • Format: Parquet
Attribute Description
Id Unique identifier for the diagnostic report.
Timestamp Time the gateway generates the report, in UTC.
InstanceId Unique identifier of the gateway instance (node) that reports the diagnostic record.
RequestId Identifier associated with the request.
ArtifactType Type of artifact associated with the execution.
TaskId Identifier of the execution task.
WorkspaceId Identifier of the associated workspace, when available.
PipelineRunId Identifier of the associated pipeline run, when available.
PipelineName Name of the associated pipeline, when available.
ActivityRunId Identifier of the activity run.
ActivityType Type of activity.
QueueTime Reported time spent queued, in seconds.
Status Reported execution status.
ExecutionStartTime Start time of the execution, in UTC.
ExecutionDurationMS Execution duration, in milliseconds.
ErrorCode Error code associated with the execution, when available.

InstanceLoad

Instance load diagnostics provide workload measurements per gateway instance, including the load signals that autoscaling uses, average queued and running evaluations, and pipeline activity. Use these logs to examine queueing and workload demand and to correlate load with historical instance-count trends.

  • Retention period: 29 days
  • Format: Parquet
Attribute Description
Timestamp Time the load report was generated, in UTC.
Id Unique identifier of the gateway instance (node) that reported the measurements.
AggregateLoadPercent Overall load signal that autoscaling uses for the gateway instance, based on evaluation demand and pipeline activity.
DefaultPoolLoadPercent Average evaluation load in the default pool, including queued and running evaluations, expressed as a percentage of the pool’s capacity.
DefaultPoolAveragePendingCount Average number of evaluations waiting to run in the default pool.
DefaultPoolAverageRunningCount Average number of evaluations running in the default pool.
DirectQueryPoolLoadPercent Average evaluation load in the DirectQuery pool, including queued and running evaluations, expressed as a percentage of the pool’s capacity.
DirectQueryPoolAveragePendingCount Average number of evaluations waiting to run in the DirectQuery pool.
DirectQueryPoolAverageRunningCount Average number of evaluations running in the DirectQuery pool.
PipelinePoolActive Whether the pipeline pool reported active work. This value is a Boolean activity indicator, not a utilization percentage.

To understand historical instance-count trends, group records into time intervals and count distinct Id values within each interval. This method shows how many instances reported logs during each interval, not necessarily how many instances were running simultaneously.

  • Last updated on