Retry transient Windows bundle directory moves

src/Aspire.Cli/Bundles/BundleService.cs

@@ -49,6 +49,17 @@ internal sealed class BundleService(
     /// </summary>
     internal const string BadSuffixPrefix = ".bad.";
 
+    // Windows scanners can briefly open freshly extracted files without delete sharing, causing
+    // Directory.Move to fail with one of these HRESULTs. Unix permits renames while files are open,
+    // and the exact error list avoids delaying deterministic failures such as ERROR_DISK_FULL.
+    // See https://learn.microsoft.com/windows/win32/debug/system-error-codes--0-499-
+    private const int AccessDeniedHResult = unchecked((int)0x80070005);
+    private const int SharingViolationHResult = unchecked((int)0x80070020);
+    private const int LockViolationHResult = unchecked((int)0x80070021);
+
+    private static readonly TimeSpan s_directoryMoveMaxRetryElapsed = TimeSpan.FromSeconds(3);
+    private static readonly TimeSpan s_directoryMoveMaxRetryDelay = TimeSpan.FromSeconds(1);
+
     /// <inheritdoc/>
     public bool IsBundle => payloadProvider.HasPayload;
 
@@ -349,7 +360,12 @@ private async Task<bool> ExtractVersionedLayoutAsync(
 
         try
         {
-            Directory.Move(tempDir, activeVersionDir);
+            await MoveDirectoryWithRetryAsync(tempDir, activeVersionDir, cancellationToken).ConfigureAwait(false);
+        }
+        catch (OperationCanceledException)
+        {
+            FileDeleteHelper.TryDeleteDirectory(tempDir);
+            throw;
         }
         catch (Exception ex) when (ex is IOException or UnauthorizedAccessException)
         {
@@ -377,6 +393,50 @@ private async Task<bool> ExtractVersionedLayoutAsync(
         return true;
     }
 
+    /// <summary>
+    /// Moves a directory, retrying transient Windows file-lock failures with bounded backoff.
+    /// </summary>
+    internal async Task MoveDirectoryWithRetryAsync(string sourcePath, string destinationPath, CancellationToken cancellationToken)
+    {
+        var delay = TimeSpan.FromMilliseconds(100);
+        var retryCount = 0;
+        var stopwatch = Stopwatch.StartNew();
+
+        while (true)
+        {
+            try
+            {
+                Directory.Move(sourcePath, destinationPath);
+                return;
+            }
+            catch (Exception ex) when (IsRetryableDirectoryMoveException(ex, environment.IsWindows()) && stopwatch.Elapsed < s_directoryMoveMaxRetryElapsed)
+            {
+                retryCount++;
+                logger.LogDebug(
+                    "Directory move from {SourcePath} to {DestinationPath} failed with HRESULT {HResult}; retrying in {DelayMs}ms (retry {RetryCount}).",
+                    sourcePath,
+                    destinationPath,
+                    ex.HResult,
+                    delay.TotalMilliseconds,
+                    retryCount);
+                await Task.Delay(delay, cancellationToken).ConfigureAwait(false);
+                delay = TimeSpan.FromMilliseconds(Math.Min(
+                    delay.TotalMilliseconds * 2,
+                    s_directoryMoveMaxRetryDelay.TotalMilliseconds));
+            }
+        }
+    }
+
+    /// <summary>
+    /// Determines whether a directory move exception represents a transient Windows file lock.
+    /// </summary>
+    internal static bool IsRetryableDirectoryMoveException(Exception exception, bool isWindows)
+    {
+        return isWindows &&
+            exception is IOException or UnauthorizedAccessException &&
+            exception.HResult is AccessDeniedHResult or SharingViolationHResult or LockViolationHResult;
+    }
+
     /// <inheritdoc/>
     public string? GetDefaultExtractDir(string processPath)
         => ComputeDefaultExtractDir(processPath, logger);

tests/Aspire.Cli.Tests/BundleServiceIntegrationTests.cs

@@ -196,6 +196,35 @@ public async Task ExtractAsync_CleansUpStaleVersionDirectories()
         }
     }
 
+    [Fact]
+    [SkipOnPlatform(TestPlatforms.Linux | TestPlatforms.OSX | TestPlatforms.FreeBSD, "Windows file sharing semantics are required.")]
+    public async Task MoveDirectoryWithRetryAsync_FileTemporarilyLocked_RetriesUntilReleased()
+    {
+        using var workspace = TemporaryWorkspace.CreateForCli(outputHelper);
+        var sourcePath = workspace.CreateDirectory("source").FullName;
+        var destinationPath = Path.Combine(workspace.WorkspaceRoot.FullName, "destination");
+        var lockedFilePath = Path.Combine(sourcePath, "locked.dll");
+        File.WriteAllText(lockedFilePath, "locked");
+
+        var service = CreateService(
+            new TestBundlePayloadProvider(CreateFakeBundlePayload()),
+            new TestLayoutDiscovery(workspace.WorkspaceRoot.FullName));
+
+        using var lockedFile = new FileStream(lockedFilePath, FileMode.Open, FileAccess.Read, FileShare.Read);
+        var moveTask = service.MoveDirectoryWithRetryAsync(
+            sourcePath,
+            destinationPath,
+            TestContext.Current.CancellationToken);
+
+        Assert.False(moveTask.IsCompleted);
+
+        lockedFile.Dispose();
+        await moveTask;
+
+        Assert.False(Directory.Exists(sourcePath));
+        Assert.True(Directory.Exists(destinationPath));
+    }
+
     [Fact]
     public async Task EnsureExtractedAndAcquireLayoutAsync_ReturnsVersionRootedLayoutAndSkipsLeasedCleanup()
     {

tests/Aspire.Cli.Tests/BundleServiceTests.cs

@@ -149,6 +149,34 @@ public void IsVersionedLayoutValid_RequiresDcpDirectory()
         Assert.False(BundleService.IsVersionedLayoutValid(dir));
     }
 
+    [Theory]
+    [InlineData(unchecked((int)0x80070005), true)] // ERROR_ACCESS_DENIED
+    [InlineData(unchecked((int)0x80070020), true)] // ERROR_SHARING_VIOLATION
+    [InlineData(unchecked((int)0x80070021), true)] // ERROR_LOCK_VIOLATION
+    [InlineData(unchecked((int)0x80070027), false)] // ERROR_HANDLE_DISK_FULL
+    [InlineData(unchecked((int)0x80070070), false)] // ERROR_DISK_FULL
+    [InlineData(unchecked((int)0x800700B7), false)] // ERROR_ALREADY_EXISTS
+    public void IsRetryableDirectoryMoveException_OnlyRetriesTransientWindowsLockErrors(int hresult, bool expected)
+    {
+        var exception = new IOException("Directory move failed.", hresult);
+
+        Assert.Equal(expected, BundleService.IsRetryableDirectoryMoveException(exception, isWindows: true));
+    }
+
+    [Fact]
+    public void IsRetryableDirectoryMoveException_RetriesUnauthorizedAccess()
+    {
+        Assert.True(BundleService.IsRetryableDirectoryMoveException(new UnauthorizedAccessException(), isWindows: true));
+    }
+
+    [Fact]
+    public void IsRetryableDirectoryMoveException_DoesNotRetryOnNonWindows()
+    {
+        var exception = new IOException("Directory move failed.", unchecked((int)0x80070020));
+
+        Assert.False(BundleService.IsRetryableDirectoryMoveException(exception, isWindows: false));
+    }
+
     [Fact]
     public void TryCleanupStaleVersions_RemovesNonActiveVersionsAndStaleTempDirs()
     {