Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Cannot Reproduce
-
10.2(EOL), 10.3(EOL), 10.4(EOL)
Description
http://buildbot.askmonty.org/buildbot/builders/win32-debug/builds/10757
innodb.undo_truncate_recover '16k,2,innodb' w2 [ fail ]
|
Test ended at 2018-10-11 20:49:09
|
|
CURRENT_TEST: innodb.undo_truncate_recover
|
mysqltest: In included file "./include/shutdown_mysqld.inc":
|
included from D:/win32-debug/build/src/mysql-test/suite/innodb/t/undo_truncate_recover.test at line 45:
|
At line 40: 'SHOW VARIABLES LIKE 'pid_file'' failed: 2006 MySQL server has gone away
|
|
The result from queries just before the failure was:
|
SET GLOBAL innodb_undo_logs = 4;
|
SET GLOBAL innodb_undo_log_truncate = 1;
|
SET GLOBAL innodb_purge_rseg_truncate_frequency = 1;
|
create table t1(keyc int primary key, c char(100)) engine = innodb;
|
begin;
|
commit;
|
begin;
|
update t1 set c = 'MariaDB';
|
update t1 set c = 'InnoDB';
|
set global debug_dbug = '+d,ib_undo_trunc';
|
commit;
|
call mtr.add_suppression("InnoDB: The transaction log size is too large");
|
SET GLOBAL innodb_fast_shutdown=0;
|
Attachments
Issue Links
- relates to
-
MDEV-17780 innodb.truncate_recover crashes in recovery due to out-of-bounds page read
-
- Closed
-
-
MDEV-17049 Enable --suite=innodb_undo on buildbot
-
- Closed
-
As far as I can tell, the statement ought to be part of the following in shutdown_mysqld.inc:
{
…
&ds_pidfile_name))
Apparently, the server was killed due to some reason while the shutdown as about to be initiated. Here is an excerpt from the server error log:
10.2 81a5b6ccd531e802bbb0f2034189095d2196de8b
Version: '10.2.19-MariaDB-debug-log' socket: '' port: 16020 Source distribution
2018-10-11 20:48:58 2080 [Note] InnoDB: Truncating UNDO tablespace 1
2018-10-11 20:48:59 2080 [Warning] InnoDB: The transaction log size is too large for innodb_log_buffer_size (1389258 >= 1048576 / 2). Trying to extend it.
2018-10-11 20:48:59 2080 [Note] InnoDB: innodb_log_buffer_size was extended to 1392640.
2018-10-11 20:48:59 2080 [Note] InnoDB: ib_undo_trunc
CURRENT_TEST: innodb.undo_truncate_recover
The last message is a suicide note from trx_purge_initiate_truncate():
DBUG_SUICIDE(););
This looks like a race condition in the test.
We do expect the server to commit suicide during the slow shutdown, but not already during the initial step of shutdown_server.