Percona XtraBackup moved to https://jira.percona.com/projects/PXB

Xtrabackup in MySQL 5.6

Asked by David Jiang

In MySQL 5.6, transaction commit does not do fsync in InnoDB engine because it can recover by MySQL binlog.
However, xtrabackup requires this "commit fsync" to get the binlog posistion, so it may cause inconsistencent data between InnoDB and binlog position.
Does Xtrabackup address this issue? And how ?

Question information

Language:
English Edit question
Status:
Answered
For:
Percona XtraBackup moved to https://jira.percona.com/projects/PXB Edit question
Assignee:
No assignee Edit question
Last query:
Last reply:
Revision history for this message
David Jiang (jiangchengyao) said : 		#1

I find MariaDB 10.0 has the same problem.
And it has been recorded in https://mariadb.com/kb/en/mariadb/documentation/optimization-and-tuning/binlog-group-commit-and-innodb_flush_log_at_trx_commit/

Revision history for this message
Alexey Kopytov (akopytov) said : 		#2

David,

Group commit does introduce a problem for XtraBackup, but it's not about fsync. It is about InnoDB log buffer not necessarily being flushed on commit, which means that the binary log position and the copy of redo log created by XtraBackup can be inconsistent. To prevent that, XtraBackup issues FLUSH ENGINE LOGS to flush redo log explicitly before finalizing its redo log copy (see bug #1320685).

Can you help with this problem?

Provide an answer of your own, or ask David Jiang for more information if necessary.

To post a message you must log in.