✨ Gateway readiness check

DysonNetwork.Gateway/Health/GatewayConstant.cs

@@ -0,0 +1,9 @@
+namespace DysonNetwork.Gateway.Health;
+
+public abstract class GatewayConstant
+{
+    public static readonly string[] ServiceNames = ["ring", "pass", "drive", "sphere", "develop", "insight", "zone"];
+    
+    // Core services stands with w/o these services the functional of entire app will broke.
+    public static readonly string[] CoreServiceNames = ["ring", "pass", "drive", "sphere"];
+}

DysonNetwork.Gateway/Health/GatewayHealthAggregator.cs

@@ -0,0 +1,60 @@
+using NodaTime;
+
+namespace DysonNetwork.Gateway.Health;
+
+public class GatewayHealthAggregator(IHttpClientFactory httpClientFactory, GatewayReadinessStore store)
+    : BackgroundService
+{
+    private async Task<ServiceHealthState> CheckService(string serviceName)
+    {
+        var client = httpClientFactory.CreateClient("health");
+        var now = SystemClock.Instance.GetCurrentInstant();
+
+        try
+        {
+            // Use the service discovery to lookup service
+            // The service defaults give every single service a health endpoint that we can use here
+            using var response = await client.GetAsync($"http://{serviceName}/health");
+
+            if (response.IsSuccessStatusCode)
+            {
+                return new ServiceHealthState(
+                    serviceName,
+                    true,
+                    now,
+                    null
+                );
+            }
+
+            return new ServiceHealthState(
+                serviceName,
+                false,
+                now,
+                $"StatusCode: {(int)response.StatusCode}"
+            );
+        }
+        catch (Exception ex)
+        {
+            return new ServiceHealthState(
+                serviceName,
+                false,
+                now,
+                ex.Message
+            );
+        }
+    }
+
+    protected override async Task ExecuteAsync(CancellationToken stoppingToken)
+    {
+        while (!stoppingToken.IsCancellationRequested)
+        {
+            foreach (var service in GatewayConstant.ServiceNames)
+            {
+                var result = await CheckService(service);
+                store.Update(result);
+            }
+
+            await Task.Delay(TimeSpan.FromSeconds(5), stoppingToken);
+        }
+    }
+}

DysonNetwork.Gateway/Health/GatewayReadinessMiddleware.cs

@@ -0,0 +1,35 @@
+namespace DysonNetwork.Gateway.Health;
+
+using Microsoft.AspNetCore.Http;
+
+public sealed class GatewayReadinessMiddleware(RequestDelegate next)
+{
+    public async Task InvokeAsync(HttpContext context, GatewayReadinessStore store)
+    {
+        if (context.Request.Path.StartsWithSegments("/health"))
+        {
+            await next(context);
+            return;
+        }
+
+        var readiness = store.Current;
+
+        // Only core services participate in readiness gating
+        var notReadyCoreServices = readiness.Services
+            .Where(kv => GatewayConstant.CoreServiceNames.Contains(kv.Key))
+            .Where(kv => !kv.Value.IsHealthy)
+            .Select(kv => kv.Key)
+            .ToArray();
+
+        if (notReadyCoreServices.Length > 0)
+        {
+            context.Response.StatusCode = StatusCodes.Status503ServiceUnavailable;
+            var unavailableServices = string.Join(", ", notReadyCoreServices);
+            context.Response.Headers["X-NotReady-Services"] = unavailableServices;
+            await context.Response.WriteAsync("Solar Network is warming up. Try again later please.");
+            return;
+        }
+
+        await next(context);
+    }
+}

DysonNetwork.Gateway/Health/GatewayReadinessStore.cs

@@ -0,0 +1,76 @@
+using NodaTime;
+
+namespace DysonNetwork.Gateway.Health;
+
+public record ServiceHealthState(
+    string ServiceName,
+    bool IsHealthy,
+    Instant LastChecked,
+    string? Error
+);
+
+public record GatewayReadinessState(
+    bool IsReady,
+    IReadOnlyDictionary<string, ServiceHealthState> Services,
+    Instant LastUpdated
+);
+
+public class GatewayReadinessStore
+{
+    private readonly Lock _lock = new();
+
+    private readonly Dictionary<string, ServiceHealthState> _services = new();
+
+    public GatewayReadinessState Current { get; private set; } = new(
+        IsReady: false,
+        Services: new Dictionary<string, ServiceHealthState>(),
+        LastUpdated: SystemClock.Instance.GetCurrentInstant()
+    );
+
+    public IReadOnlyCollection<string> ServiceNames => _services.Keys;
+
+    public GatewayReadinessStore()
+    {
+        InitializeServices(GatewayConstant.ServiceNames);
+    }
+
+    private void InitializeServices(IEnumerable<string> serviceNames)
+    {
+        lock (_lock)
+        {
+            _services.Clear();
+
+            foreach (var name in serviceNames)
+            {
+                _services[name] = new ServiceHealthState(
+                    name,
+                    IsHealthy: false,
+                    LastChecked: SystemClock.Instance.GetCurrentInstant(),
+                    Error: "Not checked yet"
+                );
+            }
+
+            RecalculateLocked();
+        }
+    }
+
+    public void Update(ServiceHealthState state)
+    {
+        lock (_lock)
+        {
+            _services[state.ServiceName] = state;
+            RecalculateLocked();
+        }
+    }
+
+    private void RecalculateLocked()
+    {
+        var isReady = _services.Count > 0 && _services.Values.All(s => s.IsHealthy);
+
+        Current = new GatewayReadinessState(
+            IsReady: isReady,
+            Services: new Dictionary<string, ServiceHealthState>(_services),
+            LastUpdated: SystemClock.Instance.GetCurrentInstant()
+        );
+    }
+}

DysonNetwork.Gateway/Health/GatewayStatusController.cs

@@ -0,0 +1,14 @@
+using Microsoft.AspNetCore.Mvc;
+
+namespace DysonNetwork.Gateway.Health;
+
+[ApiController]
+[Route("/health")]
+public class GatewayStatusController(GatewayReadinessStore readinessStore) : ControllerBase
+{
+    [HttpGet]
+    public ActionResult<GatewayReadinessState> GetHealthStatus()
+    {
+        return Ok(readinessStore.Current);
+    }
+}

DysonNetwork.Gateway/Program.cs

@@ -1,7 +1,12 @@
+using System.Text.Json;
+using System.Text.Json.Serialization;
 using System.Threading.RateLimiting;
+using DysonNetwork.Gateway.Health;
 using DysonNetwork.Shared.Http;
 using Yarp.ReverseProxy.Configuration;
 using Microsoft.AspNetCore.HttpOverrides;
+using NodaTime;
+using NodaTime.Serialization.SystemTextJson;
 
 var builder = WebApplication.CreateBuilder(args);
 
@@ -9,17 +14,19 @@ builder.AddServiceDefaults();
 
 builder.ConfigureAppKestrel(builder.Configuration, maxRequestBodySize: long.MaxValue, enableGrpc: false);
 
+builder.Services.AddSingleton<GatewayReadinessStore>();
+builder.Services.AddHostedService<GatewayHealthAggregator>();
+
 builder.Services.AddCors(options =>
 {
-    options.AddDefaultPolicy(
-        policy =>
-        {
-            policy.SetIsOriginAllowed(origin => true)
-                .AllowAnyMethod()
-                .AllowAnyHeader()
-                .AllowCredentials()
-                .WithExposedHeaders("X-Total");
-        });
+    options.AddDefaultPolicy(policy =>
+    {
+        policy.SetIsOriginAllowed(origin => true)
+            .AllowAnyMethod()
+            .AllowAnyHeader()
+            .AllowCredentials()
+            .WithExposedHeaders("X-Total");
+    });
 });
 
 builder.Services.AddRateLimiter(options =>
@@ -40,23 +47,22 @@ builder.Services.AddRateLimiter(options =>
     });
 
     options.OnRejected = async (context, token) =>
-        {
-            // Log the rejected IP
-            var logger = context.HttpContext.RequestServices
-                .GetRequiredService<ILoggerFactory>()
-                .CreateLogger("RateLimiter");
+    {
+        // Log the rejected IP
+        var logger = context.HttpContext.RequestServices
+            .GetRequiredService<ILoggerFactory>()
+            .CreateLogger("RateLimiter");
 
-            var ip = context.HttpContext.Connection.RemoteIpAddress?.ToString() ?? "unknown";
-            logger.LogWarning("Rate limit exceeded for IP: {IP}", ip);
+        var ip = context.HttpContext.Connection.RemoteIpAddress?.ToString() ?? "unknown";
+        logger.LogWarning("Rate limit exceeded for IP: {IP}", ip);
 
-            // Respond to the client
-            context.HttpContext.Response.StatusCode = StatusCodes.Status429TooManyRequests;
-            await context.HttpContext.Response.WriteAsync(
-                "Rate limit exceeded. Try again later.", token);
-        };
+        // Respond to the client
+        context.HttpContext.Response.StatusCode = StatusCodes.Status429TooManyRequests;
+        await context.HttpContext.Response.WriteAsync(
+            "Rate limit exceeded. Try again later.", token);
+    };
 });
 
-var serviceNames = new[] { "ring", "pass", "drive", "sphere", "develop", "insight", "zone" };
 
 var specialRoutes = new[]
 {
@@ -86,7 +92,7 @@ var specialRoutes = new[]
     }
 };
 
-var apiRoutes = serviceNames.Select(serviceName =>
+var apiRoutes = GatewayConstant.ServiceNames.Select(serviceName =>
 {
     var apiPath = serviceName switch
     {
@@ -105,7 +111,7 @@ var apiRoutes = serviceNames.Select(serviceName =>
     };
 });
 
-var swaggerRoutes = serviceNames.Select(serviceName => new RouteConfig
+var swaggerRoutes = GatewayConstant.ServiceNames.Select(serviceName => new RouteConfig
 {
     RouteId = $"{serviceName}-swagger",
     ClusterId = serviceName,
@@ -119,7 +125,7 @@ var swaggerRoutes = serviceNames.Select(serviceName => new RouteConfig
 
 var routes = specialRoutes.Concat(apiRoutes).Concat(swaggerRoutes).ToArray();
 
-var clusters = serviceNames.Select(serviceName => new ClusterConfig
+var clusters = GatewayConstant.ServiceNames.Select(serviceName => new ClusterConfig
 {
     ClusterId = serviceName,
     HealthCheck = new HealthCheckConfig
@@ -147,7 +153,14 @@ builder.Services
     .LoadFromMemory(routes, clusters)
     .AddServiceDiscoveryDestinationResolver();
 
-builder.Services.AddControllers();
+builder.Services.AddControllers().AddJsonOptions(options =>
+{
+    options.JsonSerializerOptions.NumberHandling = JsonNumberHandling.AllowNamedFloatingPointLiterals;
+    options.JsonSerializerOptions.PropertyNamingPolicy = JsonNamingPolicy.SnakeCaseLower;
+    options.JsonSerializerOptions.DictionaryKeyPolicy = JsonNamingPolicy.SnakeCaseLower;
+
+    options.JsonSerializerOptions.ConfigureForNodaTime(DateTimeZoneProviders.Tzdb);
+});
 
 var app = builder.Build();
 
@@ -155,14 +168,16 @@ var forwardedHeadersOptions = new ForwardedHeadersOptions
 {
     ForwardedHeaders = ForwardedHeaders.All
 };
-forwardedHeadersOptions.KnownNetworks.Clear();
+forwardedHeadersOptions.KnownIPNetworks.Clear();
 forwardedHeadersOptions.KnownProxies.Clear();
 app.UseForwardedHeaders(forwardedHeadersOptions);
 
 app.UseCors();
 
+app.UseMiddleware<GatewayReadinessMiddleware>();
+
 app.MapReverseProxy().RequireRateLimiting("fixed");
 
 app.MapControllers();
 
-app.Run();
+app.Run();