Analyze and monitor telemetry with KQL

[アーティクル]
07/23/2024

Telemetry from Business Central is stored in Azure Application Insights in the tables traces and pageViews. The language used to query data in Azure Application Insights is Kusto Query Language (KQL). This article has information and links to resources to get started learning about the KQL language.

As a simple example, follow these steps:

In the Azure portal, open your Application Insights resource.
In the Monitoring menu, select Logs.

On the New Query tab, enter the following to get the last 100 traces:

traces
| where timestamp > ago(7d)                     // look back 7 days
| take 100                                      // only take 100 rows
| project timestamp, message, customDimensions  // only choose these three columns 
| sort by timestamp desc                        // show the most recent data first

Where can I use Kusto Queries?

You can use Kusto queries as the data source in many places. For example:

The logs part of Application Insights in the Azure portal
Power BI reports
Alerts
Azure Dashboards
Jupyter Notebooks (with the Kqlmagic extension)

Where can I learn more about KQL?

Go to Kusto Query Language (KQL) in the Azure Data Explorer documentation.

Which tools can I use (KQL editors and clients)?

To get an overview of the different tools, go to Azure Data Explorer tools and integrations overview in the Azure Data Explorer documentation.

How can I query telemetry from Log Analytics?

With workspace-based resources, Azure Application Insights sends telemetry to a common Azure Log Analytics workspace, providing full access to all the features of Azure Log Analytics while keeping your application, infrastructure, and platform logs in a single consolidated location. This integration allows for common Azure role-based access control across your resources and eliminates the need for cross-app/workspace queries.

This table shows table names for Business Central telemetry when queried from Azure Application Insights and from Azure Log Analytics:

Table name in Azure Application Insights	Table name in Azure Log Analytics
traces	AppTraces
pageViews	AppPageViews

KQL example - following telemetry events for a session

In Business Central version 24 and later, the server and the browser components align on the globally unique identifiers (GUIDs) logged to the session_Id column in the traces and pageViews tables. This behavior allows you to do advanced troubleshooting such as following what happens within a session.

Use this KQL code to query what happens in a single session:

let _session_Id = 'fea0995e-525d-499a-bc2e-0fd27bfe412b' // change to the guid for the session you want to follow
;
let pv = 
pageViews
| where session_Id == _session_Id
| extend message = strcat('Page opened: ', name)
| project timestamp, session_Id, user_Id, message, clientType='WebClient'
;
let tra = 
traces
| where session_Id == _session_Id
| project timestamp, session_Id, user_Id, message, clientType= tostring(customDimensions.clientType)
;
union pv, tra
| order by timestamp asc

KQL example - following telemetry events for a user

In Business Central version 24 and later, the server and the browser components align on the GUIDs logged to the user_Id column in the traces and pageViews tables. This behavior allows you to do advanced troubleshooting such as following what a user does. To see which user corresponds to the guid logged in the user_Id column, see Assign a telemetry ID to users.

Use this KQL code to query what a user does across sessions:

let _user_Id = 'b99ed22a-4681-41e8-b5fc-91c004e1979a' // change to the guid for the user you want to follow
;
let pv = 
pageViews
| where user_Id == _user_Id
| extend message = strcat('Page opened: ', name)
| project timestamp, session_Id, user_Id, message, clientType='WebClient'
;
let tra = 
traces
| where user_Id == _user_Id
| project timestamp, session_Id, user_Id, message, clientType= tostring(customDimensions.clientType)
;
union pv, tra
| order by timestamp asc

KQL walkthrough example - understand report usage

This walkthrough guides you to create a Kusto (KQL) query to analyze how users use reports in a Business Central environment. You start with a sample query from the documentation article for report telemetry. Then, you change and refine the query step by step, learning about important KQL operators along the way. The final result shows you data for the most frequently used reports. The reports are grouped by the app/extension they're from, what users did with the report (download/preview/print/...), and the layout (Excel/Word/RDLC) that was used.

About this walkthrough

This walkthrough covers the following tasks:

Choose a query tool for KQL, either the Logs part of the Monitoring menu in Application Insights in the Azure portal, or in Kusto. Explorer (see Kusto.Explorer installation and user interface).
Get a sample query (in this example, we use the sample query for report renderings).
Limit the result rows by adding a take operator.
Limit the result columns by removing lines from the project operator.
Add a summarize operator and learn gotchas about the dynamic datatype customDimensions.
Add group by parts to the summarize operator.
Rename columns in the result.
Add an order by to see the most frequent reports first.
Add a take operator to get the top 10 most frequent reports.

Prerequisites

To complete this walkthrough, you need:

An Azure Application Insights resource.
Report telemetry from Business Central emitted to Application Insights.
A query tool for KQL, either the Logs part of the Monitoring menu in Application Insights in the Azure portal, or the Kusto. Explorer (see Kusto.Explorer installation and user interface).

Understand report usage with telemetry and KQL

First, go to the article on report telemetry Analyzing Report Telemetry

Now, copy the KQL query for successful report generation:

This KQL code can help you get started analyzing report rendering:

traces
| where timestamp > ago(60d) // adjust as needed
| where customDimensions has 'RT0006'
| where customDimensions.eventId == 'RT0006' 
// | where operation_Name == "Success report generation" // use this instead of eventId clause for versions 16.0 or earlier
| where customDimensions.result == "Success"
| project timestamp
// in which environment/company did it happen
, aadTenantId = customDimensions.aadTenantId
, environmentName = customDimensions.environmentName
, environmentType = customDimensions.environmentType
, companyName = customDimensions.companyName
// in which extension/app
, extensionId = customDimensions.extensionId
, extensionName = customDimensions.extensionName
, extensionVersion = customDimensions.extensionVersion
, extensionPublisher = customDimensions.extensionPublisher
// in which object
, alObjectId = customDimensions.alObjectId
, alObjectName = customDimensions.alObjectName
, alObjectType = customDimensions.alObjectType
// what did the user do
, documentFormat = customDimensions.documentFormat   // documentFormat dimension added in version 20.0
, LayoutAppId = customDimensions.layoutAppId         // layout dimensions added in version 20.0
, LayoutName = customDimensions.layoutName           // layout dimensions added in version 20.0
, LayoutType = customDimensions.layoutType           // layout dimensions added in version 20.0
, reportAction = customDimensions.reportAction       // reportAction dimension added in version 20.0
, reportingEngine = customDimensions.reportingEngine // reportingEngine dimension was added in version 17.3
// which user ran the report
, usertelemetryId = case(
  toint( substring(customDimensions.componentVersion,0,2)) >= 20, user_Id // user telemetry ID was introduced in the platform in version 20.0
, 'N/A'
)
// performance data
, numberOfRows = customDimensions.numberOfRows
// , serverExecutionTime = customDimensions.serverExecutionTime // the datatype for executionTime is timespan
, serverExecutionTimeInMS = toreal(totimespan(customDimensions.serverExecutionTime))/10000 // this shows how to convert timespan to milliseconds
, sqlDatabaseAccessIntent = customDimensions.sqlDatabaseAccessIntent  // sqlDatabaseAccessIntent dimension added in version 19.1
, sqlExecutes = customDimensions.sqlExecutes 
, sqlRowsRead = customDimensions.sqlRowsRead
// , totalTime = customDimensions.totalTime // the datatype for totalTime is timespan
, totalTimeInMS = toreal(totimespan(customDimensions.totalTime))/10000 // this shows how to convert timespan to milliseconds
| extend renderTimeInMS = totalTimeInMS - serverExecutionTimeInMS

The query in the previous paragraph has four parts: a table and three KQL operators separated by a pipe (|):

Which Kusto table that stores the data
where operators to filter the data
a project operator limits the number of columns
an extend operator to add new columns (calculations)

traces // which kusto table that stores the data
// where operators filter the data
| where timestamp > ago(60d) // adjust as needed
| where operation_Name == "Success report generation" // do note that in a later version of the schema, this field will not be used 
     or customDimensions.eventId == 'RT0006' // introduced in version 16.1
| where customDimensions.result == "Success"
//
// the project operator limits the number of columns
| project timestamp
, aadTenantId = customDimensions.aadTenantId
// ...
// ...
// the extend operator can be used to add new columns (calculations)
| extend renderTimeInMS = totalTimeInMS - serverExecutionTimeInMS

When writing KQL queries, it's good practice to limit the number of rows while developing it. This practice allows for quick explorations and experimentation without having to wait for the result in each modification of the query.

So, before running the query, add a take operator on a new line after the where operator and the project operator:

Your query should now look like this:

traces
| where timestamp > ago(60d) // adjust as needed
| where operation_Name == 'Success report generation' // do note that in a later version of the schema, this field will not be used 
     or customDimensions.eventId == 'RT0006'          // introduced in version 16.1
| where customDimensions.result == 'Success'
| take 5
| project timestamp
// in which environment/company did it happen
, aadTenantId = customDimensions.aadTenantId
, environmentName = customDimensions.environmentName
, environmentType = customDimensions.environmentType
, companyName = customDimensions.companyName
// in which extension/app
, extensionId = customDimensions.extensionId
, extensionName = customDimensions.extensionName
, extensionVersion = customDimensions.extensionVersion
, extensionPublisher = customDimensions.extensionPublisher
// in which object
, alObjectId = customDimensions.alObjectId
, alObjectName = customDimensions.alObjectName
, alObjectType = customDimensions.alObjectType
// what did the user do
, documentFormat = customDimensions.documentFormat   // documentFormat dimension added in version 20.0
, LayoutAppId = customDimensions.layoutAppId         // layout dimensions added in version 20.0
, LayoutName = customDimensions.layoutName           // layout dimensions added in version 20.0
, LayoutType = customDimensions.layoutType           // layout dimensions added in version 20.0
, reportAction = customDimensions.reportAction       // reportAction dimension added in version 20.0
, reportingEngine = customDimensions.reportingEngine // reportingEngine dimension was added in version 17.3
// which user ran the report
, usertelemetryId = case(
  toint( substring(customDimensions.componentVersion,0,2)) >= 20, user_Id // user telemetry id was introduced in the platform in version 20.0
, 'N/A'
)
// performance data
, numberOfRows = customDimensions.numberOfRows
// , serverExecutionTime = customDimensions.serverExecutionTime // the datatype for executionTime is timespan
, serverExecutionTimeInMS = toreal(totimespan(customDimensions.serverExecutionTime))/10000 // this shows how to convert timespan to millisecods
, sqlDatabaseAccessIntent = customDimensions.sqlDatabaseAccessIntent  // sqlDatabaseAccessIntent dimension added in version 19.1
, sqlExecutes = customDimensions.sqlExecutes 
, sqlRowsRead = customDimensions.sqlRowsRead
// , totalTime = customDimensions.totalTime // the datatype for totalTime is timespan
, totalTimeInMS = toreal(totimespan(customDimensions.totalTime))/10000 // this shows how to convert timespan to millisecods
| extend renderTimeInMS = totalTimeInMS - serverExecutionTimeInMS

Try running the query. You should now see five rows with numerous columns.

As you can see on the result, the original query includes many columns. In this example, we aren't interested in performance data and other details. So, we remove the lines that we don't need from the project part of the query:

traces
| where timestamp > ago(60d) // adjust as needed
| where operation_Name == 'Success report generation' // do note that in a later version of the schema, this field will not be used 
     or customDimensions.eventId == 'RT0006'          // introduced in version 16.1
| where customDimensions.result == 'Success'
| take 5
| project timestamp
// in which environment/company did it happen
, aadTenantId = customDimensions.aadTenantId
, environmentName = customDimensions.environmentName
, environmentType = customDimensions.environmentType
// in which extension/app
, extensionId = customDimensions.extensionId
, extensionName = customDimensions.extensionName
, extensionVersion = customDimensions.extensionVersion
, extensionPublisher = customDimensions.extensionPublisher
// in which object
, alObjectId = customDimensions.alObjectId
, alObjectName = customDimensions.alObjectName
, LayoutType = customDimensions.layoutType           // layout dimensions added in version 20.0
, reportAction = customDimensions.reportAction       // reportAction dimension added in version 20.0

Try running the query again. You should now see five rows with a lot fewer columns.

Next, we want to summarize how many reports ran. So, we add a new line with a summarize operator to the query:

traces
| where timestamp > ago(60d) // adjust as needed
| where operation_Name == 'Success report generation' // do note that in a later version of the schema, this field will not be used 
     or customDimensions.eventId == 'RT0006'          // introduced in version 16.1
| where customDimensions.result == 'Success'
| take 5
| project timestamp
// in which environment/company did it happen
, aadTenantId = customDimensions.aadTenantId
, environmentName = customDimensions.environmentName
, environmentType = customDimensions.environmentType
// in which extension/app
, extensionId = customDimensions.extensionId
, extensionName = customDimensions.extensionName
, extensionVersion = customDimensions.extensionVersion
, extensionPublisher = customDimensions.extensionPublisher
// in which object
, alObjectId = customDimensions.alObjectId
, alObjectName = customDimensions.alObjectName
, LayoutType = customDimensions.layoutType           // layout dimensions added in version 20.0
, reportAction = customDimensions.reportAction       // reportAction dimension added in version 20.0
| summarize count()

This query shows you a total count of all reports that were run. Because you have a take 5 in there, the count is also five. Now, change that limit to 100 as we're ready to group by some columns:

// Note that this query will fail with the error
// 
// Summarize group key 'alObjectId' is of a 'dynamic' type. Please use an explicit cast (for example, 'summarize ... by tostring(alObjectId)') as grouping by a 'dynamic' type is not supported.
//
traces
| where timestamp > ago(60d) // adjust as needed
| where operation_Name == 'Success report generation' // do note that in a later version of the schema, this field will not be used 
     or customDimensions.eventId == 'RT0006'          // introduced in version 16.1
| where customDimensions.result == 'Success'
| take 100
| project timestamp
// in which environment/company did it happen
, aadTenantId = customDimensions.aadTenantId
, environmentName = customDimensions.environmentName
, environmentType = customDimensions.environmentType
// in which extension/app
, extensionId = customDimensions.extensionId
, extensionName = customDimensions.extensionName
, extensionVersion = customDimensions.extensionVersion
, extensionPublisher = customDimensions.extensionPublisher
// in which object
, alObjectId = customDimensions.alObjectId
, alObjectName = customDimensions.alObjectName
, LayoutType = customDimensions.layoutType           // layout dimensions added in version 20.0
, reportAction = customDimensions.reportAction       // reportAction dimension added in version 20.0
| summarize count() by alObjectName

All of the context for the report rendering (which object ID/name, extension/app details etc.) is stored in a data structure called customDimensions. For such columns to work in the by part of the summarize operator, they must be converted to strings or numbers, otherwise you get this error when running the query:

Summarize group key 'alObjectId' is of a 'dynamic' type. Please use an explicit cast (for example, 'summarize ... by tostring(alObjectId)') as grouping by a 'dynamic' type is not supported.

Let's quickly fix that and add a tostring command to the by part of the summarize line:

traces
| where timestamp > ago(60d) // adjust as needed
| where operation_Name == 'Success report generation' // do note that in a later version of the schema, this field will not be used 
     or customDimensions.eventId == 'RT0006'          // introduced in version 16.1
| where customDimensions.result == 'Success'
| take 100
| project timestamp
// in which environment/company did it happen
, aadTenantId = customDimensions.aadTenantId
, environmentName = customDimensions.environmentName
, environmentType = customDimensions.environmentType
// in which extension/app
, extensionId = customDimensions.extensionId
, extensionName = customDimensions.extensionName
, extensionVersion = customDimensions.extensionVersion
, extensionPublisher = customDimensions.extensionPublisher
// in which object
, alObjectId = customDimensions.alObjectId
, alObjectName = customDimensions.alObjectName
, LayoutType = customDimensions.layoutType           // layout dimensions added in version 20.0
, reportAction = customDimensions.reportAction       // reportAction dimension added in version 20.0
| summarize count() by tostring(alObjectName)

The Summarize group key 'alObjectId' is of a 'dynamic' type error is common to get when you write KQL. Now you know how to fix it.

Let's add a few more columns to the by part of the summarize line and tidy up the query (remove comments and unused columns from the project part):

traces
| where timestamp > ago(60d) // adjust as needed
| where customDimensions.eventId == 'RT0006'
| where customDimensions.result == 'Success'
| take 100
| project extensionPublisher = customDimensions.extensionPublisher
, alObjectName = customDimensions.alObjectName
, LayoutType = customDimensions.layoutType
, reportAction = customDimensions.reportAction
| summarize count() 
  by tostring(alObjectName)
   , tostring(LayoutType)
   , tostring(reportAction)   
   , tostring(extensionPublisher)

If we want the columns in the result to have some nice names, we can rename them directly in the summarize operator:

traces
| where timestamp > ago(60d) // adjust as needed
| where customDimensions.eventId == 'RT0006'
| where customDimensions.result == 'Success'
| take 100
| project extensionPublisher = customDimensions.extensionPublisher
, alObjectName = customDimensions.alObjectName
, LayoutType = customDimensions.layoutType
, reportAction = customDimensions.reportAction
| summarize report_count = count() 
  by report_name   = tostring(alObjectName)
   , layout_type   = tostring(LayoutType)
   , report_action = tostring(reportAction)   
   , publisher     = tostring(extensionPublisher)

We want to see the most frequent reports in the top of the list, so we add a new line with an order by operator (remember to add _desc_ or _asc_):

traces
| where timestamp > ago(60d) // adjust as needed
| where customDimensions.eventId == 'RT0006'
| where customDimensions.result == 'Success'
| take 100
| project extensionPublisher = customDimensions.extensionPublisher
, alObjectName = customDimensions.alObjectName
, LayoutType = customDimensions.layoutType
, reportAction = customDimensions.reportAction
| summarize report_count = count() 
  by report_name   = tostring(alObjectName)
   , layout_type   = tostring(LayoutType)
   , report_action = tostring(reportAction)   
   , publisher     = tostring(extensionPublisher)   
| order by report_count desc

We just want to see the top 10 results. So we add one more operator to the end of the query:

traces
| where timestamp > ago(60d) // adjust as needed
| where customDimensions.eventId == 'RT0006'
| where customDimensions.result == 'Success'
| take 100
| project extensionPublisher = customDimensions.extensionPublisher
, alObjectName = customDimensions.alObjectName
, LayoutType = customDimensions.layoutType
, reportAction = customDimensions.reportAction
| summarize report_count = count() 
  by report_name   = tostring(alObjectName)
   , layout_type   = tostring(LayoutType)
   , report_action = tostring(reportAction)   
   , publisher     = tostring(extensionPublisher)   
| order by report_count desc
| take 10

Our query now delivers a result that looks like what we want. Final step is to deal with the initial take 100 line as the query is now ready and we're no longer experimenting. For future experiments, we just comment it out, to make it easy to add again, should we need it:

// Final version
traces
| where timestamp > ago(60d) // adjust as needed
| where customDimensions.eventId == 'RT0006'
| where customDimensions.result == 'Success'
//| take 100
| project extensionPublisher = customDimensions.extensionPublisher
, alObjectName = customDimensions.alObjectName
, LayoutType = customDimensions.layoutType
, reportAction = customDimensions.reportAction
| summarize report_count = count() 
  by report_name   = tostring(alObjectName)
   , layout_type   = tostring(LayoutType)
   , report_action = tostring(reportAction)   
   , publisher     = tostring(extensionPublisher)   
| order by report_count desc
| take 10

次の方法で共有