Kian Thompson

This commit introduces a **new configuration capability** to manage retry behavior during **Fluid container attachment**, addressing scenarios where indefinite retries are undesirable for time-sensitive operations. It implements the `Fluid.Container.DisableCloseOnAttachFailure` flag within the **container-loader** module, allowing consumers to disable internal retries for the `attach` method and receive errors directly. This **enhancement** also extends the `runWithRetry` utility in **driver-utils** to accept a `maxRetries` parameter, providing more granular control over retry limits. The change empowers consumers to implement custom retry logic or fail faster, significantly impacting the responsiveness and error handling of time-sensitive container operations.

This commit **fixes a race condition** in the **summarizer lifecycle management** within the `container-runtime` package, specifically in `summaryManager.ts`. Previously, a stop request for the **summarizer** could be ignored if it was issued while the summarizer instance was still loading, leading to unexpected summarizer continuation and "SummarizerStopTimeout" errors. The change ensures that such pre-connection stop requests are now properly honored by adding a check immediately after the summarizer finishes loading and connects. This **improves the robustness and reliability of summarizer operations**, preventing scenarios where the summarizer continues to run despite being instructed to stop, especially during parent disconnect and reconnect sequences.

This commit introduces **timeout mechanisms** to improve the **robustness and reliability** of the **Fluid Framework summarizer** within the **container runtime**. It addresses scenarios where the summarizer process could hang indefinitely during startup or shutdown, preventing proper resource cleanup and the spawning of new summarizers. Specifically, timeouts are added to the `runCore` method in `packages/runtime/container-runtime/src/summary/summaryDelayLoadedModule/summarizer.ts` and to the `stop` and `dispose` methods in `packages/runtime/container-runtime/src/summary/summaryManager.ts`. This **bug fix** ensures the summarizer is forcefully stopped or cleaned up if it fails to complete operations within a specified duration, preventing unresponsive summarizers and enhancing the overall stability of document operations.

This commit **enhances telemetry** within the **container runtime's summarizer subsystem** by adding crucial diagnostic data to various events. Specifically, it introduces `opsWithoutSummary` to `SummaryStatus:Behind` and client election telemetry, and enriches `SummarizeFailed` and `UnexpectedSummarizeError` events with `isLastSummary` and `summarizeReason`. Furthermore, `summaryGenerator.ts` now tracks `nonRuntimeOpsSinceLastSummary` and `runtimeOpsSinceLastSummary` in its `summarizeEvent`. This **new capability** provides deeper insights into summarization attempts, failures, and operational delays, aiding in debugging and performance analysis of the summarization process.

This commit significantly **enhances telemetry** for **summarizer timeouts** across the **Fluid Framework runtime**. It introduces a `telemetryContext` to track the current summarization step, enabling the existing `SummarizeTimeout` event to include detailed information about the specific code path causing delays. Furthermore, a new `"summarizeTimeout"` event is now emitted through the `ContainerRuntime` via the `ISummarizerEvents` interface, providing a standardized mechanism for downstream consumers to observe and react to these events. This **telemetry improvement** and **feature addition**, primarily affecting `@fluidframework/container-runtime`, `@fluidframework/container-runtime-definitions`, `@fluidframework/runtime-utils`, and `@fluidframework/datastore`, will greatly aid in diagnosing and resolving performance issues within the **summarization subsystem**.

This commit introduces a **retry mechanism** to the **on-demand summarizer** flow within the `container-loader` package. It **fixes a race condition** where a summarizer container might unexpectedly close and fail during on-demand summary generation if it attempts to load a new snapshot while catching up. By retrying the summary operation, this change ensures the robustness of on-demand summaries, preventing consistent failures observed in e2e tests against ODSP and FRS. The `loadSummarizerContainerAndMakeSummary` function now incorporates this retry logic to handle transient failures gracefully.

This commit **enhances the documentation** for the `orderSequentially` function within the **`runtime-definitions` package** of the Fluid Framework. A JSDoc remark is added to `packages/runtime/runtime-definitions/src/dataStoreContext.ts` to explicitly state that `orderSequentially` may internally enter **staging mode**. This **maintenance update** clarifies a potential pitfall, explaining why developers might encounter an "Already in staging mode" error if they attempt to call `enterStagingMode()` within the function's callback, thereby improving developer understanding and preventing confusion.

This commit **temporarily skips a failing test seed** (seed 63) within the **`SharedDirectory` fuzz tests** located in `packages/dds/map/src/test/mocha/directoryFuzzTests.spec.ts`. This **maintenance** action prevents continuous integration failures caused by an identified, but currently unaddressed, bug. By adding seed 63 to the skip list, the test suite remains stable while a repair item tracks the eventual fix for this underlying issue.

This commit **re-enables** a previously skipped **end-to-end test** within `packages/test/test-end-to-end-tests/src/test/summarization/summaries.spec.ts` that verifies unsummarized operations. The test was consistently flaky in the pipeline due to timeouts, particularly under service slowdowns. To address this **flakiness**, the test's timeout has been increased from the default 2000ms to **5000ms**. This **maintenance** change improves the **test pipeline's reliability** by preventing false negatives caused by service latency, ensuring more accurate validation of the **summarization** subsystem.

This commit implements a **bug fix** within the **`fluid-runner`** tool, specifically in the `packages/tools/fluid-runner/src/exportFile.ts` module. It adds a call to `waitContainerToCatchUp` inside the `createContainerAndExecute` function to ensure the container processes all pending operations before execution. This change addresses a **flaky issue** observed in Loop e2e tests where trailing operations were sometimes not fully processed, even if not directly reproducible in all environments. The modification enhances the **robustness** and **reliability** of the `fluid-runner`'s export functionality, preventing potential data inconsistencies.

This commit **disables specific E2E tests** for older Fluid client versions when running against the **ODSP driver**, addressing persistent `summaryNack` errors caused by ODSP's lack of support for "dual-commit summaries." It introduces an `isOdspCompatCompliant` check within `packages/test/test-version-utils` to identify and **skip incompatible test configurations**, specifically for versions older than `"2.0.0"`. This **maintenance** change significantly **improves the reliability of the test suite** by preventing known, expected failures and reducing flakiness in **ODSP compatibility testing**.

This commit **disables** the "Conflicting ops" **end-to-end test** located in `packages/test/test-end-to-end-tests/src/test/sharedInterval.spec.ts`. This **maintenance** change addresses intermittent failures observed during OCE rotation against the local service, aiming to reduce CI noise. By adding `.skip` to the test definition, the project temporarily foregoes automated validation for this specific `SharedInterval` scenario until the underlying issue can be resolved, improving overall test suite stability.

This commit implements a **maintenance** change by **skipping** the `test explicitSchemaControl` within the `containerRuntime.spec.ts` test suite for the **FRS** (Fluid Relay Service) and **ODSP** (OneDrive/SharePoint) drivers. This action addresses persistent **test flakiness** observed in the CI pipeline, which was causing unreliable results. While the test is skipped for these specific drivers, the core client logic it covers remains adequately validated by other existing tests. A known fix for ODSP-related flakiness was not backported to LTS due to the issue's low impact in production environments.

This commit performs a **type reorganization** by **refactoring** the `SummarizerStopReason`, `ISummarizeEventProps`, and `ISummarizerEvents` types. These types have been **moved** from the `@fluidframework/container-runtime` package to the `@fluidframework/container-runtime-definitions` package. This change primarily affects the **summary subsystem** within the container runtime, ensuring these critical summarization-related types are now available on the `IContainerRuntime` interface. The public API surface of `container-runtime` is updated to reflect their removal, streamlining type definitions and improving architectural clarity.