Why should you care?

CRM and clinical workflows treat copilot as infrastructure; single VM API is unacceptable SLA.

See it live — copy this example

Paste into an ASP.NET Core 8+ / AgentNest project, then run with dotnet run (set your API keys in user-secrets).

// AgentNest.Api/Program.cs — redundant health
builder.Services.AddHealthChecks()
    .AddCheck<AiModelHealthCheck>("ai_model")
    .AddRedis(redisConn, name: "redis")
    .AddNpgSql(pgConn, name: "postgres");

app.MapHealthChecks("/health/ready", new HealthCheckOptions
{
    Predicate = r => r.Tags.Contains("ready"),
    ResponseWriter = UIResponseWriter.WriteHealthCheckUIResponse
});

What happened?

• Combined health checks fail readiness if model, Redis, or Postgres unhealthy — load balancer stops routing to bad pods.
• Follow the steps below — typing the code yourself is the fastest way to learn.

Practice next

1. Run minimum three API replicas across availability zones.
2. Redis cluster mode or Azure Cache with replication.
3. Test failover by blocking primary OpenAI in staging.
4. Add active-active API in two regions with Front Door.
5. Persist critical queue messages to geo-redundant Service Bus.

Remember

HA requires multi-pod API, replicated Redis, DB redundancy. Readiness includes model and dependency health. Test regional OpenAI failover regularly.

Redis node failure

Redis primary fails during business hours.

Outcome: Replica promoted; API readiness flips briefly; CRM chat resumes under 2 minutes.