Details
-
Bug
-
Status: Open (View Workflow)
-
Major
-
Resolution: Unresolved
-
10.6, 10.11, 11.4, 11.8, 12.1, 12.2
-
Can result in unexpected behaviour
Description
The test innodb_fts.crash_recovery,release may be the only test that kills the server at an arbitrary point of execution during a DDL workload.
This test occasionally fails like the following:
innodb_fts.crash_recovery 'orig,release' w2 [ fail ] Found warnings/errors in server log file!
|
Test ended at 2025-10-02 15:40:25
|
line
|
2025-10-02 15:40:20 0 [ERROR] DDL_LOG: Got error 1033 when trying to execute action for entry 2 of type 'alter table'
|
^ Found warnings in /dev/shm/var/2/log/mysqld.1.err
|
In the corresponding server error log (mtr-vardir/2/log/mysqld.1.err in the archive), we can find the following:
2025-10-02 15:40:20 0 [Note] InnoDB: Starting crash recovery from checkpoint LSN=47815
|
2025-10-02 15:40:20 0 [Note] InnoDB: End of log at LSN=180257
|
2025-10-02 15:40:20 0 [Note] InnoDB: To recover: 207 pages
|
2025-10-02 15:40:20 0 [Note] InnoDB: 1 transaction(s) which must be rolled back or cleaned up in total 1 row operations to undo
|
2025-10-02 15:40:20 0 [Note] InnoDB: Trx id counter is 62
|
2025-10-02 15:40:20 0 [Note] InnoDB: Opened 3 undo tablespaces
|
2025-10-02 15:40:20 0 [Note] InnoDB: 128 rollback segments in 3 undo tablespaces are active.
|
2025-10-02 15:40:20 0 [Note] InnoDB: Removed temporary tablespace data file: "./ibtmp1"
|
2025-10-02 15:40:20 0 [Note] InnoDB: Setting file './ibtmp1' size to 12.000MiB. Physically writing the file full; Please wait ...
|
2025-10-02 15:40:20 0 [Note] InnoDB: File './ibtmp1' size is now 12.000MiB.
|
2025-10-02 15:40:20 0 [Note] InnoDB: log sequence number 180257 (memory-mapped); transaction id 63
|
...
|
2025-10-02 15:40:20 0 [Note] Recovering after a crash using tc.log
|
2025-10-02 15:40:20 0 [Note] Starting table crash recovery...
|
2025-10-02 15:40:20 0 [Note] Crash table recovery finished.
|
2025-10-02 15:40:20 0 [Note] InnoDB: Starting in background the rollback of recovered transactions
|
2025-10-02 15:40:20 0 [Note] InnoDB: Rolled back recovered transaction 31
|
2025-10-02 15:40:20 0 [Note] InnoDB: Rollback of non-prepared transactions completed
|
2025-10-02 15:40:20 0 [ERROR] mariadbd: Incorrect information in file: './test/#sql-alter-17f5b-7.frm'
|
2025-10-02 15:40:20 0 [ERROR] DDL_LOG: Got error 1033 when trying to execute action for entry 2 of type 'alter table'
|
2025-10-02 15:40:20 0 [Note] DDL_LOG: Crash recovery executed 2 entries
|
2025-10-02 15:40:20 0 [Note] Server socket created on IP: '127.0.0.1', port: '19060'.
|
2025-10-02 15:40:20 0 [Note] mariadbd: Event Scheduler: Loaded 0 events
|
2025-10-02 15:40:20 0 [Note] /tmp/workspace/Test-Package/bin/mariadbd: ready for connections.
|
My initial guess based on this output is that the .frm file had been incompletely written by the server process that had been killed. Unfortunately, no copy of that file is available in the archive.
Note: This is not the only reason why the test innodb_fts.crash_recovery,release has been failing, but this is by far the most prominent ones. Out of the 17 latest failures that I checked in the Buildbot cross-reference all but some 3 were like this. The exceptions were "Server failed to restart" due to an apparent Linux kernel bug (the same as in MDEV-37537).
The last-but-18 failure https://buildbot.mariadb.org/#/builders/482/builds/28270 is an InnoDB recovery error, which I will file separately.
Attachments
Issue Links
- relates to
-
MDEV-25180 Atomic ALTER TABLE
-
- Closed
-