Details
-
Bug
-
Status: Closed (View Workflow)
-
Minor
-
Resolution: Cannot Reproduce
-
10.6, 10.7, 10.8, 10.9, 10.10
Description
The following test failed for me locally:
|
10.7 97d16c7544c40430facda441dab89fa4b23468a0 |
atomic.alter_table 'innodb' w41 [ fail ] Found warnings/errors in server log file!
|
Test ended at 2022-08-02 08:40:07
|
line
|
2022-08-02 8:39:10 0 [ERROR] InnoDB: Operating system error number 2 in a file operation.
|
2022-08-02 8:39:10 0 [ERROR] InnoDB: The error means the system cannot find the path specified.
|
2022-08-02 8:39:10 0 [ERROR] InnoDB: File ./test/#sql-backup-55d70-4.ibd: 'delete' returned OS error 71.
|
^ Found warnings in /dev/shm/10.7/mysql-test/var/41/log/mysqld.1.err
|
In the var/41/log/mysqld.1.err, the warnings occurred in the following snippet:
|
10.7 97d16c7544c40430facda441dab89fa4b23468a0 |
Version: '10.7.5-MariaDB-debug-log' socket: '/dev/shm/10.7/mysql-test/var/tmp/41/mysqld.1.sock' port: 16760 Source distribution
|
2022-08-02 8:39:10 0 [ERROR] InnoDB: Operating system error number 2 in a file operation.
|
2022-08-02 8:39:10 0 [ERROR] InnoDB: The error means the system cannot find the path specified.
|
2022-08-02 8:39:10 0 [ERROR] InnoDB: File ./test/#sql-backup-55d70-4.ibd: 'delete' returned OS error 71.
|
2022-08-02 8:39:11 4 [Note] mariadbd: Crashing at ddl_log_alter_after_delete_backup
|
SIGKILL myself
|
The "OS error 71" is actually InnoDB’s own invented error code OS_FILE_NOT_FOUND.
Based on this output, I created a reduced version of the test, preserving only this and the previous crash point, and covering only the InnoDB recovery:
diff --git a/mysql-test/suite/atomic/alter_table.test b/mysql-test/suite/atomic/alter_table.test
|
index 1f87b7e71b8..2b415a68a16 100644
|
--- a/mysql-test/suite/atomic/alter_table.test
|
+++ b/mysql-test/suite/atomic/alter_table.test
|
@@ -32,16 +32,16 @@ RESET MASTER;
|
|
if ($engine_count == "")
|
{
|
- let $engine_count=2;
|
- let $engines='myisam','innodb';
|
+ let $engine_count=1;
|
+ let $engines='innodb';
|
}
|
if ($extra_engine == "")
|
{
|
let $extra_engine=aria;
|
}
|
|
-let $crash_count=13;
|
-let $crash_points='ddl_log_alter_after_create_frm', 'ddl_log_alter_after_create_table', 'ddl_log_alter_after_prepare_inplace','ddl_log_alter_after_copy', 'ddl_log_alter_after_log', 'ddl_log_alter_after_rename_to_backup', 'ddl_log_alter_after_rename_to_backup_log', 'ddl_log_alter_rename_frm', 'ddl_log_alter_after_rename_to_original', 'ddl_log_alter_before_rename_triggers', 'ddl_log_alter_after_rename_triggers', 'ddl_log_alter_after_delete_backup', 'ddl_log_alter_after_drop_original_table';
|
+let $crash_count=2;
|
+let $crash_points='ddl_log_alter_after_rename_triggers', 'ddl_log_alter_after_delete_backup';
|
|
let $statement_count=16;
|
let $statements='ALTER TABLE t1 ADD COLUMN c INT, COMMENT "new"', |
Unfortunately, I was unable to reproduce the failure by running either the full test or the reduced test several times. The reduced test passed 33×100 runs on both 10.6 and 10.7.
I would like to see rr record traces both of the killed server and of the recovery where such messages are being emitted.
The messages seem to be a nuisance only; I do not see any correctness problem.
Attachments
Issue Links
- relates to
-
MDEV-25180 Atomic ALTER TABLE
-
- Closed
-