Details
-
Type:
Bug
-
Status: Closed
-
Priority:
Major
-
Resolution: Fixed
-
Affects Version/s: 5.5.40-galera, 10.0.15-galera
-
Fix Version/s: 5.5.41-galera, 10.0.16-galera
-
Component/s: wsrep
-
Labels:None
Description
Start a node with wsrep provider and wsrep_causal_reads=ON and then execute the following commands :
MariaDB [test]> show status like 'wsrep_ready'; +---------------+-------+ | Variable_name | Value | +---------------+-------+ | wsrep_ready | ON | +---------------+-------+ 1 row in set (0.64 sec) MariaDB [test]> set global wsrep_provider="/usr/lib/galera/libgalera_smm.so"; Query OK, 0 rows affected (2.42 sec) MariaDB [test]> show status like 'wsrep_ready'; ERROR 1205 (HY000): Lock wait timeout exceeded; try restarting transaction
Gliffy Diagrams
Attachments
Activity
- All
- Comments
- Work Log
- History
- Activity
- Transitions
Only one small question, why call mtr.add_suppression("safe_mutex: Found wrong usage of mutex.*"); is needed ? Sounds like a bug.
R: Jan
Its because in wsrep_cluster_address_update LOCK_global_system_variables is released
temporarily during closure of client connections.
/* stop replication is heavy operation, and includes closing all client $
connections. Closing clients may need to get LOCK_global_system_variables$
at least in MariaDB.$
$
Note: releasing LOCK_global_system_variables may cause race condition, if $
there can be several concurrent clients changing wsrep_provider$
*/$
mysql_mutex_unlock(&LOCK_global_system_variables);$
wsrep_stop_replication(thd);$
mysql_mutex_lock(&LOCK_global_system_variables);$
I will check further is this is really needed here and remove if not.
ok to push.
http://lists.askmonty.org/pipermail/commits/2014-December/007214.html