[MDEV-17357] Test failure on galera.galera_pc_ignore_sb Created: 2018-10-03  Updated: 2020-01-06  Resolved: 2019-12-23

Status: Closed
Project: MariaDB Server
Component/s: Galera, Tests
Affects Version/s: 10.1
Fix Version/s: N/A

Type: Bug Priority: Major
Reporter: Jan Lindström (Inactive) Assignee: Jan Lindström (Inactive)
Resolution: Won't Fix Votes: 0
Labels: None

Issue Links:
Relates
relates to MDEV-20888 Galera test failure on galera.galera_... Closed

 Description   

Surfaced after galera library upgrade:

galera.galera_pc_ignore_sb 'innodb_plugin' w4 [ fail ]
        Test ended at 2018-10-01 09:44:31
 
CURRENT_TEST: galera.galera_pc_ignore_sb
mysqltest: At line 38: query 'SET GLOBAL wsrep_provider_options = '$wsrep_provider_options_orig';' failed: 1210: Incorrect arguments to SET
 
The result from queries just before the failure was:
SET GLOBAL wsrep_provider_options = 'pc.ignore_sb=true';
Killing server ...
CREATE TABLE t1 (f1 INTEGER) ENGINE=InnoDB;
INSERT INTO t1 VALUES (1);
DROP TABLE t1;
SELECT VARIABLE_VALUE = 1 FROM INFORMATION_SCHEMA.GLOBAL_STATUS WHERE VARIABLE_NAME = 'wsrep_cluster_size';
VARIABLE_VALUE = 1
1
SELECT VARIABLE_VALUE = 'ON' FROM INFORMATION_SCHEMA.GLOBAL_STATUS WHERE VARIABLE_NAME = 'wsrep_ready';
VARIABLE_VALUE = 'ON'
1
SET GLOBAL wsrep_cluster_address = '';



 Comments   
Comment by Jan Lindström (Inactive) [ 2018-10-04 ]

After several tries had to give up current version:

http://buildbot.askmonty.org/buildbot/builders/kvm-rpm-centos74-amd64/builds/4633/steps/mtr/logs/stdio

galera.galera_pc_ignore_sb 'innodb_plugin' w4 [ fail ]
        Test ended at 2018-10-03 17:31:50
 
CURRENT_TEST: galera.galera_pc_ignore_sb
mysqltest: In included file "./include/galera_wait_ready.inc": 
included from /usr/share/mysql-test/suite/galera/include/start_mysqld.inc at line 16:
included from /usr/share/mysql-test/suite/galera/t/galera_pc_ignore_sb.test at line 43:
At line 28: "Server did not transition to READY state"
 
The result from queries just before the failure was:
SET @wsrep_cluster_address_orig = @@GLOBAL.wsrep_cluster_address;
SET @wsrep_provider_options_orig = @@GLOBAL.wsrep_provider_options;
SET GLOBAL wsrep_provider_options ='pc.ignore_sb=true';
Killing server ...
CREATE TABLE t1 (f1 INTEGER) ENGINE=InnoDB;
INSERT INTO t1 VALUES (1);
DROP TABLE t1;
SELECT VARIABLE_VALUE = 1 FROM INFORMATION_SCHEMA.GLOBAL_STATUS WHERE VARIABLE_NAME = 'wsrep_cluster_size';
VARIABLE_VALUE = 1
1
SELECT VARIABLE_VALUE = 'ON' FROM INFORMATION_SCHEMA.GLOBAL_STATUS WHERE VARIABLE_NAME = 'wsrep_ready';
VARIABLE_VALUE = 'ON'
1
SET GLOBAL wsrep_cluster_address = '';
SET GLOBAL wsrep_cluster_address = @wsrep_cluster_address_orig;
 
 - saving '/dev/shm/var/4/log/galera.galera_pc_ignore_sb-innodb_plugin/' to '/dev/shm/var/log/galera.galera_pc_ignore_sb-innodb_plugin/'
 
Retrying test galera.galera_pc_ignore_sb, attempt(2/3)...

Somehow we need to move from split brain situation to normal cluster reliable way. Note that test case does not fail every time only most of the time and more on release builds.

Comment by Mario Karuza (Inactive) [ 2018-10-10 ]

Problem is triggered after execution of galera.galera_drop_database test introduced with MDEV-16656. This test should leave cluster in original state before the end of it but it seems that this is not case.
Problem which is reported here is that after change of 'wsrep_cluster_address' mysqld.1 is not in PRIMARY and will timeout.

Generated at Thu Feb 08 08:35:51 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.