[MDEV-40791] SET GLOBAL innodb_log_archive=OFF triggers full flush - Jira

XML

Word

Printable

Details

Type: Bug
Status: Closed (View Workflow)
Priority: Critical
Resolution: Fixed
Affects Version/s: 13.0
Fix Version/s: 13.0.2
Component/s: Storage Engine - InnoDB
Labels:
- performance

Bug Category:
Related to performance
Release Note Summary:
An unnecessary write of the entire buffer pool could be triggered by SET GLOBAL innodb_log_archive=OFF.

Description

During the testing of MDEV-14992 it turned out that disabling the log archiving by executing

SET GLOBAL innodb_log_archive=OFF;

would trigger a full buffer pool flush:


      wait_lsn= 0;
    retry_after_checkpoint:
      IF_WIN(log_resize_release_locks(),);
      if (wait_lsn)
      {
        mysql_mutex_lock(&buf_pool.flush_list_mutex);
        buf_flush_wait(wait_lsn, false);
        mysql_mutex_unlock(&buf_pool.flush_list_mutex);
      }
      latch.wr_unlock();
      continue;

In my test scenario, the log file format is being changed in place.

If we made innodb_log_archive=ON write the sequence bit exactly in the same way as innodb_log_archive=OFF, then the wait for an extra checkpoint should be avoidable.

Attachments

Issue Links

blocks

MDEV-14992 BACKUP SERVER to mounted file system

In Progress

relates to

MDEV-37949 Implement innodb_log_archive

Closed

MDEV-40683 SET GLOBAL innodb_log_archive=OFF may break recovery and backup

Closed

Activity

People

Assignee:: Marko Mäkelä

Reporter:: Marko Mäkelä

Assigned for Implementation:: Marko Mäkelä

Assigned for Review:: Thirunarayanan Balathandayuthapani

Votes:: 0 Vote for this issue

Watchers:: 1 Start watching this issue

Dates

Created:: 2026-08-17 12:07

Updated:: 2026-08-20 14:45

Resolved:: 2026-08-20 14:45

Time Tracking

Estimated:

Not Specified

Remaining:

Logged:

7.25h

Git Integration

Error rendering 'com.xiplink.jira.git.jira_git_plugin:git-issue-webpanel'. Please contact your Jira administrators.

a848493c6fe031f23606144420c1ca1e467cbd81:storage/innobase/log/log0log.cc
796	wait_lsn= 0;
797	retry_after_checkpoint:
798	IF_WIN(log_resize_release_locks(),);
799	if (wait_lsn)
800	{
801	mysql_mutex_lock(&buf_pool.flush_list_mutex);
802	buf_flush_wait(wait_lsn, false);
803	mysql_mutex_unlock(&buf_pool.flush_list_mutex);
804	}
805	latch.wr_unlock();
806	continue;