[Do not review] DRAFT: Testing hot-reload pipeline test flakiness fix

src/Cli.Tests/EndToEndTests.cs

@@ -367,7 +367,7 @@ public void TestUpdateGraphQLPathRuntimeSettings(string path, bool isSuccess)
     /// Ensures that invalid links provided for Cors.Origins result in failed engine startup 
     /// due to validation failure.
     /// </summary>
-    [Ignore]
+
     [DataTestMethod]
     [DataRow("http://locahost1 https://localhost2", true, DisplayName = "Success in updating Host.Cors.Origins.")]
     public void TestUpdateHostCorsOriginsRuntimeSettings(string path, bool isSuccess)

src/Config/DabConfigEvents.cs

@@ -15,8 +15,10 @@ public static class DabConfigEvents
     public const string POSTGRESQL_QUERY_EXECUTOR_ON_CONFIG_CHANGED = "POSTGRESQL_QUERY_EXECUTOR_ON_CONFIG_CHANGED";
     public const string DOCUMENTOR_ON_CONFIG_CHANGED = "DOCUMENTOR_ON_CONFIG_CHANGED";
     public const string AUTHZ_RESOLVER_ON_CONFIG_CHANGED = "AUTHZ_RESOLVER_ON_CONFIG_CHANGED";
-    public const string GRAPHQL_SCHEMA_REFRESH_ON_CONFIG_CHANGED = "GRAPHQL_SCHEMA_REFRESH_ON_CONFIG_CHANGED";
+    //public const string GRAPHQL_SCHEMA_REFRESH_ON_CONFIG_CHANGED = "GRAPHQL_SCHEMA_REFRESH_ON_CONFIG_CHANGED";
     public const string GRAPHQL_SCHEMA_EVICTION_ON_CONFIG_CHANGED = "GRAPHQL_SCHEMA_EVICTION_ON_CONFIG_CHANGED";
     public const string GRAPHQL_SCHEMA_CREATOR_ON_CONFIG_CHANGED = "GRAPHQL_SCHEMA_CREATOR_ON_CONFIG_CHANGED";
     public const string LOG_LEVEL_INITIALIZER_ON_CONFIG_CHANGE = "LOG_LEVEL_INITIALIZER_ON_CONFIG_CHANGE";
+    // Fired after all other hot-reload events have been invoked.
+    public const string HOT_RELOAD_ALL_DONE = "HOT_RELOAD_ALL_DONE";
 }

src/Config/HotReloadEventHandler.cs

@@ -32,9 +32,11 @@ public HotReloadEventHandler()
             { DOCUMENTOR_ON_CONFIG_CHANGED, null },
             { AUTHZ_RESOLVER_ON_CONFIG_CHANGED, null },
             { GRAPHQL_SCHEMA_CREATOR_ON_CONFIG_CHANGED, null },
-            { GRAPHQL_SCHEMA_REFRESH_ON_CONFIG_CHANGED, null },
+            //{ GRAPHQL_SCHEMA_REFRESH_ON_CONFIG_CHANGED, null },
             { GRAPHQL_SCHEMA_EVICTION_ON_CONFIG_CHANGED, null },
-            { LOG_LEVEL_INITIALIZER_ON_CONFIG_CHANGE, null }
+            { LOG_LEVEL_INITIALIZER_ON_CONFIG_CHANGE, null },
+            // Completion signal for tests/consumers that need to await full hot-reload lifecycle.
+            { HOT_RELOAD_ALL_DONE, null }
         };
     }
 

src/Config/RuntimeConfigLoader.cs

@@ -107,11 +107,14 @@ protected void SignalConfigChanged(string message = "")
             // Order of event firing matters: Eviction must be done before creating a new schema and then updating the schema.
             OnConfigChangedEvent(new HotReloadEventArgs(GRAPHQL_SCHEMA_EVICTION_ON_CONFIG_CHANGED, message));
             OnConfigChangedEvent(new HotReloadEventArgs(GRAPHQL_SCHEMA_CREATOR_ON_CONFIG_CHANGED, message));
-            OnConfigChangedEvent(new HotReloadEventArgs(GRAPHQL_SCHEMA_REFRESH_ON_CONFIG_CHANGED, message));
+            //OnConfigChangedEvent(new HotReloadEventArgs(GRAPHQL_SCHEMA_REFRESH_ON_CONFIG_CHANGED, message));
         }
 
         // Log Level Initializer is outside of if statement as it can be updated on both development and production mode.
         OnConfigChangedEvent(new HotReloadEventArgs(LOG_LEVEL_INITIALIZER_ON_CONFIG_CHANGE, message));
+
+        // Finally, fire completion signal so tests can deterministically await full hot-reload.
+        OnConfigChangedEvent(new HotReloadEventArgs(HOT_RELOAD_ALL_DONE, message));
     }
 
     /// <summary>

src/Core/Configurations/RuntimeConfigProvider.cs

@@ -64,11 +64,30 @@ public RuntimeConfigProvider(RuntimeConfigLoader runtimeConfigLoader)
     /// </seealso>
     private void RaiseChanged()
     {
-        //First use of GetConfig during hot reload, in order to do validation of
-        //config file before any changes are made for hot reload.
-        //In case validation fails, an exception will be thrown and hot reload will be canceled.
+        // First, validate and apply the hot-reloaded config (sets LKG or restores previous).
         ValidateConfig();
 
+        // Ensure dependent components (metadata providers, schema, etc.) are refreshed BEFORE signalling the change token.
+        // This makes consumers that await the provider's change token observe the updated state deterministically.
+        // Handlers are invoked synchronously here to avoid races.
+        try
+        {
+            if (_configLoader.RuntimeConfig is not null && RuntimeConfigLoadedHandlers.Count > 0)
+            {
+                InvokeConfigLoadedHandlersAsync().GetAwaiter().GetResult();
+            }
+        }
+        catch (Exception)
+        {
+            // If handler refresh fails, restore LKG and propagate a validation error similar to ValidateConfig failure path.
+            _configLoader.RestoreLkgConfig();
+            throw new DataApiBuilderException(
+                message: "Failed to refresh runtime components after hot reload.",
+                statusCode: HttpStatusCode.ServiceUnavailable,
+                subStatusCode: DataApiBuilderException.SubStatusCodes.ErrorInInitialization);
+        }
+
+        // Finally, rotate and signal the change token so listeners can react.
         DabChangeToken previousToken = Interlocked.Exchange(ref _changeToken, new DabChangeToken());
         previousToken.SignalChange();
     }

src/Core/Properties/launchSettings.json

@@ -0,0 +1,8 @@
+{
+  "profiles": {
+    "Azure.DataApiBuilder.Core": {
+      "commandName": "Project",
+      "commandLineArgs": "validate"
+    }
+  }
+}

src/Core/Services/MetadataProviders/MetadataProviderFactory.cs

@@ -62,10 +62,34 @@ private void ConfigureMetadataProviders()
 
         public void OnConfigChanged(object? sender, HotReloadEventArgs args)
         {
+            // Build new providers from current config
+            var newProviders = new Dictionary<string, ISqlMetadataProvider>();
+            foreach ((string dataSourceName, DataSource dataSource) in _runtimeConfigProvider.GetConfig().GetDataSourceNamesToDataSourcesIterator())
+            {
+                ISqlMetadataProvider provider = dataSource.DatabaseType switch
+                {
+                    DatabaseType.CosmosDB_NoSQL => new CosmosSqlMetadataProvider(_runtimeConfigProvider, _fileSystem),
+                    DatabaseType.MSSQL => new MsSqlMetadataProvider(_runtimeConfigProvider, _queryManagerFactory, _logger, dataSourceName, _isValidateOnly),
+                    DatabaseType.DWSQL => new MsSqlMetadataProvider(_runtimeConfigProvider, _queryManagerFactory, _logger, dataSourceName, _isValidateOnly),
+                    DatabaseType.PostgreSQL => new PostgreSqlMetadataProvider(_runtimeConfigProvider, _queryManagerFactory, _logger, dataSourceName, _isValidateOnly),
+                    DatabaseType.MySQL => new MySqlMetadataProvider(_runtimeConfigProvider, _queryManagerFactory, _logger, dataSourceName, _isValidateOnly),
+                    _ => throw new NotSupportedException(dataSource.DatabaseTypeNotSupportedMessage),
+                };
+                newProviders.Add(dataSourceName, provider);
+            }
+
+            // Initialize all new providers before making them visible
+            foreach ((_, ISqlMetadataProvider provider) in newProviders)
+            {
+                provider.InitializeAsync().GetAwaiter().GetResult();
+            }
+
+            // Atomically swap the provider set
             _metadataProviders.Clear();
-            ConfigureMetadataProviders();
-            // Blocks the current thread until initialization is finished.
-            this.InitializeAsync().GetAwaiter().GetResult();
+            foreach ((string name, ISqlMetadataProvider provider) in newProviders)
+            {
+                _metadataProviders[name] = provider;
+            }
         }
 
         /// <inheritdoc />

src/Service.Tests/Configuration/ConfigurationTests.cs

@@ -3068,7 +3068,6 @@ private static async void ValidateMutationSucceededAtDbLayer(TestServer server,
         /// </summary>
         /// <param name="entityType">Type of the entity</param>
         /// <param name="requestPath">Request path for performing POST API requests on the entity</param>
-        [Ignore]
         [DataTestMethod]
         [TestCategory(TestCategory.MSSQL)]
         [DataRow(EntitySourceType.Table, "/api/Book", DisplayName = "Location Header validation - Table, Base Route not configured")]
@@ -3920,7 +3919,7 @@ public void LogLevelSerialization(LogLevel expectedLevel)
         /// <summary>
         /// Tests different log level filters that are valid and check that they are deserialized correctly
         /// </summary>
-        [Ignore]
+
         [DataTestMethod]
         [TestCategory(TestCategory.MSSQL)]
         [DataRow(LogLevel.Trace, typeof(RuntimeConfigValidator))]

src/Service.Tests/Configuration/HotReload/AuthorizationResolverHotReloadTests.cs

@@ -32,7 +32,7 @@ public class AuthorizationResolverHotReloadTests
     /// and uses the same test server instance.
     /// This test is currently flakey, failing intermittently in our pipeline, and is therefore ignored.
     /// </summary>
-    [Ignore]
+
     [TestMethod]
     [DoNotParallelize]
     [TestCategory(TestCategory.MSSQL)]