Why should you care?

Global ERP customers require active-passive regions and evidence that AI workloads meet internal cloud standards.

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.Infrastructure/RegionFailoverChatClient.cs
public sealed class RegionFailoverChatClient(IChatClient primary, IChatClient secondary, ILogger log) : IChatClient
{
    public async Task<ChatResponse> GetResponseAsync(
        IEnumerable<ChatMessage> messages, ChatOptions? options = null, CancellationToken ct = default)
    {
        try { return await primary.GetResponseAsync(messages, options, ct); }
        catch (HttpRequestException ex)
        {
            log.LogWarning(ex, "Primary region failed; failing over");
            return await secondary.GetResponseAsync(messages, options, ct);
        }
    }
}

What happened?

• RegionFailoverChatClient tries East US Azure OpenAI first, falls back to West Europe on HTTP failure for HA.
• Follow the steps below — typing the code yourself is the fastest way to learn.

Practice next

1. Deploy paired OpenAI resources in two regions.
2. Register RegionFailoverChatClient as IChatClient decorator.
3. Run GameDay failover drill quarterly.
4. Add health-based proactive failover before hard errors.
5. Use Azure Front Door health probes for regional API traffic.

Remember

Enterprise infra includes multi-region model failover. Decorate IChatClient for automatic secondary routing. Replicate or tolerate lag on vector stores during DR.

Regional Azure outage

East US OpenAI degraded during ERP month-end.

Outcome: RegionFailoverChatClient routes to West Europe; finance close completes on time.