MySQL分支 Percona 5.6.15 發布

jopen 10年前發布 | 13K 次閱讀 Percona

Percona 為 MySQL 數據庫服務器進行了改進，在功能和性能上較 MySQL 有著很顯著的提升。該版本提升了在高負載情況下的 InnoDB 的性能、為 DBA 提供一些非常有用的性能診斷工具；另外有更多的參數和命令來控制服務器行為。

Percona Server 只包含 MySQL 的服務器版，并沒有提供相應對 MySQL 的 Connector 和 GUI 工具進行改進。

Percona Server 使用了一些 google-mysql-tools, Proven Scaling, Open Query 對 MySQL 進行改造。

與內置的 MySQL 存儲引擎相比，XtraDB 提供了一些極大的改進，但它不是一款獨立產品，也無法輕松放入現有 MySQL 安裝。因此，如果您想使用這款新引擎，則必須使用提供它的產品。

Percona Server 就是這樣一款產品，由領先的 MySQL 咨詢公司 Percona 發布。Percona Server 是一款獨立的數據庫產品，為用戶提供了換出其 MySQL 安裝并換入 Percona Server 產品的能力。通過這樣做，就可以利用 XtraDB 存儲引擎。Percona Server 聲稱可以完全與 MySQL 兼容，因此從理論上講，您無需更改軟件中的任何代碼。這確實是一個很大的優勢，適合在您尋找快速性能改進時控制質量。因此，采用 Percona Server 的一個很好的理由是，利用 XtraDB 引擎來盡可能地減少代碼更改。

此外，他們是 XtraDB 存儲引擎的原作者。Percona 將此代碼用作開源代碼，因此您可以在其他產品中找到它，但引擎的最初創建者與編寫此產品的是同一個人，所以您可以隨心所欲地使用此信息。

下面是 Percona Server 的聲明，該聲明來自它們自己的網站：

可擴展性：處理更多事務；在強大的服務器上進行擴展
性能：使用了 XtraDB 的 Percona Server 速度非常快
可靠性：避免損壞，提供崩潰安全 (crash-safe) 復制
管理：在線備份，在線表格導入/導出
診斷：高級分析和檢測
靈活性：可變的頁面大小，改進的緩沖池管理

Percona 團隊的最終聲明是 “Percona Server 是由 Oracle 發布的最接近官方 MySQL Enterprise 發行版的版本”，因此與其他更改了大量基本核心 MySQL 代碼的分支有所區別。Percona Server 的一個缺點是他們自己管理代碼，不接受外部開發人員的貢獻，以這種方式確保他們對產品中所包含功能的控制。

同步基于 MySQL 5.6.15/5.5.34/5.1.73.此版本增強了線程池在高并發下的性能，增強mysqlbinlog支持SSL和壓縮支持，降低了信息模式里用戶統計的性能開銷及一些 Bug 修正。

5.6.15完全改進：

New Features

Thread Pool high priority scheduling is now enabled by default by changing the default thread_pool_high_prio_tickets value from0to4294967295.

Percona Server now supports Low priority queue throttling. This feature should improve Thread Pool performance under high concurrency in a situation when thread groups are oversubscribed.

Introduced new thread_pool_high_prio_mode to provide more fine-grained control over high priority scheduling either globally or per connection in Thread Pool.

Percona Server has extended mysqlbinlog to provideSSLand compression support.

Percona Server has reduced the performance overhead of the User Statistics feature.

Bugs Fixed

INSTALL PLUGINstatement would crash server if User Statistics were enabled. Bug fixed #1011047.

innodb_log_checksum_algorithm did not have any effect when set frommy.cnfor mysqld command line, it would have effect only if set dynamically. Bug fixed #1248505.

Server would crash on shutdown if Atomic write support for Fusion-io devices feature is enabled. Bug fixed #1255628 (Jan Lindstr?m).

Percona Server would crash when data was select from XTRADB_RSEG table when InnoDB system table space was initialized with lower then default number of rollback segments. Bug fixed #1260259.

Certain types of workloads (large result sets, blobs, slow clients) can have longer waits on network I/O (socket reads and writes). Whenever server waits, this should be communicated to the Thread Pool, so it can start new query by either waking a waiting thread or sometimes creating a new one. Ported MariaDB patch MDEV-156, bug fixed #1159743.

mysqldump --innodb-optimize-keyswas generating incorrectCREATE TABLEstatements for partitioned tables. Bug fixed #1233841.

Fixed errors when server was compiled with-DWITH_LIBWRAP=ONoption. Bug fixed #1240442.

Ifinnobase_atomic_writeswas used on separate undo files that do not exists would lead to operating system error. Bug fixed #1255638 (Jan Lindstr?m).

Default value for thread_pool_max_threads has been changed from500to100 000(the maximum supported number of connections), because limiting the total number of threads in the Thread Pool can result in deadlocks and uneven distribution of worker threads between thread groups in case of stalled connections. Bug fixed #1258097.

PURGE CHANGED_PAGE_BITMAPS BEFOREstatement would delete the changed page data after the specified LSN and up to the start of the next bitmap file. If this data were to be used for fast incremental backups, its absence would cause Percona XtraBackup to fall back to the full-scan incremental backup mode. Bug fixed #1260035 (Andrew Gaul).

Server performance could degrade under heavy load or it could deadlock on shutdown while performing purge. Bug fixed #1236696.

Server could crash under heavy load if InnoDB compression was used. Bug fixed #1240371.

Redo log checksum mismatches would be diagnosed using the data page checksum algorithm setting instead of redo log checksum algorithm one. Bug fixed #1250148.