| Bug #38769 | Backup: write table data does not stop iterating when all drivers are handled | ||
|---|---|---|---|
| Submitted: | 13 Aug 2008 13:00 | Modified: | 29 Sep 2008 23:14 |
| Reporter: | Jørgen Løland | Email Updates: | |
| Status: | Closed | Impact on me: | |
| Category: | MySQL Server: Backup | Severity: | S5 (Performance) |
| Version: | 6.0.7-backup | OS: | Any |
| Assigned to: | Jørgen Løland | CPU Architecture: | Any |
[13 Aug 2008 13:03]
Bugs System
A patch for this bug has been committed. After review, it may be pushed to the relevant source trees for release in the next version. You can access the patch from: http://lists.mysql.com/commits/51526 2683 Jorgen Loland 2008-08-13 Bug#38769 - Backup: write table data does not stop iterating when all drivers are handled Fix: stop iteration that adds drivers to the scheduler when there are no more drivers in Image_info.m_snap
[13 Aug 2008 13:06]
Jørgen Løland
The attached patch fixes the problem (checked with single-stepping in gdb). All backup tests passed with the fix.
[1 Sep 2008 15:24]
Rafal Somla
Good to push.
[2 Sep 2008 7:39]
Bugs System
A patch for this bug has been committed. After review, it may be pushed to the relevant source trees for release in the next version. You can access the patch from: http://lists.mysql.com/commits/53036 2688 Jorgen Loland 2008-09-02 Bug#38769 - Backup: write table data does not stop iterating when all drivers are handled Fix: stop iteration that adds drivers to the scheduler when there are no more drivers in Image_info.m_snap
[2 Sep 2008 7:40]
Bugs System
A patch for this bug has been committed. After review, it may be pushed to the relevant source trees for release in the next version. You can access the patch from: http://lists.mysql.com/commits/53037 2688 Jorgen Loland 2008-09-02 Bug#38769 - Backup: write table data does not stop iterating when all drivers are handled Fix: stop iteration that adds drivers to the scheduler when there are no more drivers in Image_info.m_snap
[28 Sep 2008 19:23]
Øystein Grøvlen
Pushed to 6.0 main for 6.0.8
[29 Sep 2008 23:14]
Paul DuBois
No user impact. No changelog entry needed.
Description: In data_backup.cc#write_table_data, the for-loop that adds drivers to the scheduler or the inactive list iterates 256 times regardless of the number of drivers: 444 // add unknown "at end" drivers to scheduler, rest to inactive list 445 446 for (uint n=0; n < 256; ++n) 447 { 448 Snapshot_info *i= info.m_snap[n]; 449 450 if (!i) 451 continue; Very little processing is done in the extra iterations; because of lines 450-451, each loop is short-circuited. Hence, this is a pure performance issue. How to repeat: Code investigation Suggested fix: Replace "for (uint n=0; n < 256; ++n)" with "for (uint n=0; n < info.snap_count(); ++n) Nothing in Image_info::add_snapshot indicates that there can be holes in m_snap so that there can be drivers at m_snap[n] where n >= snap_count