Work with custom event in Application Insights

Application Insights can offer you lots of built-in telemetries such as Page Views and Exceptions. But quite often we need to track some customize/business performances. Some examples that we are using now are:

  • SharePoint crawled items from content source X
  • Daily usage of a web application of all users from company ABC
  • Usage and performance of different version of API

To meet these challenges, Application Insights offers API for custom event and metrics.

In this article I will show how to monitor and analysis API performance by using custom events, including export it into external database for archiving and future analysis.

Report events

Our ASP.Net Web API application is already instrumented with Application Insights. The goal is to also report the performance and usage of API in below format.

By following the documentation, we have below code in the server side for sending the performance data of each API request:

1
2
3
4
5
6
7
8
9
10
var properties = new Dictionary<string, string>{
{ "ElapsedMilliseconds", apiCall.ElapsedMilliseconds.ToString()},
{ "ContractName", apiCall.ContractName},
{ "AbsoluteUri", apiCall.AbsoluteUri},
{ "Action", apiCall.Action },
{ "UserId", apiCall.UserId },
{ "AccountId", apiCall.AccountId },
{ "AccountName", apiCall.AccountName }
};
_appInsightTelemetryClient.TrackEvent(apiCall.ContractName + " - " + apiCall.Action, properties);

Then you can verify the raw data you got in Application Insights

And also in the analytics

You can see that all customized attributes in this event are in customDimensions object.

Analysis events

From here, you can run analytic queries, such as

Find the 5 most slowest API actions in last 24 hours

1
2
3
4
5
6
7
customEvents 
| where timestamp >= ago(24h)
| extend action = tostring(customDimensions.["Action"])
| extend ElapsedSeconds = todouble(customDimensions.["ElapsedMilliseconds"]) / 1000
| summarize avg(ElapsedSeconds) by action
| top 5 by avg_ElapsedSeconds desc
| render barchart
1
2
3
4
5
6
customEvents 
| where timestamp >= ago(7d)
| extend action = tostring(customDimensions.["Action"])
| summarize count() by action
| order by count_ desc
| render barchart

Percentile of one action “GetUser” in last 7 days

1
2
3
4
5
6
customEvents
| where timestamp >= ago(7d)
| where customDimensions.["Action"] == "GetUser"
| extend ElapsedSeconds = todouble(customDimensions.["ElapsedMilliseconds"]) / 1000
| summarize percentiles(ElapsedSeconds, 50, 90, 95) by bin(timestamp, 1h)
| render timechart

Export to SQL database

As I described in my previous blog Application Insights Export and PowerBI, you can export custom event telemetry data to a SQL database.

Create the database table

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
CREATE TABLE [dbo].[mywebapi-ai-events]
(
[ID] varchar(50),
[TimeStamp] varchar(50),
ElapsedMilliseconds varchar(20),
AbsoluteUri varchar(300),
[UserId] varchar(200),
[Action] varchar(100),
[AccountId] varchar(200),
ContractName varchar(200),
AccountName varchar(200),
CONSTRAINT [PK_Source] PRIMARY KEY CLUSTERED
(
[ID] ASC
)
)

Setup input in Streaming Analytics

Setup output in Streaming Analytics

Setup query in Streaming Analytics

1
2
3
4
5
6
7
8
9
10
11
12
13
SELECT
C.internal.data.id as ID
, C.context.data.eventTime as [TimeStamp]
,GetRecordPropertyValue(GetArrayElement(C.context.custom.dimensions, 0),'ElapsedMilliseconds') as ElapsedMilliseconds
,GetRecordPropertyValue(GetArrayElement(C.context.custom.dimensions, 1),'AbsoluteUri') as AbsoluteUri
,GetRecordPropertyValue(GetArrayElement(C.context.custom.dimensions, 2),'UserId') as UserId
,GetRecordPropertyValue(GetArrayElement(C.context.custom.dimensions, 3),'Action') as Action
,GetRecordPropertyValue(GetArrayElement(C.context.custom.dimensions, 4),'AccountId') as AccountId
,GetRecordPropertyValue(GetArrayElement(C.context.custom.dimensions, 5),'ContractName') as ContractName
,GetRecordPropertyValue(GetArrayElement(C.context.custom.dimensions, 6),'AccountName') as AccountName
INTO
[output-ai-apicustomevents]
FROM [input-ai-apicustomevents] C

Also see some dicussion here

Finally the data will be stored in database

Share Comments