2020 年 8 月August 2020

這些功能和 Azure Databricks 平臺改進已于2020年8月發行。These features and Azure Databricks platform improvements were released in August 2020.

注意

發行是暫存的。Releases are staged. 在初始發行日期之後,您的 Azure Databricks 帳戶可能不會更新到一周。Your Azure Databricks account may not be updated until up to a week after the initial release date.

重要

版本3.26 已發行給加拿大中部和印度中部區域中的客戶。Version 3.26 has been released to customers in the Canada Central and Central India regions only. 所有其他區域都將在3.27 發行時取得3.26 功能。All other regions will get the 3.26 features at the same time that 3.27 is released.

權杖管理 API 是 GA,而系統管理員可以使用管理主控台來授與及撤銷使用者對權杖的存取權Token Management API is GA and admins can use the Admin Console to grant and revoke user access to tokens

2020年8月26日-年9月1日:3.27 版August 26 - September 1, 2020: Version 3.27

權杖管理現已正式推出。Token management is now generally available. Azure Databricks 系統管理員可以使用權杖管理 API 和管理主控台來管理使用者的 Azure Databricks 個人存取權杖。Azure Databricks administrators can use the Token Management API and the Admin Console to manage their users’ Azure Databricks personal access tokens. 以系統管理員身分,您可以:As an admin, you can:

  • 監視及撤銷使用者的個人存取權杖。Monitor and revoke users’ personal access tokens.
  • 控制您工作區中未來權杖的存留期。Control the lifetime of future tokens in your workspace.
  • 控制哪些使用者可以透過許可權 API 或管理主控台來建立和使用權杖。Control which users can create and use tokens via the Permissions API or in the Admin Console.

在從公開預覽轉換為 GA 時,管理 API 參數的權杖已 created_by 變更為 created_by_id ,並新增了新的參數 created_by_usernameIn the transition from Public Preview to GA, the Token Management API parameter created_by was changed to created_by_id, and a new parameter, created_by_username was added.

如需詳細資訊,請參閱 管理個人存取權杖For more information, see Manage personal access tokens.

已增加 Shiny 應用程式的訊息大小限制Message size limits for Shiny apps increased

2020年8月26日-年9月1日:3.27 版August 26 - September 1, 2020: Version 3.27

發亮應用程式的應用程式大小上限已從 10 MB 增加到 20 MB。The maximum application size for Shiny apps has been increased from 10 MB to 20 MB. 如果您應用程式的總大小超過此限制,請參閱 亮亮的常見問題 以取得建議。If your application’s total size exceeds this limit, refer to the Shiny FAQ for recommendations.

以本地模式設定叢集的改良指示Improved instructions for setting up a cluster in local mode

2020年8月26日-年9月1日:3.27 版August 26 - September 1, 2020: Version 3.27

在叢集 UI 中:In the cluster UI:

  • 如果您建立具有0個背景工作的叢集,則會出現工具提示,建議您使用原生模式,並在) 顯示相關聯的設定設定 (spark.master local[*]If you create a cluster with 0 workers, a tool tip appears recommending that you use local mode and showing the associated configuration setting (spark.master local[*]).
  • 您無法再設定叢集 spark.master local[*] ,除非叢集有0個背景工作角色。You can no longer set spark.master local[*] for a cluster, unless the cluster has 0 workers.

與執行相關聯筆記本的檢視版本View version of notebook associated with a run

2020年8月26日-年9月1日:3.27 版August 26 - September 1, 2020: Version 3.27

您現在可以從 [實驗] 提要欄位,顯示與執行相關聯的筆記本版本。From the Experiments sidebar, you can now display the version of a notebook associated with a run. 如需詳細資訊,請參閱 觀看筆記本實驗For details, see View notebook experiment.

Databricks Runtime 7.2 GADatabricks Runtime 7.2 GA

2020年8月20日August 20, 2020

Databricks Runtime 7.2 提供許多額外的功能和 Databricks Runtime 7.1 的增強功能,包括:Databricks Runtime 7.2 brings many additional features and improvements over Databricks Runtime 7.1, including:

  • 自動載入 器現已正式推出:自動載入器是一種有效率的方法,可將大量檔案以累加方式擷取至 Delta Lake。Auto Loader is generally available: Auto Loader is an efficient method for incrementally ingesting a large number of files into Delta Lake. 現在已正式推出,並新增了下列功能:It is now GA and adds the following features:
    • 目錄清單模式選項:自動載入器除了現有的檔案通知模式之外,還會新增一個新的目錄清單模式,以判斷何時有新的檔案。Directory listing mode option: Auto Loader adds a new directory listing mode, in addition to the existing file notification mode, for determining when there are new files.
    • 雲端資源管理 API:您現在可以使用我們的 Scala API 來管理自動載入器所建立的雲端資源。Cloud resource management API: You can now use our Scala API to manage cloud resources created by Auto Loader. 您可以使用此 API 列出 notification services,並卸載特定的 notification services。You can list notification services and tear down specific notification services using this API.
    • 速率限制選項:您現在可以使用 cloudFiles.maxBytesPerTrigger 選項來限制每個 microbatch 中所處理的資料量。Rate limiting option: You can now use the cloudFiles.maxBytesPerTrigger option to limit the amount of data processed in each microbatch.
    • 選項驗證:自動載入器現在會驗證您提供的選項。validationOption validation: Auto Loader now validates the options you provide.validation 將會失敗。will fail. 若要略過選項驗證,請將設定 cloudFiles.validateOptionsfalseTo skip option validation, set cloudFiles.validateOptions to false.
  • 使用 clone 有效率地複製 Delta 資料表Efficiently copy a Delta table with clone.
  • 增強功能:Improvements:
    • 雪花連接器已升級為版本2.8.1,其中包含 Spark 3.0 支援。Snowflake connector has been upgraded to version 2.8.1, which includes Spark 3.0 support.
    • 認證傳遞改善Credential passthrough improvements
    • TensorBoard 改進TensorBoard improvements
    • 升級的 Python 和 R 程式庫Upgraded Python and R libraries

如需詳細資訊,請參閱完整的 Databricks Runtime 7.2 (不支援的) 版本資訊。For details, see the complete Databricks Runtime 7.2 (Unsupported) release notes.

Databricks Runtime 7.2 ML GADatabricks Runtime 7.2 ML GA

2020年8月20日August 20, 2020

適用于 Machine Learning 的 Databricks Runtime 7.2 建置於 Databricks Runtime 7.2 之上,並引進全新和改良的 Python 和系統程式庫。Databricks Runtime 7.2 for Machine Learning is built on top of Databricks Runtime 7.2 and brings new and improved Python and system libraries. 如需詳細資訊,請參閱 Machine Learning (不支援的完整 Databricks Runtime 7.2) 版本資訊。For details, see the complete Databricks Runtime 7.2 for Machine Learning (Unsupported) release notes.

Databricks Runtime 7.2 Genomics GADatabricks Runtime 7.2 Genomics GA

2020年8月20日August 20, 2020

適用于 Genomics 的 Databricks Runtime 7.2 建立在 Databricks Runtime 7.2 之上,並大幅加快將常值 numpy 1D 和2D 浮點數類型 ndarrays 轉換成 JAVA 陣列的速度。Databricks Runtime 7.2 for Genomics is built on top of Databricks Runtime 7.2 and significantly speeds up the conversion of literal numpy 1D and 2D float-typed ndarrays to Java arrays. 「發光全 基因組關聯」研究檔 會反映其使用方式。The Glow genome-wide association study documentation reflects the usage.

如需詳細資訊,請參閱 Genomics 的完整 Databricks Runtime 7.2 (不支援的) 版本資訊。For details, see the complete Databricks Runtime 7.2 for Genomics (Unsupported) release notes.

存取權限 API (公開預覽)Permissions API (Public Preview)

2020 年 8 月 18 日August 18, 2020

Databricks 很高興宣佈許可權 API 的公開預覽,可讓您管理的許可權:Databricks is pleased to announce the public preview of the Permissions API, which lets you manage permissions for:

  • 權杖Tokens
  • 叢集Clusters
  • 集區Pools
  • 工作Jobs
  • NotebooksNotebooks
  • 資料夾 (目錄) Folders (directories)
  • MLflow 已註冊的模型MLflow registered models

如需詳細資訊,請參閱 許可權 APIFor more information, see Permissions API.

Databricks Connect 7.1 (GA)Databricks Connect 7.1 (GA)

2020 年8 月 12 日August 12, 2020

Databricks Connect 現在支援 Databricks Runtime 7.1。Databricks Connect now supports Databricks Runtime 7.1.

在 Databricks Runtime 7.1 中,Databricks 建議您一律使用最新版本的 Databricks Connect。In Databricks Runtime 7.1, Databricks recommends that you always use the most recent version of Databricks Connect.

叢集程式庫的可重複安裝順序Repeatable installation order for cluster libraries

12-25 年8月,2020:版本3.26August 12-25, 2020: Version 3.26

在執行 Databricks Runtime 7.2 或更新版本的叢集上,Azure Databricks 現在會依照安裝的連續處理所有叢集程式庫。On a cluster running Databricks Runtime 7.2 or above, Azure Databricks now processes all cluster libraries in the order that they were installed.

從 MLflow 註冊的模型頁面建立模型 (公開預覽)Create model from MLflow registered models page (Public Preview)

12-25 年8月,2020:版本3.26August 12-25, 2020: Version 3.26

您現在可以從 [MLflow 註冊的模型] 頁面建立新的模型。You can now create a new model from the MLflow registered models page. 如需詳細資訊,請參閱 建立新的已註冊模型,並將記錄的模型指派給它For details, see Create a new registered model and assign a logged model to it.

Databricks 容器服務支援 GPU 映像Databricks Container Services supports GPU images

12-25 年8月,2020:版本3.26August 12-25, 2020: Version 3.26

您現在可以在具有 Gpu 的叢集上使用 Databricks 容器服務 ,以建立具有自訂程式庫的便攜深度學習環境。You can now use Databricks Container Services on clusters with GPUs to create portable deep learning environments with customized libraries.

如需詳細資訊,請參閱 DATABRICKS GPU 叢集上的容器服務For details, see Databricks Container Services on GPU clusters.