MDEV-29878 Galera test failure on MDEV-26575

Test MDEV-26575 fails when it runs after MDEV-25389. This is because the latter simulates a failure while an applier thread is created in `start_wsrep_THD()`. The failure was not handled correctly and would not cleanup the created THD from the global `server_threads`. A subsequent shutdown would hang and eventually fail trying to close this THD. Reviewed-by: Jan Lindström <jan.lindstrom@mariadb.com>

Author: Daniele Sciascia

mysql-test/suite/galera/r/MDEV-25389.result

@@ -1,5 +1,7 @@
 connection node_2;
 connection node_1;
+connection node_1;
+connection node_2;
 connection node_2;
 call mtr.add_suppression("WSREP: Failed to create/initialize system thread");
 SET GLOBAL debug_dbug='+d,simulate_failed_connection_1';

mysql-test/suite/galera/t/MDEV-25389.test

@@ -2,6 +2,10 @@
 --source include/have_debug.inc
 --source include/have_debug_sync.inc
 
+--let $node_1=node_1
+--let $node_2=node_2
+--source ../galera/include/auto_increment_offset_save.inc
+
 --connection node_2
 call mtr.add_suppression("WSREP: Failed to create/initialize system thread");
 SET GLOBAL debug_dbug='+d,simulate_failed_connection_1';
@@ -11,3 +15,10 @@ SELECT @@wsrep_slave_threads;
 SET GLOBAL debug_dbug='';
 SET GLOBAL wsrep_slave_threads=1;
 SELECT @@wsrep_slave_threads;
+
+# MDEV-29878: this test caused a subsequent test to fail
+# during shutdown. Do a restart here, to make sure the
+# issue is fixed.
+--source include/restart_mysqld.inc
+
+--source ../galera/include/auto_increment_offset_restore.inc

sql/wsrep_mysqld.cc

@@ -2901,7 +2901,7 @@ int wsrep_create_trigger_query(THD *thd, uchar** buf, size_t* buf_len)
 
 void* start_wsrep_THD(void *arg)
 {
- THD *thd;
+ THD *thd= NULL;
 
  Wsrep_thd_args* thd_args= (Wsrep_thd_args*) arg;
 
@@ -2939,10 +2939,6 @@ void* start_wsrep_THD(void *arg)
  thd->thr_create_utime= microsecond_interval_timer();
  if (MYSQL_CALLBACK_ELSE(thread_scheduler, init_new_connection_thread, (), 0))
  {
- close_connection(thd, ER_OUT_OF_RESOURCES);
- statistic_increment(aborted_connects,&LOCK_status);
- // This will signal error to wsrep_slave_threads_update
- wsrep_thread_create_failed.store(true, std::memory_order_relaxed);
  WSREP_DEBUG("start_wsrep_THD: init_new_connection_thread failed");
  goto error;
  }
@@ -2964,11 +2960,6 @@ void* start_wsrep_THD(void *arg)
  wsrep_assign_from_threadvars(thd);
  if (wsrep_store_threadvars(thd))
  {
- close_connection(thd, ER_OUT_OF_RESOURCES);
- statistic_increment(aborted_connects,&LOCK_status);
- MYSQL_CALLBACK(thread_scheduler, end_thread, (thd, 0));
- delete thd;
- delete thd_args;
  goto error;
  }
 
@@ -3061,6 +3052,18 @@ void* start_wsrep_THD(void *arg)
 error:
  WSREP_ERROR("Failed to create/initialize system thread");
 
+ if (thd)
+ {
+ close_connection(thd, ER_OUT_OF_RESOURCES);
+ statistic_increment(aborted_connects, &LOCK_status);
+ server_threads.erase(thd);
+ delete thd;
+ my_thread_end();
+ }
+ delete thd_args;
+ // This will signal error to wsrep_slave_threads_update
+ wsrep_thread_create_failed.store(true, std::memory_order_relaxed);
+
  /* Abort if its the first applier/rollbacker thread. */
  if (!mysqld_server_initialized)
  unireg_abort(1);

sql/wsrep_thd.cc

@@ -23,8 +23,7 @@
 #include "rpl_rli.h"
 #include "log_event.h"
 #include "sql_parse.h"
-#include "mysqld.h" // start_wsrep_THD();
-#include "wsrep_applier.h" // start_wsrep_THD();
+#include "wsrep_mysqld.h" // start_wsrep_THD();
 #include "mysql/service_wsrep.h"
 #include "debug_sync.h"
 #include "slave.h"