Details
-
Bug
-
Status: Open (View Workflow)
-
Major
-
Resolution: Unresolved
-
10.5.8
-
None
-
None
-
Linux, Ubuntu 18.04, n.a.
Description
When we dropped the Primary Key of some partitioned tables the InnoDB Buffer Pool Statistics status became completely wrong:
1. Restart the database to resize the InnoDB Buffer Pool
2. Run the DDL:
SQL> alter table history drop primary key, add index (`itemid`,`clock`);
Query OK, 79946045 rows affected (15 min 9.689 sec)
Records: 79946045 Duplicates: 0 Warnings: 0
SQL> alter table history_str drop primary key, add index (`itemid`,`clock`);
Query OK, 119857625 rows affected (1 hour 34 min 14.117 sec)
Records: 119857625 Duplicates: 0 Warnings: 0
SQL> alter table history_uint drop primary key, add index (`itemid`,`clock`);
Query OK, 171818919 rows affected (36 min 14.557 sec)
Records: 171818919 Duplicates: 0 Warnings: 0
3. Get Status:
SQL> SHOW GLOBAL STATUS LIKE 'innodb_buffer_pool_pag%';
-------------------------------------------------------------+
| Variable_name | Value |
-------------------------------------------------------------+
| Innodb_buffer_pool_pages_data | 1924783 | -> data cannot be bigger than pages total! |
| Innodb_buffer_pool_pages_free | 1 | |
| Innodb_buffer_pool_pages_misc | 18446744073709175312 | --> negative number |
| Innodb_buffer_pool_pages_total | 1548480 | 24G / 16k should be: 1572864 ---------------------------------------- |
So all the number for a proper monitoring are completely wrong!
