Details
-
Bug
-
Status: In Progress (View Workflow)
-
Major
-
Resolution: Unresolved
-
10.5, 10.6, 10.2(EOL), 10.3(EOL), 10.4(EOL)
Description
In stack traces posted to MDEV-25773, we can observe that DROP DATABASE (or DROP TABLE) would end up waiting for the buffer pool mutex, for allocating a page so that the to-be-dropped table could be opened.
This is unnecessary and could cause a crash, in case the data file is corrupted. In MDEV-21069 we already removed writes to the .ibd file on DROP TABLE. There should be no need to read the .ibd file either.
Attachments
Issue Links
- relates to
-
MDEV-25801 DROP DATABASE very slow after innodb undo log truncate
- Closed
-
MDEV-21069 Crash on DROP TABLE if the data file is corrupted
- Closed
-
MDEV-25773 sysbench-TPCC failed to prepare with extremely slow execution
- Closed