Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Fixed
-
N/A
Description
The failure comes in two flavors.
One looks like this:
|
bb-10.6-monty-innodb f179fdefd1 |
2021-05-11 19:14:17 0 [ERROR] InnoDB: Expected tablespace id 67 but found 0 in the file ./test/tt13.ibd
|
2021-05-11 19:14:17 0 [Note] DDL_LOG: Crash recovery executed 2 entries
|
That is, the error message is thrown once, and then the recovery finishes. I'm not sure whether it's a problem or a bogus message here.
Another one looks like this:
2021-05-12 3:39:01 0 [Note] InnoDB: Creating #2 encryption thread id 140169671862016 total threads 4.
|
2021-05-12 3:39:01 0 [Note] InnoDB: Creating #3 encryption thread id 140169663469312 total threads 4.
|
2021-05-12 3:39:01 0 [Note] InnoDB: Creating #4 encryption thread id 140169655076608 total threads 4.
|
2021-05-12 3:39:01 0 [Note] InnoDB: Buffer pool(s) load completed at 210512 3:39:01
|
2021-05-12 3:39:01 0 [ERROR] InnoDB: Expected tablespace id 10 but found 0 in the file ./test/D.ibd
|
2021-05-12 3:39:01 0 [ERROR] InnoDB: Expected tablespace id 10 but found 0 in the file ./test/D.ibd
|
2021-05-12 3:39:01 0 [ERROR] InnoDB: Expected tablespace id 10 but found 0 in the file ./test/D.ibd
|
...
|
2021-05-12 3:39:01 0 [ERROR] InnoDB: Expected tablespace id 10 but found 0 in the file ./test/D.ibd
|
Thousands of identical messages are produced within a second. I don't know if it's an endless loop or it would finish eventually, as the test was killing the server after encountering those.
In the first case, the server was running DDL in 2 threads when it was sigkilled. The last queries according to the general log were
27 Query DROP /* QNO 16561 CON_ID 27 */ TABLE IF EXISTS vv11, tt13 |
26 Query ALTER /* QNO 507 CON_ID 26 */ TABLE t3 ENGINE=InnoDB, ADD COLUMN IF NOT EXISTS icol3 MEDIUMINT UNSIGNED ZEROFILL DEFAULT 2, LOCK=SHARED |
In the second case, a previously stored backup from MariaBackup was being restored.
Logs and data directories are available for both cases. The rr profiles are available for the first one.
At least the recovery failure on previously created datadir from case 1 is reproducible on bb-10.6-monty-innodb 98ff31770.
Attachments
Issue Links
- relates to
-
MDEV-17567 Atomic DDL
-
- Closed
-