mirror of
https://github.com/MariaDB/server.git
synced 2025-11-27 05:41:41 +03:00
7bffe468b2
When high priority replication slave applier encounters lock conflict in innodb, it will force the conflicting lock holder transaction (victim) to rollback. This is a must in multi-master sychronous replication model to avoid cluster lock-up. This high priority victim abort (aka "brute force" (BF) abort), is started from innodb lock manager while holding the victim's transaction's (trx) mutex. Depending on the execution state of the victim transaction, it may happen that the BF abort will call for THD::awake() to wake up the victim transaction for the rollback. Now, if BF abort requires THD::awake() to be called, then the applier thread executed locking protocol of: victim trx mutex -> victim THD::LOCK_thd_data If, at the same time another DBMS super user issues KILL command to abort the same victim, it will execute locking protocol of: victim THD::LOCK_thd_data -> victim trx mutex. These two locking protocol acquire mutexes in opposite order, hence unresolvable mutex locking deadlock may occur. The fix in this commit adds THD::wsrep_aborter flag to synchronize who can kill the victim This flag is set both when BF is called for from innodb and by KILL command. Either path of victim killing will bail out if victim's wsrep_killed is already set to avoid mutex conflicts with the other aborter execution. THD::wsrep_aborter records the aborter THD's ID. This is needed to preserve the right to kill the victim from different locations for the same aborter thread. It is also good error logging, to see who is reponsible for the abort. A new test case was added in galera.galera_bf_kill_debug.test for scenario where wsrep applier thread and manual KILL command try to kill same idle victim
141 lines
3.6 KiB
Plaintext
141 lines
3.6 KiB
Plaintext
--source include/galera_cluster.inc
|
|
--source include/have_innodb.inc
|
|
--source include/have_debug.inc
|
|
--source include/have_debug_sync.inc
|
|
|
|
#
|
|
# Test case 7:
|
|
# 1. Start a transaction on node_2,
|
|
# and leave it pending while holding a row locked
|
|
# 2. set sync point pause applier
|
|
# 3. send a conflicting write on node_1, it will pause
|
|
# at the sync point
|
|
# 4. though another connection to node_2, kill the local
|
|
# transaction
|
|
#
|
|
|
|
--connection node_2
|
|
CREATE TABLE t1(a int not null primary key auto_increment,b int) engine=InnoDB;
|
|
insert into t1 values (NULL,1);
|
|
|
|
#
|
|
# connection node_2a runs a local transaction, that is victim of BF abort
|
|
# and victim of KILL command by connection node_2
|
|
#
|
|
--connect node_2a, 127.0.0.1, root, , test, $NODE_MYPORT_2
|
|
--connection node_2a
|
|
truncate t1;
|
|
insert into t1 values (1,0);
|
|
|
|
# start a transaction that will conflict with later applier
|
|
begin;
|
|
update t1 set b=2 where a=1;
|
|
|
|
--connection node_2
|
|
set session wsrep_sync_wait=0;
|
|
--let $wait_condition = SELECT COUNT(*) = 1 FROM INFORMATION_SCHEMA.PROCESSLIST WHERE USER = 'root' AND COMMAND = 'Sleep' LIMIT 1
|
|
--source include/wait_condition.inc
|
|
|
|
--let $k_thread = `SELECT ID FROM INFORMATION_SCHEMA.PROCESSLIST WHERE USER = 'root' AND COMMAND = 'Sleep' LIMIT 1`
|
|
|
|
# connection node_2b is for controlling debug syn points
|
|
# first set a sync point for applier, to pause during BF aborting
|
|
# and before THD::awake would be called
|
|
#
|
|
--connect node_2b, 127.0.0.1, root, , test, $NODE_MYPORT_2
|
|
--connection node_2b
|
|
SET GLOBAL debug_dbug = "d,sync.before_wsrep_thd_abort";
|
|
|
|
#
|
|
# replicate an update, which will BF abort the victim node_2a
|
|
# however, while applier in node 2 is handling the abort,
|
|
# it will pause in sync point set by node_2b
|
|
#
|
|
--connection node_1
|
|
select * from t1;
|
|
update t1 set b= 1 where a=1;
|
|
|
|
#
|
|
# wait until the applying of above update has reached the sync point
|
|
# in node 2
|
|
#
|
|
--connection node_2b
|
|
SET SESSION DEBUG_SYNC = "now WAIT_FOR sync.before_wsrep_thd_abort_reached";
|
|
|
|
--connection node_2
|
|
#
|
|
# pause KILL execution before awake
|
|
#
|
|
SET DEBUG_SYNC= 'before_awake_no_mutex SIGNAL awake_reached WAIT_FOR continue_kill';
|
|
--disable_query_log
|
|
--send_eval KILL $k_thread
|
|
--enable_query_log
|
|
|
|
|
|
--connection node_2b
|
|
SET DEBUG_SYNC='now WAIT_FOR awake_reached';
|
|
|
|
# release applier and KILL operator
|
|
SET GLOBAL debug_dbug = "";
|
|
SET DEBUG_SYNC = "now SIGNAL signal.before_wsrep_thd_abort";
|
|
SET DEBUG_SYNC = "now SIGNAL continue_kill";
|
|
|
|
--connection node_2
|
|
--reap
|
|
|
|
--connection node_2a
|
|
--error 0,1213
|
|
select * from t1;
|
|
|
|
--connection node_2
|
|
SET DEBUG_SYNC = "RESET";
|
|
|
|
drop table t1;
|
|
|
|
--disconnect node_2a
|
|
#
|
|
# Test case 7:
|
|
# run a transaction in node 2, and set a sync point to pause the transaction
|
|
# in commit phase.
|
|
# Through another connection to node 2, kill the committing transaction by
|
|
# KILL QUERY command
|
|
#
|
|
|
|
--connect node_2a, 127.0.0.1, root, , test, $NODE_MYPORT_2
|
|
--connection node_2a
|
|
--let $connection_id = `SELECT CONNECTION_ID()`
|
|
|
|
CREATE TABLE t1 (i int primary key);
|
|
|
|
# Set up sync point
|
|
SET DEBUG_SYNC = "before_wsrep_ordered_commit SIGNAL bwoc_reached WAIT_FOR bwoc_continue";
|
|
|
|
# Send insert which will block in the sync point above
|
|
--send INSERT INTO t1 VALUES (1)
|
|
|
|
--connection node_2
|
|
SET DEBUG_SYNC = "now WAIT_FOR bwoc_reached";
|
|
|
|
--disable_query_log
|
|
--disable_result_log
|
|
# victim has passed the point of no return, kill is not possible anymore
|
|
--eval KILL QUERY $connection_id
|
|
--enable_result_log
|
|
--enable_query_log
|
|
|
|
SET DEBUG_SYNC = "now SIGNAL bwoc_continue";
|
|
SET DEBUG_SYNC='RESET';
|
|
--connection node_2a
|
|
--error 0,1213
|
|
--reap
|
|
|
|
--connection node_2
|
|
# victim was able to complete the INSERT
|
|
select * from t1;
|
|
|
|
--disconnect node_2a
|
|
|
|
--connection node_1
|
|
drop table t1;
|
|
|