A lightweight .NET library for safely shutting down applications while allowing critical operations to complete.

Overview

When an application receives SIGTERM, the library:

1. Starts draining the application.

2. Changes the readiness health check to 503, preventing new traffic.

3. Keeps the liveness health check at 200 so the application remains alive.

4. Waits for all registered critical operations to finish.

5. Shuts down the application after the last critical operation completes.

                    SIGTERM

                       │

                       ▼

              ┌─────────────────┐

              │ Start Draining  │

              └────────┬────────┘

                       │

              Readiness = 503

              Liveness  = 200

                       │

                       ▼

                No new traffic

                       │

                       ▼

          Wait for critical operations

                       │

                       ▼

              Last operation ends

                       │

                       ▼

             Application shutdown

Critical Operation Tracker

Use the ICriticalOperationTracker to register operations that must not be interrupted during shutdown.

public async Task ProcessOrderAsync(

    CancellationToken cancellationToken)

{

    using var operation = _tracker.BeginOperation();

    await ProcessOrderInternalAsync();

}

The operation is automatically released when the using scope ends, including when an exception occurs.

Shutdown Coordination

The tracker uses a TaskCompletionSource to signal when all critical operations have completed.

Operation A ────────────────┐

Operation B ────────┐       │

Operation C ────┐   │       │

                │   │       │

              SIGTERM        │

                │            │

                ▼            ▼

             Draining

                │

                ▼

        Wait for all operations

                │

                ▼

          TaskCompletionSource

                │

                ▼

            Shutdown

There is no polling; shutdown waits asynchronously until the tracker is completed.

Health Checks

The application should expose separate readiness and liveness endpoints.

| State | Readiness | Liveness |

|---|---:|---:|

| Normal | 200 | 200 |

| Draining | 503 | 200 |

Why?

• Readiness 503 tells Kubernetes/load balancers to stop sending new requests.

• Liveness 200 keeps the process alive while critical operations finish.

Kubernetes

Example probes:

livenessProbe:

  httpGet:

    path: /health/live

    port: 8080

readinessProbe:

  httpGet:

    path: /health/ready

    port: 8080

Configure a suitable termination grace period:

spec:

  terminationGracePeriodSeconds: 60

The value should be based on the maximum expected duration of your critical operations.

Recommended Usage

Track only operations that are genuinely business-critical, such as:

• Payment processing

• Booking/order confirmation

• Important external API calls

• Publishing critical events

• Completing transactional workflows

Normal read operations generally do not need to be tracked.

Key Design

SIGTERM

   │

   ▼

Start draining

   │

   ├── Readiness → 503

   └── Liveness  → 200

   │

   ▼

Wait for critical operations

   │

   ▼

All operations completed

   │

   ▼

Shutdown application

This allows the application to stop receiving new traffic without abruptly terminating important business operations.

"""