切换Application Insights至连接字符串后异常无法在Azure中显示

微软已宣布Application Insights的Instrumentation Key将被弃用，推荐使用连接字符串。但切换后，通过ILogger记录的日志可正常显示在Azure Application Insights中，抛出的异常却无法被捕获显示。

切换前可正常工作的代码 #

private static IServiceCollection ConfigureServices()
{
    //here I just read app settings.json file
    IConfiguration config = LoadConfiguration();
    IServiceCollection services = new ServiceCollection();

    services.AddLogging(builder =>
        {
            builder.AddConfiguration(config);
            builder.AddConsole();
            builder.AddApplicationInsights(config.GetSection("InstrumentationKey").Value)
                .AddFilter("", LogLevel.Trace);
        })
        .AddSingleton<IExampleService, ExampleService>();
    
    return services;
}

切换至连接字符串后的代码 #

private static IServiceCollection ConfigureServices()
{
    IConfiguration config = LoadConfiguration();
    IServiceCollection services = new ServiceCollection();

    services.AddLogging(builder =>
        {
            builder.AddConfiguration(config);
            builder.AddConsole();
            builder.AddApplicationInsights(
                telConfig => telConfig.ConnectionString = config.GetSection("AIConnectionString").Value),
                ailoptions => {}).AddFilter("", LogLevel.Trace);
        })
        .AddSingleton<IExampleService, ExampleService>();

    return services;
}

问题分析 #

使用连接字符串配置时，默认的异常追踪逻辑可能未被正确初始化，或者ILogger的异常收集选项未显式启用，导致抛出的异常无法被Application Insights捕获。

解决方案 #

1. 显式配置异常追踪选项 #

修改AddApplicationInsights的配置，开启TrackExceptionsAsExceptionTelemetry选项，确保ILogger记录的异常被转换为Exception Telemetry：

private static IServiceCollection ConfigureServices()
{
    IConfiguration config = LoadConfiguration();
    IServiceCollection services = new ServiceCollection();

    services.AddLogging(builder =>
        {
            builder.AddConfiguration(config);
            builder.AddConsole();
            builder.AddApplicationInsights(
                telConfig => 
                {
                    telConfig.ConnectionString = config.GetSection("AIConnectionString").Value;
                },
                ailoptions => 
                {
                    // 显式开启异常追踪
                    ailoptions.TrackExceptionsAsExceptionTelemetry = true;
                }).AddFilter("", LogLevel.Trace);
        })
        .AddSingleton<IExampleService, ExampleService>();

    return services;
}

2. 针对后台服务场景添加Telemetry Worker Service #

如果是Worker Service或后台服务，需添加AddApplicationInsightsTelemetryWorkerService来确保未处理的异常被收集：

services.AddLogging(builder =>
    {
        // 原有配置...
    })
    .AddApplicationInsightsTelemetryWorkerService(config => 
    {
        config.ConnectionString = config.GetSection("AIConnectionString").Value;
    })
    .AddSingleton<IExampleService, ExampleService>();

3. 确保异常通过ILogger正确记录 #

直接抛出的未处理异常需要依赖Telemetry服务收集，而通过ILogger.LogError记录的异常更易被捕获，建议在catch块中显式记录：

public void DoWork()
{
    try
    {
        throw new InvalidOperationException("测试异常");
    }
    catch (Exception ex)
    {
        _logger.LogError(ex, "执行操作时发生异常");
        throw; // 可选：保留异常抛出，让上层或Telemetry服务处理
    }
}

4. 验证日志级别配置 #

确保appsettings.json中Application Insights的日志级别包含Error及以上：

"Logging": {
    "LogLevel": {
        "Default": "Trace"
    },
    "ApplicationInsights": {
        "LogLevel": {
            "Default": "Trace"
        }
    }
}

5. 临时禁用采样验证 #

自适应采样可能过滤异常，可临时禁用采样排查：

telConfig.EnableAdaptiveSampling = false;

内容的提问来源于stack exchange，提问作者Aamon