ZipLogger.Metrics.AspNetCore 0.6.0

ZipLogger.Metrics.AspNetCore

Request metrics (APM) for ASP.NET Core apps on ZipLogger: a middleware that times every request and ships it to your workspace, powering the Metrics page's avg/p95 latency and throughput series per service.

// Program.cs
builder.Services.AddZipLoggerMetrics(builder.Configuration, service: "orders-api");

var app = builder.Build();
app.UseZipLoggerMetrics();   // early, so timings cover the whole pipeline

Configuration (reuses your existing ZipLogger settings — apps already shipping logs need nothing new):

{
  "ZipLogger": {
    "Endpoint": "https://app.ziplogger.ai",
    "ApiKey": "zk_..."
  }
}

Optional overrides under ZipLogger:Metrics (Service, MetricName, ExcludePaths, QueueCapacity, BatchSize, FlushInterval, …). When endpoint/key are unset the middleware is a no-op, so local runs work without an account.

Sessions across services

The browser SDK's instrumentFetch() sends the user's session id as W3C baggage (session.id=…) beside traceparent. UseZipLoggerMetrics() copies it onto the request's Activity as the OpenTelemetry session.id attribute — and as Activity baggage, so it flows on to the services this one calls — which is what lets ZipLogger's investigation view show every trace of a session, not only the one request. Nothing is set when the header is absent. Allow the header in CORS for cross-origin APIs: Access-Control-Allow-Headers: traceparent, baggage.

Background jobs (workers)

Non-HTTP hosts get the same treatment via ZipLoggerMetrics — time a job run and it ships as job.duration with the job name as a label:

// worker host: same registration, no middleware needed
builder.Services.AddZipLoggerMetrics(builder.Configuration, service: "orders-workers");

// in a BackgroundService (inject ZipLoggerMetrics)
using (metrics.TimeJob("nightly-sweep"))
{
    await RunSweepAsync(ct);
}

Behavior

Same delivery semantics as every ZipLogger client: enqueuing never blocks a request and never throws. Timings buffer in a bounded queue (default 10,000; drop-over-grow with a DroppedCount), ship as batches (default 100 / 2 s linger) to /ingest/v1/metrics, retry transient failures (429 honoring Retry-After, 5xx, network) with exponential backoff, and flush on graceful shutdown. Each event carries method, route (the route pattern, keeping label cardinality low), and status labels.

Need a ZipLogger server? Sign up free at ziplogger.ai — the free tier covers 1,000 logs/day. Full documentation: ziplogger.ai/docs.html