[MDEV-26499] galera_3nodes.galera_ipv6_mysqldump MTR failed: mysql_shutdown failed Created: 2021-08-30  Updated: 2023-10-31

Status: Stalled
Project: MariaDB Server
Component/s: Galera
Affects Version/s: 10.6
Fix Version/s: 10.6

Type: Bug Priority: Major
Reporter: Ramesh Sivaraman Assignee: Julius Goryavsky
Resolution: Unresolved Votes: 0
Labels: None

Issue Links:
Blocks
blocks MDEV-26498 Galera test failures on 10.6/10.7 Open
Relates
relates to MDEV-24481 galera_3nodes.galera_vote_rejoin_mysq... Closed
relates to MDEV-32635 galera_shutdown_nonprim: mysql_shutd... Closed

 Description   

stdio.log

10.6.5 7271cf48d62196abc0c578d00ab3ca09b56aae77

galera_3nodes.galera_ipv6_mysqldump 'innodb' w1 [ fail ]
        Test ended at 2021-08-29 06:12:51
 
CURRENT_TEST: galera_3nodes.galera_ipv6_mysqldump
mysqltest: In included file "./include/shutdown_mysqld.inc": 
included from ./include/restart_mysqld.inc at line 10:
included from /usr/share/mysql/mysql-test/suite/galera_3nodes/t/galera_ipv6_mysqldump.test at line 85:
At line 50: mysql_shutdown failed



 Comments   
Comment by Daniele Sciascia [ 2022-11-24 ]

I found an issue and a test case that causes mysql_shutdown in mysqltest client to fail as described in this ticket.
The MTR test to reproduce:

--source include/galera_cluster.inc                      
--source include/have_debug_sync.inc                     
                                                         
--let $node_1=node_1                                     
--let $node_2=node_2                                     
--source include/auto_increment_offset_save.inc          
                                                         
--connection node_2                                      
SET GLOBAL debug_dbug="+d,simulate_shutdown_client_slow";
--source include/restart_mysqld.inc                      
                                                         
--source include/auto_increment_offset_restore.inc       

and this DBUG_EXECUTE_IF in sql_parse.cc:

diff --git a/sql/sql_parse.cc b/sql/sql_parse.cc
index 0261e7480f1..71d228f3338 100644
--- a/sql/sql_parse.cc
+++ b/sql/sql_parse.cc
@@ -2206,6 +2206,7 @@ bool dispatch_command(enum enum_server_command command, THD *thd,
     my_eof(thd);
     kill_mysql(thd);
     error=TRUE;
+    DBUG_EXECUTE_IF("simulate_shutdown_client_slow", my_sleep(1000000););
     break;
   }
 #endif

I tested this in 10.4, but I presume it's the same issue. I could take this ticket over, since it is stalled and I have a tentative fix.

Generated at Thu Feb 08 09:45:46 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.