[MDEV-9289] Batch DDL causes MariaDB 10.0.22 Galera Cluster hang on Ubuntu 12.04 Created: 2015-12-15  Updated: 2019-12-12  Resolved: 2019-12-12

Status: Closed
Project: MariaDB Server
Component/s: Galera
Affects Version/s: 10.0.22-galera
Fix Version/s: N/A

Type: Bug Priority: Major
Reporter: Zhixin Zhang Assignee: Seppo Jaakola
Resolution: Won't Fix Votes: 0
Labels: galera
Environment:

Distributor ID: Ubuntu
Description: Ubuntu 12.04.4 LTS
Release: 12.04
Codename: precise

Linux web104 3.2.0-58-generic #88-Ubuntu SMP Tue Dec 3 17:37:58 UTC 2013 x86_64 x86_64 x86_64 GNU/Linux


Attachments: Text File system_variables.txt     File test_sql.sql    

 Description   

How to reproduce:
Set up 3 mariadb galera cluster nodes: A, B, C. create one table on node A like this:

create table foo (a int);

And then do batch ddl like this.

create table foo_0 like foo;
create table foo_1 like foo;
create table foo_2 like foo;
...
create table foo_99 like foo;

Then the cluster hanged. run show processlist on node b, the output was like this:

MariaDB [(none)]> show processlist;
+----+-------------+-----------+------+---------+------+--------------------+------------------------------+----------+
| Id | User        | Host      | db   | Command | Time | State              | Info                         | Progress |
+----+-------------+-----------+------+---------+------+--------------------+------------------------------+----------+
|  1 | system user |           | hebi | Sleep   |    7 | exit open_tables() | create table foo_28 like foo |    0.000 |
|  2 | system user |           | NULL | Sleep   | 4377 | wsrep aborter idle | NULL                         |    0.000 |
 

run percona-toolkit pt-pmp on node b, the output was like this

Tue Dec 15 13:21:13 CST 2015
     78 pthread_cond_wait,toku_cond_wait(toku_pthread.h:309),kwait(toku_pthread.h:309),work_on_kibbutz(toku_pthread.h:309),start_thread(libpthread.so.0),clone(libc.so.6),??
     31 pthread_cond_wait,galera::Monitor::enter(libgalera_smm.so),galera::TrxHandle*)(libgalera_smm.so),galera::TrxHandle*)(libgalera_smm.so),gcs_action(libgalera_smm.so),bool&)(libgalera_smm.so),galera::ReplicatorSMM::async_recv(libgalera_smm.so),galera_recv(libgalera_smm.so),wsrep_replication_process(wsrep_thd.cc:314),start_wsrep_THD(mysqld.cc:5471),start_thread(libpthread.so.0),clone(libc.so.6),??
     10 libaio::??(libaio.so.1),os_aio_linux_collect(os0file.cc:5157),os_aio_linux_handle(os0file.cc:5157),fil_aio_wait(fil0fil.cc:5687),io_handler_thread(srv0start.cc:532),start_thread(libpthread.so.0),clone(libc.so.6),??
 

Accoding to the output, it seems that the tokudb engine causes the issue. So I disabled the tokudb engine, the issue was solved.

BTW, I can't reproduced the issue on 10.0.22-galera+ubuntu 14.04, pxc , 10.1.9-galera.



 Comments   
Comment by Jan Lindström (Inactive) [ 2019-12-12 ]

Support for 10.0-galera has ended.

Generated at Thu Feb 08 07:33:33 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.