Graceful Shutdown in ASP.NET Core Services

Pods killed mid-request caused 502s during every deploy until we implemented proper shutdown.

The Pattern

builder.Services.Configure<HostOptions>(opts =>
{
    opts.ShutdownTimeout = TimeSpan.FromSeconds(25);
});

app.Lifetime.ApplicationStopping.Register(() =>
{
    // Stop accepting new work
    _healthCheck.MarkNotReady();
});

Kubernetes Alignment

• terminationGracePeriodSeconds: 30
• preStop sleep 5s (remove from endpoints before drain)
• App drain: wait for in-flight HTTP to complete (max 25s)
• Force kill at 30s

Readiness probe must flip before SIGTERM for zero-downtime rolls. The preStop hook exists because endpoint updates lag behind the signal.