A sink for Serilog that writes events to Microsoft Application Insights. This sink comes with several defaults that send
Serilog LogEvent
messages to Application Insights as either EventTelemetry
or TraceTelemetry
.
The simplest way to configure Serilog to send data to a Application Insights dashboard via instrumentation key is to use current active telemetry configuration which is already initialised in most application types like ASP.NET Core, Azure Functions etc.:
var log = new LoggerConfiguration()
.WriteTo.ApplicationInsights(TelemetryConfiguration.Active, TelemetryConverter.Traces)
.CreateLogger();
.. or as EventTelemetry
:
var log = new LoggerConfiguration()
.WriteTo.ApplicationInsights(TelemetryConfiguration.Active, TelemetryConverter.Events)
.CreateLogger();
You can also pass an instrumentation key and this sink will create a new
TelemetryConfiguration
based on it, however it's actively discouraged compared to using already initialised telemetry configuration, as your telemetry won't be properly correlated.
Note: Whether you choose Events
or Traces
, if the LogEvent contains any exceptions it will always be sent
as ExceptionTelemetry
.
The
singleton TelemetryConfiguration.Active
has been deprecated in the Application Insights SDK on .NET Core in favor of dependency injection pattern
.
Therefore, now we need to pass in the TelemetryConfiguration
instance that was added either
by services.AddApplicationInsightsTelemetryWorkerService()
(if you're developing
a non-http application)
or services.AddApplicationInsightsTelemetry()
(if you're developing
an ASP.Net Core application) during Startup
in ConfigureServices
.
Log.Logger = new LoggerConfiguration()
.WriteTo.ApplicationInsights(
serviceProvider.GetRequiredService<TelemetryConfiguration>(),
TelemetryConverter.Traces)
.CreateLogger();
However, you probably want to setup your Logger as close to the entry point of your application as possible, so that any
startup errors can be caught and properly logged. The problem is that now we're in a chicken-and-egg situation: we want
to setup the logger early, but we need the TelemetryConfiguration
which still haven't been added to our DI container.
Luckily from version 4.0.x of the Serilog.Extensions.Hosting
we have the possibility to configure a bootstrap logger
to capture early errors, and then change it using DI dependant services once they are configured.
// dotnet add package serilog.extensions.hosting -v 4.0.0-*
public static class Program
{
public static void Main(string[] args)
{
Log.Logger = new LoggerConfiguration()
.WriteTo.Console()
.CreateBootstrapLogger();
try
{
CreateHostBuilder(args).Build().Run();
}
catch (Exception ex)
{
Log.Fatal(ex, "An unhandled exception occurred during bootstrapping");
}
finally
{
Log.CloseAndFlush();
}
}
public static IHostBuilder CreateHostBuilder(string[] args) =>
Host.CreateDefaultBuilder(args)
.UseSerilog((context, services, loggerConfiguration) => loggerConfiguration
.WriteTo.ApplicationInsights(
services.GetRequiredService<TelemetryConfiguration>(),
TelemetryConverter.Traces))
.ConfigureWebHostDefaults(webBuilder => { webBuilder.UseStartup<Startup>(); });
}
Configuring in code, as shown above, is recommended because the existing TelemetryClient
can be injected.
The following configuration shows how to create an ApplicationInsights sink with ReadFrom.Configuration(configuration).
{
"Serilog": {
"Using": [
"Serilog.Sinks.ApplicationInsights"
],
"MinimumLevel": {
"Default": "Information",
"Override": {
"Microsoft": "Warning",
"System": "Warning"
}
},
"WriteTo": [
{
"Name": "ApplicationInsights",
"Args": {
"connectionString": "[your connection string here]",
"telemetryConverter":
"Serilog.Sinks.ApplicationInsights.TelemetryConverters.TraceTelemetryConverter, Serilog.Sinks.ApplicationInsights"
}
}
],
"Enrich": [ "FromLogContext" ],
"Properties": {
"Application": "Sample"
}
}
}
The telemetryConverter
has to be specified with the full type name and the assembly name.
A connectionString
can be omitted if it's supplied in the APPLICATIONINSIGHTS_CONNECTION_STRING
environment variable.
By default, trace telemetry submits:
- rendered message in trace's standard message property.
- severity in trace's standard severityLevel property.
- timestamp in trace's standard timestamp property.
- messageTemplate in customDimensions.
- custom log properties as customDimensions.
Event telemetry submits:
- message template as event name.
- renderedMessage in customDimensions.
- timestamp in event's standard timestamp property.
- custom log properties as customDimensions.
Exception telemetry submits:
- exception as standard AI exception.
- severity in trace's standard severityLevel property.
- timestamp in trace's standard timestamp property.
- custom log properties as customDimensions.
Note that log context properties are also included in customDimensions when Serilog is configured with
.Enrich.FromLogContext()
.
By default custom properties are converted to compact JSON, for instance:
var position = new { Latitude = 25, Longitude = 134 };
var elapsedMs = 34;
var numbers = new int[] { 1, 2, 3, 4 };
Logger.Information("Processed {@Position} in {Elapsed:000} ms., str {str}, numbers: {numbers}",
position, elapsedMs, "test", numbers);
will produce the following properties in customDimensions:
Property | Value |
---|---|
Elapsed | 34 |
Position | {"Latitude":25,"Longitude":134} |
numbers | [1,2,3,4] |
This is a breaking change from v2 which was producing these properties:
Property | Value |
---|---|
Elapsed | 34 |
Position.Latitude | 25 |
Position.Longitude | 134 |
numbers.0 | 1 |
numbers.1 | 2 |
numbers.2 | 3 |
numbers.3 | 4 |
You can revert the old behavior by overriding standard telemetry formatter, for instance:
private class DottedOutTraceTelemetryConverter : TraceTelemetryConverter
{
public override IValueFormatter ValueFormatter => new ApplicationInsightsDottedValueFormatter();
}
Additionally, you can also customize whether to send the LogEvents at all, if so which type(s) of Telemetry to send
and also what to send (all or no LogEvent properties at all) by passing your own ITelemetryConverter
instead
of TelemetryConverter.Traces
or TelemetryConverter.Events
by either implementing your own ITelemetryConverter
or
deriving from TraceTelemetryConverter
or EventTelemetryConverter
and overriding specific bits.
Log.Logger = new LoggerConfiguration()
.WriteTo.ApplicationInsights(configuration, new CustomConverter())
.CreateLogger();
// ...
private class CustomConverter : TraceTelemetryConverter
{
public override IEnumerable<ITelemetry> Convert(LogEvent logEvent, IFormatProvider formatProvider)
{
// first create a default TraceTelemetry using the sink's default logic
// .. but without the log level, and (rendered) message (template) included in the Properties
foreach (ITelemetry telemetry in base.Convert(logEvent, formatProvider))
{
// then go ahead and post-process the telemetry's context to contain the user id as desired
if (logEvent.Properties.ContainsKey("UserId"))
{
telemetry.Context.User.Id = logEvent.Properties["UserId"].ToString();
}
// post-process the telemetry's context to contain the operation id
if (logEvent.Properties.ContainsKey("operation_Id"))
{
telemetry.Context.Operation.Id = logEvent.Properties["operation_Id"].ToString();
}
// post-process the telemetry's context to contain the operation parent id
if (logEvent.Properties.ContainsKey("operation_parentId"))
{
telemetry.Context.Operation.ParentId = logEvent.Properties["operation_parentId"].ToString();
}
// typecast to ISupportProperties so you can manipulate the properties as desired
ISupportProperties propTelematry = (ISupportProperties)telemetry;
// find redundant properties
var removeProps = new[] { "UserId", "operation_parentId", "operation_Id" };
removeProps = removeProps.Where(prop => propTelematry.Properties.ContainsKey(prop)).ToArray();
foreach (var prop in removeProps)
{
// remove redundant properties
propTelematry.Properties.Remove(prop);
}
yield return telemetry;
}
}
}
If you want to skip sending a particular LogEvent, just return null
from your own converter method.
The easiest way to customise included properties is to subclass one of the ITelemetryConverter
implementations. For
instance, let's include renderedMessage
in event telemetry:
private class IncludeRenderedMessageConverter : EventTelemetryConverter
{
public override void ForwardPropertiesToTelemetryProperties(LogEvent logEvent,
ISupportProperties telemetryProperties, IFormatProvider formatProvider)
{
base.ForwardPropertiesToTelemetryProperties(logEvent, telemetryProperties, formatProvider,
includeLogLevel: false,
includeRenderedMessage: true,
includeMessageTemplate: false);
}
}
As explained by the Application Insights documentation , the default behaviour of the AI client is to buffer messages and send them to AI in batches whenever the client seems fit. However, this may lead to lost messages when your application terminates while there are still unsent messages in said buffer.
You can control when AI shall flush its messages, for example when your application closes:
- Create a custom
TelemetryClient
and hold on to it in a field or property:
// private TelemetryClient _telemetryClient;
// ...
_telemetryClient = new TelemetryClient()
{
InstrumentationKey = "<My AI Instrumentation Key>"
};
- Use that custom
TelemetryClient
to initialize the Sink:
var log = new LoggerConfiguration()
.WriteTo
.ApplicationInsights(_telemetryClient, TelemetryConverter.Events)
.CreateLogger();
- Call .Flush() on the TelemetryClient whenever you deem necessary, i.e. Application Shutdown:
_telemetryClient.Flush();
// The AI documentation mentions that calling `Flush()` *can* be asynchronous and non-blocking so
// depending on the underlying Channel to AI you might want to wait some time
// specific to your application and its connectivity constraints for the flush to finish.
await Task.Delay(1000);
// or
System.Threading.Thread.Sleep(1000);
Application Insight's operation id is pushed out if you set operationId
LogEvent property. If it's present, AI's
operation id will be overridden by the value from this property.
This can be set like so:
public class OperationIdEnricher : ILogEventEnricher
{
public void Enrich(LogEvent logEvent, ILogEventPropertyFactory propertyFactory)
{
if (logEvent.Properties.TryGetValue("RequestId", out var requestId))
{
logEvent.AddPropertyIfAbsent(new LogEventProperty("operationId", requestId));
}
}
}
Application Insight supports component version and is pushed out if you set version
log event property. If it's
present, AI's operation version will include the value from this property.
Azure functions has out of the box integration with Application Insights, which automatically logs functions execution start, end, and any exception. Please refer to the original documenation on how to enable it.
This sink can enrich AI messages, preserving operation_Id and other context information which is already provided by functions runtime. The easiest way to configure Serilog in this case is to use the injected TelemetryClient which should be automatically configured by the environment through the APPLICATIONINSIGHTS_CONNECTION_STRING appsetting. You can, for instance, initialise logging in the static constructor:
[assembly: FunctionsStartup(typeof(MyFunctions.Startup))]
namespace MyFunctions
{
public class Startup : FunctionsStartup
{
public override void Configure(IFunctionsHostBuilder builder)
{
builder.Services.AddSingleton<ILoggerProvider>((sp) =>
{
Log.Logger = new LoggerConfiguration()
.Enrich.FromLogContext()
.WriteTo.ApplicationInsights(sp.GetRequiredService<TelemetryClient>(), TelemetryConverter.Traces)
.CreateLogger();
return new SerilogLoggerProvider(Log.Logger, true);
});
}
}
}
Copyright © 2022 Serilog Contributors - Provided under the Apache License, Version 2.0.
See also: Serilog Documentation