Merge remote-tracking branch 'origin/enh/lock-resiliency' into enh/lock-resiliency

Author: sfmskywalker

src/modules/Elsa.Resilience.Core/Services/DefaultTransientExceptionStrategy.cs

@@ -56,14 +56,13 @@ public bool IsTransient(Exception exception)
             return true;
 
         // Check if the exception message contains any transient patterns
-        if (!string.IsNullOrEmpty(exception.Message))
-        {
-            foreach (var pattern in TransientExceptionMessagePatterns)
-            {
-                if (exception.Message.Contains(pattern, StringComparison.OrdinalIgnoreCase))
-                    return true;
-            }
-        }
+        var message = exception.Message;
+        
+        if (!string.IsNullOrEmpty(message) &&
+            TransientExceptionMessagePatterns
+                .Where(pattern => message.Contains(pattern, StringComparison.OrdinalIgnoreCase))
+                .Any())
+            return true;
 
         return false;
     }

src/modules/Elsa.Workflows.Runtime.Distributed/Services/DistributedWorkflowClient.cs

@@ -20,8 +20,12 @@ public class DistributedWorkflowClient(
     : IWorkflowClient
 {
     private readonly LocalWorkflowClient _localWorkflowClient = ActivatorUtilities.CreateInstance<LocalWorkflowClient>(serviceProvider, workflowInstanceId);
-    private readonly Lazy<ResiliencePipeline> _retryPipeline = new(() => CreateRetryPipeline(transientExceptionDetector, logger, workflowInstanceId));
+    private readonly ITransientExceptionDetector _transientExceptionDetector = transientExceptionDetector;
+    private readonly ILogger<DistributedWorkflowClient> _logger = logger;
+    private readonly Lazy<ResiliencePipeline> _retryPipeline = new(CreateRetryPipelineForInstance);
 
+    private ResiliencePipeline CreateRetryPipelineForInstance() =>
+        CreateRetryPipeline(_transientExceptionDetector, _logger, WorkflowInstanceId);
     public string WorkflowInstanceId => workflowInstanceId;
 
     public async Task<CreateWorkflowInstanceResponse> CreateInstanceAsync(CreateWorkflowInstanceRequest request, CancellationToken cancellationToken = default)
@@ -133,17 +137,19 @@ private static ResiliencePipeline CreateRetryPipeline(
         ILogger<DistributedWorkflowClient> logger,
         string workflowInstanceId)
     {
+        const int maxRetryAttempts = 3;
+
         return new ResiliencePipelineBuilder()
             .AddRetry(new()
             {
-                MaxRetryAttempts = 3,
+                MaxRetryAttempts = maxRetryAttempts,
                 Delay = TimeSpan.FromMilliseconds(500),
                 BackoffType = DelayBackoffType.Exponential,
                 UseJitter = true,
-                ShouldHandle = new PredicateBuilder().Handle<Exception>(transientExceptionDetector.IsTransient),
+                ShouldHandle = new PredicateBuilder().Handle<Exception>(ex => transientExceptionDetectionService.IsTransient(ex)),
                 OnRetry = args =>
                 {
-                    logger.LogWarning(args.Outcome.Exception, "Transient error acquiring lock for workflow instance {WorkflowInstanceId}. Attempt {AttemptNumber} of {MaxAttempts}.", workflowInstanceId, args.AttemptNumber + 1, 3);
+                    logger.LogWarning(args.Outcome.Exception, "Transient error acquiring lock for workflow instance {WorkflowInstanceId}. Attempt {AttemptNumber} of {MaxAttempts}.", workflowInstanceId, args.AttemptNumber + 1, maxRetryAttempts);
                     return ValueTask.CompletedTask;
                 }
             })

test/component/Elsa.Workflows.ComponentTests/Scenarios/DistributedLockResilience/DistributedLockResilienceTests.cs

@@ -81,6 +81,10 @@ private static ResiliencePipeline CreateRetryPipeline(ITransientExceptionDetecto
             .AddRetry(new()
             {
                 MaxRetryAttempts = MaxRetryAttempts,
+                // NOTE: The test retry policy intentionally differs from the production configuration.
+                // - We use a short, constant delay (10ms) to keep tests fast.
+                // - We disable jitter and exponential backoff to make timing deterministic and assertions stable.
+                //   The production pipeline uses a larger delay with exponential backoff and jitter for robustness.
                 Delay = TimeSpan.FromMilliseconds(10),
                 BackoffType = DelayBackoffType.Constant,
                 UseJitter = false,