[stress] Sync namespace federated credentials periodically and on sta…

…rtup (#9063) * Watcher namespace fed cred management fixes * Update stress watcher dockerfile to use net 8 * Add back delete code, remove terminating skip * Bump cluster version to 1.29.8
Azure · Oct 3, 2024 · c32711b · c32711b
1 parent a76f3d2
commit c32711b
Show file tree

Hide file tree

Showing 5 changed files with 98 additions and 12 deletions.
diff --git a/tools/stress-cluster/chaos/examples/stress-deployment-example/Chart.lock b/tools/stress-cluster/chaos/examples/stress-deployment-example/Chart.lock
@@ -1,6 +1,6 @@
 dependencies:
 - name: stress-test-addons
   repository: https://stresstestcharts.blob.core.windows.net/helm/
-  version: 0.3.2
-digest: sha256:6eee71a7e8a4c0dc06d5fbbce39ef63237a0db0b7fc2da66e98e96b68985b764
-generated: "2024-05-23T11:37:41.371010465-04:00"
+  version: 0.3.3
+digest: sha256:1cffb5ed8ea74953ab7611f9e2de2163af2c3f0918afb9928f71210da9c19a4a
+generated: "2024-10-02T16:18:41.429777815-04:00"
diff --git a/tools/stress-cluster/cluster/azure/cluster/cluster.bicep b/tools/stress-cluster/cluster/azure/cluster/cluster.bicep
@@ -13,7 +13,7 @@ param updateNodes bool = false
 // monitoring parameters
 param workspaceId string
 
-var kubernetesVersion = '1.29.4'
+var kubernetesVersion = '1.29.8'
 var nodeResourceGroup = 'rg-nodes-${dnsPrefix}-${clusterName}-${groupSuffix}'
 
 var systemAgentPool = {

diff --git a/tools/stress-cluster/services/Stress.Watcher/Dockerfile b/tools/stress-cluster/services/Stress.Watcher/Dockerfile
@@ -1,8 +1,8 @@
-FROM mcr.microsoft.com/dotnet/sdk:6.0-cbl-mariner2.0 AS build
+FROM mcr.microsoft.com/dotnet/sdk:8.0-cbl-mariner2.0 AS build
 
 COPY ./src /src
 
-RUN cd /src && dotnet publish -c Release -o /stresswatcher -r linux-x64 -f net6.0 -p:PublishSingleFile=true --self-contained
+RUN cd /src && dotnet publish -c Release -o /stresswatcher -r linux-x64 -f net8.0 -p:PublishSingleFile=true --self-contained
 
 FROM mcr.microsoft.com/azure-cli:cbl-mariner2.0
 

diff --git a/tools/stress-cluster/services/Stress.Watcher/src/NamespaceEventHandler.cs b/tools/stress-cluster/services/Stress.Watcher/src/NamespaceEventHandler.cs
@@ -28,6 +28,8 @@ public class NamespaceEventHandler
         // Concurrent Federated Identity Credentials writes under the same managed identity are not supported
         private static readonly SemaphoreSlim FederatedCredentialWriteSemaphore = new(1, 1);
 
+        private Dictionary<string, UserAssignedIdentityResource> WorkloadAppCache = [];
+
         public List<string> WorkloadAppPool;
         public string WorkloadAppIssuer;
 
@@ -62,6 +64,57 @@ public NamespaceEventHandler(
                 .CreateLogger();
         }
 
+        public async Task SyncCredentials()
+        {
+            try
+            {
+                Logger.Information($"Waiting for federated credential write semaphore");
+                await FederatedCredentialWriteSemaphore.WaitAsync();
+                await _syncCredentials();
+            }
+            finally
+            {
+                Logger.Information("Releasing federated credential write semaphore");
+                FederatedCredentialWriteSemaphore.Release();
+            }
+        }
+
+        public async Task _syncCredentials()
+        {
+            Logger.Information("Syncing namespaced federated credentials, this may take a minute...");
+
+            var namespaces = await Client.ListNamespaceAsync();
+            foreach (var app in WorkloadAppPool)
+            {
+                var resourceId = UserAssignedIdentityResource.CreateResourceIdentifier(SubscriptionId, ClusterGroup, app);
+                var userAssignedIdentity = ArmClient.GetUserAssignedIdentityResource(resourceId);
+                var identityResource = await userAssignedIdentity.GetAsync();
+                var fedCreds = userAssignedIdentity.GetFederatedIdentityCredentials();
+                await foreach (var item in fedCreds.GetAllAsync())
+                {
+                    if (!namespaces.Items.Any(ns => item.Data.Name == CreateFederatedIdentityCredentialName(ns)))
+                    {
+                        if (!string.IsNullOrEmpty(WatchNamespace) && item.Data.Name != CreateFederatedIdentityCredentialName(WatchNamespace))
+                        {
+                            Logger.Information($"Skipping delete federated credential '{item.Data.Name}' because it is not the watched namespace '{WatchNamespace}'");
+                            continue;
+                        }
+                        // Only perform delete operations for namespace state that may have changed if the watcher was not running.
+                        // Any create operations will be handled after initialization as the watch stream processes all active namespaces on startup
+                        Logger.Information($"Deleting federated identity credential '{item.Data.Name}' for managed identity '{app}' as the corresponding namespace no longer exists.");
+                        WorkloadAppCache.Remove(item.Data.Name);
+                        var lro = await item.DeleteAsync(Azure.WaitUntil.Completed);
+                    }
+                    else
+                    {
+                        WorkloadAppCache[item.Data.Name] = identityResource.Value;
+                    }
+                }
+            }
+
+            Logger.Information($"Federated credential sync complete. Cached {WorkloadAppCache.Count} federated credentials.");
+        }
+
         public async Task Watch(CancellationToken cancellationToken)
         {
             string resourceVersion = null;
@@ -116,7 +169,7 @@ public async Task Watch(CancellationToken cancellationToken)
 
         public void HandleNamespaceEvent(WatchEventType eventType, V1Namespace ns)
         {
-            if (ExcludedNamespaces.Contains(ns.Name()))
+            if (ExcludedNamespaces.Contains(ns.Name()) || string.IsNullOrEmpty(ns.Name()))
             {
                 return;
             }
@@ -156,7 +209,12 @@ public void HandleNamespaceEvent(WatchEventType eventType, V1Namespace ns)
 
         public string CreateFederatedIdentityCredentialName(V1Namespace ns)
         {
-            return $"stress-{ns.Name()}";
+            return CreateFederatedIdentityCredentialName(ns.Name());
+        }
+
+        public string CreateFederatedIdentityCredentialName(string ns)
+        {
+            return $"stress-{ns}";
         }
 
         public async Task InitializeWorkloadIdForNamespace(V1Namespace ns)
@@ -175,14 +233,21 @@ public async Task InitializeWorkloadIdForNamespace(V1Namespace ns)
             var identityData = await selectedWorkloadIdentity.GetAsync();
             var selectedWorkloadAppId = identityData.Value.Data.ClientId.ToString();
 
-            var meta = new V1ObjectMeta(){
+            var meta = new V1ObjectMeta()
+            {
                 Name = ns.Name(),
                 NamespaceProperty = ns.Name(),
                 Annotations = new Dictionary<string, string>(){
                     { "azure.workload.identity/client-id", selectedWorkloadAppId }
                 }
             };
             var serviceAccount = new V1ServiceAccount(metadata: meta);
+            var allAccounts = await Client.ListNamespacedServiceAccountAsync(ns.Name());
+            if (allAccounts.Items.Any(sa => sa.Name() == ns.Name()))
+            {
+                Logger.Information($"Service account '{ns.Name()}/{ns.Name()}' already exists, skipping creation.");
+                return;
+            }
             await Client.CreateNamespacedServiceAccountAsync(serviceAccount, ns.Name());
             Logger.Information($"Created service account '{ns.Name()}/{ns.Name()}' with workload client id '{selectedWorkloadAppId}'");
         }
@@ -200,6 +265,12 @@ public async Task<UserAssignedIdentityResource> CreateFederatedIdentityCredentia
             Logger.Information($"Waiting for federated credential write semaphore");
             await FederatedCredentialWriteSemaphore.WaitAsync();
 
+            if (WorkloadAppCache.ContainsKey(credentialName))
+            {
+                Logger.Information($"Found cache entry for federated credential {credentialName}, returning identity {WorkloadAppCache[credentialName].Data.ClientId}");
+                return await WorkloadAppCache[credentialName].GetAsync();
+            }
+
             foreach (var workloadApp in WorkloadAppPool)
             {
                 var userAssignedIdentityResourceId = UserAssignedIdentityResource.CreateResourceIdentifier(SubscriptionId, ClusterGroup, workloadApp);
@@ -246,15 +317,22 @@ public async Task<UserAssignedIdentityResource> CreateFederatedIdentityCredentia
             Logger.Information($"Creating/updating federated identity credential '{credentialName}' " +
                                $"with subject '{subject}' for managed identity '{selectedWorkloadApp}'");
             var lro = await federatedIdentityCredential.UpdateAsync(Azure.WaitUntil.Completed, fedCredData);
+            WorkloadAppCache[credentialName] = selectedIdentity;
             Logger.Information($"Created federated identity credential '{lro.Value.Data.Name}'");
 
             return selectedIdentity;
         }
 
         public async Task DeleteFederatedIdentityCredential(V1Namespace ns)
         {
+            Logger.Information($"Waiting for federated credential write semaphore");
+            await FederatedCredentialWriteSemaphore.WaitAsync();
+
             var credentialName = CreateFederatedIdentityCredentialName(ns);
             var workloadApp = "";
+
+            WorkloadAppCache.Remove(credentialName);
+
             foreach (var app in WorkloadAppPool)
             {
                 var resourceId = UserAssignedIdentityResource.CreateResourceIdentifier(SubscriptionId, ClusterGroup, app);
@@ -284,9 +362,6 @@ public async Task DeleteFederatedIdentityCredential(V1Namespace ns)
                     SubscriptionId, ClusterGroup, workloadApp, credentialName);
             var federatedIdentityCredential = ArmClient.GetFederatedIdentityCredentialResource(federatedIdentityCredentialResourceId);
 
-            Logger.Information($"Waiting for federated credential write semaphore");
-            await FederatedCredentialWriteSemaphore.WaitAsync();
-
             Logger.Information($"Deleting federated identity credential '{credentialName}' for managed identity '{workloadApp}'");
             var lro = await federatedIdentityCredential.DeleteAsync(Azure.WaitUntil.Completed);
             Logger.Information($"Deleted federated identity credential '{credentialName}'");

diff --git a/tools/stress-cluster/services/Stress.Watcher/src/Program.cs b/tools/stress-cluster/services/Stress.Watcher/src/Program.cs
@@ -79,6 +79,8 @@ static async Task Program(Options options)
             var namespaceEventHandler = new NamespaceEventHandler(
                 client, armClient, workloadConfig.SubscriptionId, workloadConfig.ClusterGroup,
                 workloadConfig.WorkloadAppPool, workloadConfig.WorkloadAppIssuer, options.Namespace);
+            await namespaceEventHandler.SyncCredentials();
+            _ = PollAndSyncCredentials(namespaceEventHandler, 288);  // poll every 12 hours
 
             var cts = new CancellationTokenSource();
             var taskList = new List<Task>
@@ -164,5 +166,14 @@ static WorkloadAuthConfig GetWorkloadConfigValues(Options options, Boolean isLoc
                 ClusterGroup = clusterGroup
             };
         }
+
+        static async Task PollAndSyncCredentials(NamespaceEventHandler namespaceHandler, int minutes)
+        {
+            while (true)
+            {
+                await Task.Delay(TimeSpan.FromMinutes(minutes));
+                await namespaceHandler.SyncCredentials();
+            }
+        }
     }
 }