Redis Enterprise Software 发行说明 6.0（2020 年 5 月）

数据库访问的 ACL 和 RBAC 改进。主动-主动数据库支持 Redis Streams。

Redis 企业软件

Redis Enterprise Software （RS） 6.0 现已推出！ 此新版本捆绑了开源 Redis 6，使用基于角色的访问控制 （RBAC） 实现增强的访问控制列表 （ACL） 功能以进行数据库访问，并增加了对主动-主动数据库的 Redis Streams 的支持。

版本信息

升级说明

按照这些说明从 RS 5.4.0 及更高版本升级到 RS 6.0。 对于主动-主动部署，此版本要求您升级 CRDB 功能集版本。

生命周期结束

Redis Enterprise Software 6.0 和之前的 RS 版本的生命周期终止 （EOL） 可在此处找到。 Redis 模块的 EOL 可以在这里找到。

  • 对 Red Hat Enterprise Linux 6 和 Oracle Linux 6 作系统平台的支持将于 2020 年 11 月 30 日结束。
  • 对 Ubuntu 14.04 （Trusty Tahr） 作系统平台的支持将于 2020 年 11 月 30 日结束。

新功能

开源 Redis 6

RS 6.0 捆绑了最新的开源 Redis 6。 有关更多信息，请查看 Dive into Redis 6 一文。

访问控制列表 （ACL）

RS 6.0 基于 OSS Redis 6，能够使用用户及其数据访问权限（包括他们可以执行的命令和可以访问的密钥）来管理和控制与数据库的连接。

在 OSS Redis 中，ACL 是按用户针对每个数据库单独管理的。在 Redis Enterprise Software 中，Redis ACL 针对集群中的数据库进行管理。 有关更多信息，请查看 Redis Enterprise Software 用户管理文档。

基于角色的访问控制 （RBAC）

RS 6.0 利用 Redis ACL 实施基于角色的访问控制，从而轻松扩展和管理数据访问权限。使用角色可以最大限度地减少管理具有许多数据库、多个用户和各种访问控制列表的集群所涉及的开销。 有关更多信息，请查看 Redis Enterprise Software 用户管理文档。

对 Redis Streams 的主动-主动支持

RS 6.0 增加了对使用无冲突复制数据类型 （CRDT） 的主动-主动异地分布式数据库上的 Redis 流的支持。您现在可以使用所有 Redis 流命令，包括主动-主动数据库上的使用者组。要启用它，请在升级过程中将主动-主动数据库功能集版本升级到最新版本（功能集版本 = 2）。 有关更多信息，请查看活动数据库上的 Redis Streams。

Redis 模块

以下 Redis 模块的 GA 版本捆绑在 RS 6.0 中：

要将更新的模块与数据库一起使用，必须升级数据库上的模块。

其他功能

  • 添加了配置使用 S3 协议的存储服务的功能。可以配置备份位置以及 RDB 文件的导入和导出位置。存储服务必须具有有效的 SSL 证书。要连接到与 S3 兼容的存储位置，请运行：rladmin cluster config s3_url <url>

  • crdb-cli 工具已更新，因此它现在显示主动-主动数据库的功能集版本和每个实例的协议版本。例如：

    $ crdb-cli crdb list --verbose
CRDB-GUID               NAME                REPL-ID	PROTOCOL    FEATURESET    DB-ID     CLUSTER-FQDN
969122be-...	loremipsum1           1	      0 	   	 0	    bdb:1       cluster1.local
969122be-...	loremipsum1           2	      0 	   	 0	    bdb:1       cluster2.local

  • Added REST API and rladmin commands to modify the timeout for automatically disconnecting an inactive admin console session.

    • Using the rladmin run: rladmin cluster config cm_session_timeout_minutes <int_value>

    • Using the REST API:

      curl --request PUT \
--url https://localhost:9443/v1/cluster \
--header 'content-type: application/json' \
--data '{
    "cm_session_timeout_minutes": 10
}'

  • Added no_of_expires metrics for database metrics and for shard metrics. You can access this metric from the REST API:

    • no_of_expires shows the current number of volatile keys in the database.
    • expired_objects shows the rate of keys expired in DB (expirations/sec).

  • Added the ability to customize the welcome message on the login page in the admin console console.

Important fixes

  • RS39121, RS35335 - Optimized the XREAD and XREADGROUP commands so when using them in a non blocking fashion (without BLOCK keyword) they use the shared connection instead of a dedicated connection.
  • RS26448 - Fixed ‘ram overhead’ metric calculation for ROF databases using AOF.
  • RS31190 - Fixed a bug that causes databases with OSS Cluster API enabled to override the ‘preferred IP type’ attribute from ‘external’ to ‘internal’ (the default value).
  • RS34009 - Updated the modules loading procedure for clusters with FIPS compliance enabled.
  • RS38233 - Improved the Redis cleanup job handling the persistent directory
  • RS39228 - Fixed a bug in the WAIT command that in some cases was released after a longer period than requested.
  • RS39749 - Fixed a bug that blocked eviction while LUA scripts were in progress.
  • RS43996 - Fixed a bug when aborting an upgrade to RS 6.0.

Known limitations

-RS81463 - A shard may crash when resharding an Active-Active database with Auto Tiering . Specifically, the shard will crash when volatile keys or Active-Active tombstone keys reside in Flash memory.

Active-Active databases

  • RS51359 - Active-Active databases, using replication and Append Only File (AOF) for database persistence, are suffering from memory leaks on replica shards, causing them to grow bigger than the master shards. Customers are advised to upgrade to RS 6.0.12 TBD. Meanwhile you can use snapshots for database persistence or restart the replica shards TBD.

Installation limitations

Several Redis Enterprise Software installation reference files are installed to the directory /etc/opt/redislabs/ even if you use custom installation directories.

As a workaround to install Redis Enterprise Software without using any root directories, do the following before installing Redis Enterprise Software:

  1. Create all custom, non-root directories you want to use with Redis Enterprise Software.

  2. Mount /etc/opt/redislabs to one of the custom, non-root directories.

Upgrade

  • RS 5.4.2 introduced new Active-Active Redis Database capabilities that improve its compatibility with open source Redis. Now the string data-type in Active-Active Redis Database is implicitly and dynamically typed, just like open source Redis. To use the new capabilities on nodes that are upgraded from version RS 5.4.2 or lower, you must upgrade the Active-Active Redis Database protocol.
  • When you upgrade an Active-Active Redis with active AOF from version RS 5.4.2 or lower to version RS 5.4.4 or higher:
    • If replication is enabled, you must run the BGREWRITEAOF command on all replica shards after the upgrade.
    • If replication is not enabled, you must run the BGREWRITEAOF command on all shards after the upgrade.
  • Starting from RS 5.4.2, to preserve the current Redis major.minor version during database upgrade you must use the keep_redis_version option instead of keep_current_version.
  • Dynatrace agent installed on the cluster nodes can hamper the working on Envoy process leading to failure of UI and REST API. Prior upgrading we recommend removing Dynatrace completely or try upgrading to newer versions.

Redis commands

  • The capability of disabling specific Redis commands does not work on commands specific to Redis Modules.
  • Starting from RS 5.4.2 and after you upgrade an Active-Active database, TYPE commands for string data-type in Active-Active databases return "string" (OSS Redis standard).
RATE THIS PAGE
Back to top ↑