mariadb/sql/service_wsrep.cc

404 lines
11 KiB
C++
Raw Normal View History

/* Copyright 2018-2023 Codership Oy <info@codership.com>
2019-01-23 12:30:00 +01:00
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; version 2 of the License.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA */
#include "mariadb.h"
#include "mysql/service_wsrep.h"
#include "wsrep/key.hpp"
#include "wsrep_thd.h"
#include "wsrep_trans_observer.h"
#include "sql_class.h"
#include "debug_sync.h"
10.4 wsrep group commit fixes (#1224) * MDEV-16509 Improve wsrep commit performance with binlog disabled Release commit order critical section early after trx_commit_low() if binlog is not transaction coordinator. In order to avoid two phase commit, binlog_hton is not registered for THD during IO_CACHE population. Implemented a test which verifies that the transactions release commit order early. This optimization will change behavior during recovery as the commit is not two phase when binlog is off. Fixed and recorded wsrep-recover-v25 and wsrep-recover to match the behavior. * MDEV-18730 Ordering for wsrep binlog group commit Previously out of order execution was allowed for wsrep commits. Established proper ordering by populating wait_for_commit for every wsrep THD and making group commit leader to wait for prior commits before proceeding to trx_group_commit_leader(). * MDEV-18730 Added a test case to verify correct commit ordering * MDEV-16509, MDEV-18730 Review fixes Use WSREP_EMULATE_BINLOG() macro to decide if the binlog_hton should be registered. Whitespace/syntax fixes and cleanups. * MDEV-16509 Require binlog for galera_var_innodb_disallow_writes test If the commit to InnoDB is done in one phase, the native InnoDB behavior is that the transaction is committed in memory before it is persisted to disk. This means that the innodb_disallow_writes=ON may not prevent transaction to become visible to other readers before commit is completely over. On the other hand, if the commit is two phase (as it is with binlog), the transaction will be blocked in prepare phase. Fixed the test to use binlog, which enforces two phase commit, which in turn makes commit to block before the changes become visible to other connections. This guarantees that the test produces expected result.
2019-03-15 06:09:13 +01:00
#include "log.h"
2019-01-23 12:30:00 +01:00
extern "C" my_bool wsrep_on(const THD *thd)
{
return my_bool(WSREP(thd));
}
extern "C" void wsrep_thd_LOCK(const THD *thd)
{
mysql_mutex_lock(&thd->LOCK_thd_data);
}
MDEV-29293 MariaDB stuck on starting commit state This commit contains a merge from 10.5-MDEV-29293-squash into 10.6. Although the bug MDEV-29293 was not reproducible with 10.6, the fix contains several improvements for wsrep KILL query and BF abort handling, and addresses the following issues: * MDEV-30307 KILL command issued inside a transaction is problematic for galera replication: This commit will remove KILL TOI replication, so Galera side transaction context is not lost during KILL. * MDEV-21075 KILL QUERY maintains nodes data consistency but breaks GTID sequence: This is fixed as well as KILL does not use TOI, and thus does not change GTID state. * MDEV-30372 Assertion in wsrep-lib state: This was caused by BF abort or KILL when local transaction was in the middle of group commit. This commit disables THD::killed handling during commit, so the problem is avoided. * MDEV-30963 Assertion failure !lock.was_chosen_as_deadlock_victim in trx0trx.h:1065: The assertion happened when the victim was BF aborted via MDL while it was committing. This commit changes MDL BF aborts so that transactions which are committing cannot be BF aborted via MDL. The RQG grammar attached in the issue could not reproduce the crash anymore. Original commit message from 10.5 fix: MDEV-29293 MariaDB stuck on starting commit state The problem seems to be a deadlock between KILL command execution and BF abort issued by an applier, where: * KILL has locked victim's LOCK_thd_kill and LOCK_thd_data. * Applier has innodb side global lock mutex and victim trx mutex. * KILL is calling innobase_kill_query, and is blocked by innodb global lock mutex. * Applier is in wsrep_innobase_kill_one_trx and is blocked by victim's LOCK_thd_kill. The fix in this commit removes the TOI replication of KILL command and makes KILL execution less intrusive operation. Aborting the victim happens now by using awake_no_mutex() and ha_abort_transaction(). If the KILL happens when the transaction is committing, the KILL operation is postponed to happen after the statement has completed in order to avoid KILL to interrupt commit processing. Notable changes in this commit: * wsrep client connections's error state may remain sticky after client connection is closed. This error message will then pop up for the next client session issuing first SQL statement. This problem raised with test galera.galera_bf_kill. The fix is to reset wsrep client error state, before a THD is reused for next connetion. * Release THD locks in wsrep_abort_transaction when locking innodb mutexes. This guarantees same locking order as with applier BF aborting. * BF abort from MDL was changed to do BF abort on server/wsrep-lib side first, and only then do the BF abort on InnoDB side. This removes the need to call back from InnoDB for BF aborts which originate from MDL and simplifies the locking. * Removed wsrep_thd_set_wsrep_aborter() from service_wsrep.h. The manipulation of the wsrep_aborter can be done solely on server side. Moreover, it is now debug only variable and could be excluded from optimized builds. * Remove LOCK_thd_kill from wsrep_thd_LOCK/UNLOCK to allow more fine grained locking for SR BF abort which may require locking of victim LOCK_thd_kill. Added explicit call for wsrep_thd_kill_LOCK/UNLOCK where appropriate. * Wsrep-lib was updated to version which allows external locking for BF abort calls. Changes to MTR tests: * Disable galera_bf_abort_group_commit. This test is going to be removed (MDEV-30855). * Make galera_var_retry_autocommit result more readable by echoing cases and expectations into result. Only one expected result for reap to verify that server returns expected status for query. * Record galera_gcache_recover_manytrx as result file was incomplete. Trivial change. * Make galera_create_table_as_select more deterministic: Wait until CTAS execution has reached MDL wait for multi-master conflict case. Expected error from multi-master conflict is ER_QUERY_INTERRUPTED. This is because CTAS does not yet have open wsrep transaction when it is waiting for MDL, query gets interrupted instead of BF aborted. This should be addressed in separate task. * A new test galera_bf_abort_registering to check that registering trx gets BF aborted through MDL. * A new test galera_kill_group_commit to verify correct behavior when KILL is executed while the transaction is committing. Co-authored-by: Seppo Jaakola <seppo.jaakola@iki.fi> Co-authored-by: Jan Lindström <jan.lindstrom@galeracluster.com> Signed-off-by: Julius Goryavsky <julius.goryavsky@mariadb.com>
2023-04-19 15:51:55 +02:00
extern "C" int wsrep_thd_TRYLOCK(const THD *thd)
{
return mysql_mutex_trylock(&thd->LOCK_thd_data);
}
2019-01-23 12:30:00 +01:00
extern "C" void wsrep_thd_UNLOCK(const THD *thd)
{
mysql_mutex_unlock(&thd->LOCK_thd_data);
}
extern "C" void wsrep_thd_kill_LOCK(const THD *thd)
{
mysql_mutex_lock(&thd->LOCK_thd_kill);
}
extern "C" void wsrep_thd_kill_UNLOCK(const THD *thd)
{
mysql_mutex_unlock(&thd->LOCK_thd_kill);
}
2019-01-23 12:30:00 +01:00
extern "C" const char* wsrep_thd_client_state_str(const THD *thd)
{
return wsrep::to_c_string(thd->wsrep_cs().state());
}
extern "C" const char* wsrep_thd_client_mode_str(const THD *thd)
{
return wsrep::to_c_string(thd->wsrep_cs().mode());
}
extern "C" const char* wsrep_thd_transaction_state_str(const THD *thd)
{
return wsrep::to_c_string(thd->wsrep_cs().transaction().state());
}
extern "C" const char *wsrep_thd_query(const THD *thd)
{
if (!thd)
return "NULL";
switch(thd->lex->sql_command)
2019-11-01 14:23:18 +01:00
{
// Mask away some security related details from error log
2019-11-01 14:23:18 +01:00
case SQLCOM_CREATE_USER:
return "CREATE USER";
case SQLCOM_GRANT:
return "GRANT";
case SQLCOM_REVOKE:
return "REVOKE";
case SQLCOM_SET_OPTION:
if (thd->lex->definer)
return "SET PASSWORD";
2019-11-01 14:23:18 +01:00
/* fallthrough */
default:
return (thd->query() ? thd->query() : "NULL");
2019-11-01 14:23:18 +01:00
}
return "NULL";
2019-01-23 12:30:00 +01:00
}
extern "C" query_id_t wsrep_thd_transaction_id(const THD *thd)
{
return thd->wsrep_cs().transaction().id().get();
}
extern "C" long long wsrep_thd_trx_seqno(const THD *thd)
{
const wsrep::client_state& cs= thd->wsrep_cs();
if (cs.mode() == wsrep::client_state::m_toi)
{
return cs.toi_meta().seqno().get();
}
else
{
return cs.transaction().ws_meta().seqno().get();
}
}
extern "C" void wsrep_thd_self_abort(THD *thd)
{
thd->wsrep_cs().bf_abort(wsrep::seqno(0));
}
extern "C" const char* wsrep_get_sr_table_name()
{
return wsrep_sr_table_name_full;
}
extern "C" my_bool wsrep_get_debug()
{
return wsrep_debug;
}
/*
Test if this connection is a true local (user) connection and not
a replication or wsrep applier thread.
Note that this is only usable for galera (as there are other kinds
of system threads, and only if WSREP_NNULL() is tested by the caller.
*/
2019-01-23 12:30:00 +01:00
extern "C" my_bool wsrep_thd_is_local(const THD *thd)
{
MDEV-21096 async slave crash with gtid_log_pos table access (#1413) The original crash happened when async replication IO thread was updating mysql.gtid_slave_pos table. Operations on this table should remain node local, but it appears that protection (THD::wsrep_ignore_table flag) to prevent wsrep replication for this table mas missing for innodb write_row() and update_row(). It was somewhat difficult to reproduce the issue, because mtr seems to create the affected table mysql.gtid_log_pos as of Aria engine type, and Aria engine operations will not be replicated anyhow. It looks, though, that in release installation, mysql.gtid_slave_pos table is of InnoDB engine. It was possible to trigger somewhat related problem by running test galera.galera_as_slave_gtid with configuration: gtid_pos_auto_engines=InnoDB. However, this test mode, causes earlier crash when replication background thread creates aditional table: mysql.gtid_slave_pos_InnoDB, and this table create triggered wsrep TOI replication, which also failed for assertion. Actually, async replication IO and background threads should not replicate anything to cluster. This pull request contains new test galera.galera_as_slave_gtid_auto_engine, which basically just runs galera.galera_as_slave_gtid with configuration of gtid_pos_auto_engines=InnoDB. Test galera.galera_as_slave_gtid is also modified for better code reuse. Actual fix for MDEV-21096 is in storage/innobase/handler/ha_innodb.cc, where THD::wsrep_ignore_table flag is now honored before wsrep key population. There is additional fix in sql/service_wsrep.cc where async replication IO and background threads are marked as non-local. This fences these threads out of wsrep replication altogether. Note that this change, actually makes the use of THD::wsrep_ignore-table redundant. We may want to refactor THD::wsrep_ignore_table out in the future, if there is no other use case for it in sight.
2019-11-25 10:19:33 +01:00
/*
async replication IO and background threads have nothing to
replicate in the cluster, marking them as non-local here to
prevent write set population and replication
MDEV-21096 async slave crash with gtid_log_pos table access (#1413) The original crash happened when async replication IO thread was updating mysql.gtid_slave_pos table. Operations on this table should remain node local, but it appears that protection (THD::wsrep_ignore_table flag) to prevent wsrep replication for this table mas missing for innodb write_row() and update_row(). It was somewhat difficult to reproduce the issue, because mtr seems to create the affected table mysql.gtid_log_pos as of Aria engine type, and Aria engine operations will not be replicated anyhow. It looks, though, that in release installation, mysql.gtid_slave_pos table is of InnoDB engine. It was possible to trigger somewhat related problem by running test galera.galera_as_slave_gtid with configuration: gtid_pos_auto_engines=InnoDB. However, this test mode, causes earlier crash when replication background thread creates aditional table: mysql.gtid_slave_pos_InnoDB, and this table create triggered wsrep TOI replication, which also failed for assertion. Actually, async replication IO and background threads should not replicate anything to cluster. This pull request contains new test galera.galera_as_slave_gtid_auto_engine, which basically just runs galera.galera_as_slave_gtid with configuration of gtid_pos_auto_engines=InnoDB. Test galera.galera_as_slave_gtid is also modified for better code reuse. Actual fix for MDEV-21096 is in storage/innobase/handler/ha_innodb.cc, where THD::wsrep_ignore_table flag is now honored before wsrep key population. There is additional fix in sql/service_wsrep.cc where async replication IO and background threads are marked as non-local. This fences these threads out of wsrep replication altogether. Note that this change, actually makes the use of THD::wsrep_ignore-table redundant. We may want to refactor THD::wsrep_ignore_table out in the future, if there is no other use case for it in sight.
2019-11-25 10:19:33 +01:00
async replication SQL thread, applies client transactions from
mariadb master and will be replicated into cluster
*/
MDEV-21096 async slave crash with gtid_log_pos table access (#1413) The original crash happened when async replication IO thread was updating mysql.gtid_slave_pos table. Operations on this table should remain node local, but it appears that protection (THD::wsrep_ignore_table flag) to prevent wsrep replication for this table mas missing for innodb write_row() and update_row(). It was somewhat difficult to reproduce the issue, because mtr seems to create the affected table mysql.gtid_log_pos as of Aria engine type, and Aria engine operations will not be replicated anyhow. It looks, though, that in release installation, mysql.gtid_slave_pos table is of InnoDB engine. It was possible to trigger somewhat related problem by running test galera.galera_as_slave_gtid with configuration: gtid_pos_auto_engines=InnoDB. However, this test mode, causes earlier crash when replication background thread creates aditional table: mysql.gtid_slave_pos_InnoDB, and this table create triggered wsrep TOI replication, which also failed for assertion. Actually, async replication IO and background threads should not replicate anything to cluster. This pull request contains new test galera.galera_as_slave_gtid_auto_engine, which basically just runs galera.galera_as_slave_gtid with configuration of gtid_pos_auto_engines=InnoDB. Test galera.galera_as_slave_gtid is also modified for better code reuse. Actual fix for MDEV-21096 is in storage/innobase/handler/ha_innodb.cc, where THD::wsrep_ignore_table flag is now honored before wsrep key population. There is additional fix in sql/service_wsrep.cc where async replication IO and background threads are marked as non-local. This fences these threads out of wsrep replication altogether. Note that this change, actually makes the use of THD::wsrep_ignore-table redundant. We may want to refactor THD::wsrep_ignore_table out in the future, if there is no other use case for it in sight.
2019-11-25 10:19:33 +01:00
return (
thd->system_thread != SYSTEM_THREAD_SLAVE_BACKGROUND &&
thd->system_thread != SYSTEM_THREAD_SLAVE_IO &&
thd->wsrep_cs().mode() == wsrep::client_state::m_local);
2019-01-23 12:30:00 +01:00
}
extern "C" my_bool wsrep_thd_is_applying(const THD *thd)
{
return thd->wsrep_cs().mode() == wsrep::client_state::m_high_priority;
}
extern "C" my_bool wsrep_thd_is_toi(const THD *thd)
{
return thd->wsrep_cs().mode() == wsrep::client_state::m_toi;
}
extern "C" my_bool wsrep_thd_is_local_toi(const THD *thd)
{
return thd->wsrep_cs().mode() == wsrep::client_state::m_toi &&
thd->wsrep_cs().toi_mode() == wsrep::client_state::m_local;
}
extern "C" my_bool wsrep_thd_is_in_rsu(const THD *thd)
{
return thd->wsrep_cs().mode() == wsrep::client_state::m_rsu;
}
extern "C" my_bool wsrep_thd_is_BF(const THD *thd, my_bool sync)
{
my_bool status = FALSE;
if (thd && WSREP(thd))
{
if (sync) mysql_mutex_lock(&thd->LOCK_thd_data);
status = (wsrep_thd_is_applying(thd) || wsrep_thd_is_toi(thd));
if (sync) mysql_mutex_unlock(&thd->LOCK_thd_data);
}
return status;
}
extern "C" my_bool wsrep_thd_is_SR(const THD *thd)
{
return thd && thd->wsrep_cs().transaction().is_streaming();
}
extern "C" void wsrep_handle_SR_rollback(THD *bf_thd,
THD *victim_thd)
{
DBUG_ASSERT(victim_thd);
DBUG_ASSERT(wsrep_thd_is_SR(victim_thd));
2019-01-23 12:30:00 +01:00
if (!victim_thd || !wsrep_on(bf_thd)) return;
WSREP_DEBUG("handle rollback, for deadlock: thd %llu trx_id %" PRIu64 " frags %zu conf %s",
2019-01-23 12:30:00 +01:00
victim_thd->thread_id,
victim_thd->wsrep_trx_id(),
victim_thd->wsrep_sr().fragments_certified(),
wsrep_thd_transaction_state_str(victim_thd));
/* Note: do not store/reset globals before wsrep_bf_abort() call
to avoid losing BF thd context. */
MDEV-29293 MariaDB stuck on starting commit state This commit contains a merge from 10.5-MDEV-29293-squash into 10.6. Although the bug MDEV-29293 was not reproducible with 10.6, the fix contains several improvements for wsrep KILL query and BF abort handling, and addresses the following issues: * MDEV-30307 KILL command issued inside a transaction is problematic for galera replication: This commit will remove KILL TOI replication, so Galera side transaction context is not lost during KILL. * MDEV-21075 KILL QUERY maintains nodes data consistency but breaks GTID sequence: This is fixed as well as KILL does not use TOI, and thus does not change GTID state. * MDEV-30372 Assertion in wsrep-lib state: This was caused by BF abort or KILL when local transaction was in the middle of group commit. This commit disables THD::killed handling during commit, so the problem is avoided. * MDEV-30963 Assertion failure !lock.was_chosen_as_deadlock_victim in trx0trx.h:1065: The assertion happened when the victim was BF aborted via MDL while it was committing. This commit changes MDL BF aborts so that transactions which are committing cannot be BF aborted via MDL. The RQG grammar attached in the issue could not reproduce the crash anymore. Original commit message from 10.5 fix: MDEV-29293 MariaDB stuck on starting commit state The problem seems to be a deadlock between KILL command execution and BF abort issued by an applier, where: * KILL has locked victim's LOCK_thd_kill and LOCK_thd_data. * Applier has innodb side global lock mutex and victim trx mutex. * KILL is calling innobase_kill_query, and is blocked by innodb global lock mutex. * Applier is in wsrep_innobase_kill_one_trx and is blocked by victim's LOCK_thd_kill. The fix in this commit removes the TOI replication of KILL command and makes KILL execution less intrusive operation. Aborting the victim happens now by using awake_no_mutex() and ha_abort_transaction(). If the KILL happens when the transaction is committing, the KILL operation is postponed to happen after the statement has completed in order to avoid KILL to interrupt commit processing. Notable changes in this commit: * wsrep client connections's error state may remain sticky after client connection is closed. This error message will then pop up for the next client session issuing first SQL statement. This problem raised with test galera.galera_bf_kill. The fix is to reset wsrep client error state, before a THD is reused for next connetion. * Release THD locks in wsrep_abort_transaction when locking innodb mutexes. This guarantees same locking order as with applier BF aborting. * BF abort from MDL was changed to do BF abort on server/wsrep-lib side first, and only then do the BF abort on InnoDB side. This removes the need to call back from InnoDB for BF aborts which originate from MDL and simplifies the locking. * Removed wsrep_thd_set_wsrep_aborter() from service_wsrep.h. The manipulation of the wsrep_aborter can be done solely on server side. Moreover, it is now debug only variable and could be excluded from optimized builds. * Remove LOCK_thd_kill from wsrep_thd_LOCK/UNLOCK to allow more fine grained locking for SR BF abort which may require locking of victim LOCK_thd_kill. Added explicit call for wsrep_thd_kill_LOCK/UNLOCK where appropriate. * Wsrep-lib was updated to version which allows external locking for BF abort calls. Changes to MTR tests: * Disable galera_bf_abort_group_commit. This test is going to be removed (MDEV-30855). * Make galera_var_retry_autocommit result more readable by echoing cases and expectations into result. Only one expected result for reap to verify that server returns expected status for query. * Record galera_gcache_recover_manytrx as result file was incomplete. Trivial change. * Make galera_create_table_as_select more deterministic: Wait until CTAS execution has reached MDL wait for multi-master conflict case. Expected error from multi-master conflict is ER_QUERY_INTERRUPTED. This is because CTAS does not yet have open wsrep transaction when it is waiting for MDL, query gets interrupted instead of BF aborted. This should be addressed in separate task. * A new test galera_bf_abort_registering to check that registering trx gets BF aborted through MDL. * A new test galera_kill_group_commit to verify correct behavior when KILL is executed while the transaction is committing. Co-authored-by: Seppo Jaakola <seppo.jaakola@iki.fi> Co-authored-by: Jan Lindström <jan.lindstrom@galeracluster.com> Signed-off-by: Julius Goryavsky <julius.goryavsky@mariadb.com>
2023-04-19 15:51:55 +02:00
mysql_mutex_lock(&victim_thd->LOCK_thd_data);
if (!(bf_thd && bf_thd != victim_thd))
2019-01-23 12:30:00 +01:00
{
DEBUG_SYNC(victim_thd, "wsrep_before_SR_rollback");
}
if (bf_thd)
{
wsrep_bf_abort(bf_thd, victim_thd);
}
else
{
wsrep_thd_self_abort(victim_thd);
}
MDEV-29293 MariaDB stuck on starting commit state This commit contains a merge from 10.5-MDEV-29293-squash into 10.6. Although the bug MDEV-29293 was not reproducible with 10.6, the fix contains several improvements for wsrep KILL query and BF abort handling, and addresses the following issues: * MDEV-30307 KILL command issued inside a transaction is problematic for galera replication: This commit will remove KILL TOI replication, so Galera side transaction context is not lost during KILL. * MDEV-21075 KILL QUERY maintains nodes data consistency but breaks GTID sequence: This is fixed as well as KILL does not use TOI, and thus does not change GTID state. * MDEV-30372 Assertion in wsrep-lib state: This was caused by BF abort or KILL when local transaction was in the middle of group commit. This commit disables THD::killed handling during commit, so the problem is avoided. * MDEV-30963 Assertion failure !lock.was_chosen_as_deadlock_victim in trx0trx.h:1065: The assertion happened when the victim was BF aborted via MDL while it was committing. This commit changes MDL BF aborts so that transactions which are committing cannot be BF aborted via MDL. The RQG grammar attached in the issue could not reproduce the crash anymore. Original commit message from 10.5 fix: MDEV-29293 MariaDB stuck on starting commit state The problem seems to be a deadlock between KILL command execution and BF abort issued by an applier, where: * KILL has locked victim's LOCK_thd_kill and LOCK_thd_data. * Applier has innodb side global lock mutex and victim trx mutex. * KILL is calling innobase_kill_query, and is blocked by innodb global lock mutex. * Applier is in wsrep_innobase_kill_one_trx and is blocked by victim's LOCK_thd_kill. The fix in this commit removes the TOI replication of KILL command and makes KILL execution less intrusive operation. Aborting the victim happens now by using awake_no_mutex() and ha_abort_transaction(). If the KILL happens when the transaction is committing, the KILL operation is postponed to happen after the statement has completed in order to avoid KILL to interrupt commit processing. Notable changes in this commit: * wsrep client connections's error state may remain sticky after client connection is closed. This error message will then pop up for the next client session issuing first SQL statement. This problem raised with test galera.galera_bf_kill. The fix is to reset wsrep client error state, before a THD is reused for next connetion. * Release THD locks in wsrep_abort_transaction when locking innodb mutexes. This guarantees same locking order as with applier BF aborting. * BF abort from MDL was changed to do BF abort on server/wsrep-lib side first, and only then do the BF abort on InnoDB side. This removes the need to call back from InnoDB for BF aborts which originate from MDL and simplifies the locking. * Removed wsrep_thd_set_wsrep_aborter() from service_wsrep.h. The manipulation of the wsrep_aborter can be done solely on server side. Moreover, it is now debug only variable and could be excluded from optimized builds. * Remove LOCK_thd_kill from wsrep_thd_LOCK/UNLOCK to allow more fine grained locking for SR BF abort which may require locking of victim LOCK_thd_kill. Added explicit call for wsrep_thd_kill_LOCK/UNLOCK where appropriate. * Wsrep-lib was updated to version which allows external locking for BF abort calls. Changes to MTR tests: * Disable galera_bf_abort_group_commit. This test is going to be removed (MDEV-30855). * Make galera_var_retry_autocommit result more readable by echoing cases and expectations into result. Only one expected result for reap to verify that server returns expected status for query. * Record galera_gcache_recover_manytrx as result file was incomplete. Trivial change. * Make galera_create_table_as_select more deterministic: Wait until CTAS execution has reached MDL wait for multi-master conflict case. Expected error from multi-master conflict is ER_QUERY_INTERRUPTED. This is because CTAS does not yet have open wsrep transaction when it is waiting for MDL, query gets interrupted instead of BF aborted. This should be addressed in separate task. * A new test galera_bf_abort_registering to check that registering trx gets BF aborted through MDL. * A new test galera_kill_group_commit to verify correct behavior when KILL is executed while the transaction is committing. Co-authored-by: Seppo Jaakola <seppo.jaakola@iki.fi> Co-authored-by: Jan Lindström <jan.lindstrom@galeracluster.com> Signed-off-by: Julius Goryavsky <julius.goryavsky@mariadb.com>
2023-04-19 15:51:55 +02:00
mysql_mutex_unlock(&victim_thd->LOCK_thd_data);
if (bf_thd)
{
wsrep_store_threadvars(bf_thd);
}
2019-01-23 12:30:00 +01:00
}
extern "C" my_bool wsrep_thd_bf_abort(THD *bf_thd, THD *victim_thd,
2019-01-23 12:30:00 +01:00
my_bool signal)
{
mysql_mutex_assert_owner(&victim_thd->LOCK_thd_kill);
MDEV-29293 MariaDB stuck on starting commit state This commit contains a merge from 10.5-MDEV-29293-squash into 10.6. Although the bug MDEV-29293 was not reproducible with 10.6, the fix contains several improvements for wsrep KILL query and BF abort handling, and addresses the following issues: * MDEV-30307 KILL command issued inside a transaction is problematic for galera replication: This commit will remove KILL TOI replication, so Galera side transaction context is not lost during KILL. * MDEV-21075 KILL QUERY maintains nodes data consistency but breaks GTID sequence: This is fixed as well as KILL does not use TOI, and thus does not change GTID state. * MDEV-30372 Assertion in wsrep-lib state: This was caused by BF abort or KILL when local transaction was in the middle of group commit. This commit disables THD::killed handling during commit, so the problem is avoided. * MDEV-30963 Assertion failure !lock.was_chosen_as_deadlock_victim in trx0trx.h:1065: The assertion happened when the victim was BF aborted via MDL while it was committing. This commit changes MDL BF aborts so that transactions which are committing cannot be BF aborted via MDL. The RQG grammar attached in the issue could not reproduce the crash anymore. Original commit message from 10.5 fix: MDEV-29293 MariaDB stuck on starting commit state The problem seems to be a deadlock between KILL command execution and BF abort issued by an applier, where: * KILL has locked victim's LOCK_thd_kill and LOCK_thd_data. * Applier has innodb side global lock mutex and victim trx mutex. * KILL is calling innobase_kill_query, and is blocked by innodb global lock mutex. * Applier is in wsrep_innobase_kill_one_trx and is blocked by victim's LOCK_thd_kill. The fix in this commit removes the TOI replication of KILL command and makes KILL execution less intrusive operation. Aborting the victim happens now by using awake_no_mutex() and ha_abort_transaction(). If the KILL happens when the transaction is committing, the KILL operation is postponed to happen after the statement has completed in order to avoid KILL to interrupt commit processing. Notable changes in this commit: * wsrep client connections's error state may remain sticky after client connection is closed. This error message will then pop up for the next client session issuing first SQL statement. This problem raised with test galera.galera_bf_kill. The fix is to reset wsrep client error state, before a THD is reused for next connetion. * Release THD locks in wsrep_abort_transaction when locking innodb mutexes. This guarantees same locking order as with applier BF aborting. * BF abort from MDL was changed to do BF abort on server/wsrep-lib side first, and only then do the BF abort on InnoDB side. This removes the need to call back from InnoDB for BF aborts which originate from MDL and simplifies the locking. * Removed wsrep_thd_set_wsrep_aborter() from service_wsrep.h. The manipulation of the wsrep_aborter can be done solely on server side. Moreover, it is now debug only variable and could be excluded from optimized builds. * Remove LOCK_thd_kill from wsrep_thd_LOCK/UNLOCK to allow more fine grained locking for SR BF abort which may require locking of victim LOCK_thd_kill. Added explicit call for wsrep_thd_kill_LOCK/UNLOCK where appropriate. * Wsrep-lib was updated to version which allows external locking for BF abort calls. Changes to MTR tests: * Disable galera_bf_abort_group_commit. This test is going to be removed (MDEV-30855). * Make galera_var_retry_autocommit result more readable by echoing cases and expectations into result. Only one expected result for reap to verify that server returns expected status for query. * Record galera_gcache_recover_manytrx as result file was incomplete. Trivial change. * Make galera_create_table_as_select more deterministic: Wait until CTAS execution has reached MDL wait for multi-master conflict case. Expected error from multi-master conflict is ER_QUERY_INTERRUPTED. This is because CTAS does not yet have open wsrep transaction when it is waiting for MDL, query gets interrupted instead of BF aborted. This should be addressed in separate task. * A new test galera_bf_abort_registering to check that registering trx gets BF aborted through MDL. * A new test galera_kill_group_commit to verify correct behavior when KILL is executed while the transaction is committing. Co-authored-by: Seppo Jaakola <seppo.jaakola@iki.fi> Co-authored-by: Jan Lindström <jan.lindstrom@galeracluster.com> Signed-off-by: Julius Goryavsky <julius.goryavsky@mariadb.com>
2023-04-19 15:51:55 +02:00
mysql_mutex_assert_owner(&victim_thd->LOCK_thd_data);
2019-01-23 12:30:00 +01:00
my_bool ret= wsrep_bf_abort(bf_thd, victim_thd);
/*
Send awake signal if victim was BF aborted or does not
have wsrep on. Note that this should never interrupt RSU
as RSU has paused the provider.
*/
if ((ret || !wsrep_on(victim_thd)) && signal)
{
MDEV-21910 Deadlock between BF abort and manual KILL command 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
2020-05-19 10:12:26 +02:00
victim_thd->wsrep_aborter= bf_thd->thread_id;
victim_thd->awake_no_mutex(KILL_QUERY_HARD);
MDEV-21910 Deadlock between BF abort and manual KILL command 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
2020-05-19 10:12:26 +02:00
} else {
WSREP_DEBUG("wsrep_thd_bf_abort skipped awake, signal %d", signal);
}
2019-01-23 12:30:00 +01:00
return ret;
}
extern "C" my_bool wsrep_thd_skip_locking(const THD *thd)
{
return thd && thd->wsrep_skip_locking;
}
extern "C" my_bool wsrep_thd_order_before(const THD *left, const THD *right)
{
if (wsrep_thd_is_BF(left, false) &&
wsrep_thd_is_BF(right, false) &&
wsrep_thd_trx_seqno(left) < wsrep_thd_trx_seqno(right)) {
2019-01-23 12:30:00 +01:00
WSREP_DEBUG("BF conflict, order: %lld %lld\n",
(long long)wsrep_thd_trx_seqno(left),
(long long)wsrep_thd_trx_seqno(right));
return TRUE;
}
WSREP_DEBUG("waiting for BF, trx order: %lld %lld\n",
(long long)wsrep_thd_trx_seqno(left),
(long long)wsrep_thd_trx_seqno(right));
return FALSE;
}
extern "C" my_bool wsrep_thd_is_aborting(const MYSQL_THD thd)
{
mysql_mutex_assert_owner(&thd->LOCK_thd_data);
const wsrep::client_state& cs(thd->wsrep_cs());
const enum wsrep::transaction::state tx_state(cs.transaction().state());
switch (tx_state)
2019-01-23 12:30:00 +01:00
{
case wsrep::transaction::s_must_abort:
return (cs.state() == wsrep::client_state::s_exec ||
cs.state() == wsrep::client_state::s_result);
case wsrep::transaction::s_aborting:
return true;
default:
return false;
}
2019-01-23 12:30:00 +01:00
return false;
}
static inline enum wsrep::key::type
map_key_type(enum Wsrep_service_key_type type)
{
switch (type)
{
case WSREP_SERVICE_KEY_SHARED: return wsrep::key::shared;
case WSREP_SERVICE_KEY_REFERENCE: return wsrep::key::reference;
case WSREP_SERVICE_KEY_UPDATE: return wsrep::key::update;
case WSREP_SERVICE_KEY_EXCLUSIVE: return wsrep::key::exclusive;
}
return wsrep::key::exclusive;
}
extern "C" int wsrep_thd_append_key(THD *thd,
const struct wsrep_key* key,
int n_keys,
enum Wsrep_service_key_type key_type)
{
Wsrep_client_state& client_state(thd->wsrep_cs());
DBUG_ASSERT(client_state.transaction().active());
int ret= 0;
for (int i= 0; i < n_keys && ret == 0; ++i)
{
wsrep::key wsrep_key(map_key_type(key_type));
for (size_t kp= 0; kp < key[i].key_parts_num; ++kp)
{
wsrep_key.append_key_part(key[i].key_parts[kp].ptr, key[i].key_parts[kp].len);
}
ret= client_state.append_key(wsrep_key);
}
/*
In case of `wsrep_gtid_mode` when WS will be replicated, we need to set
`server_id` for events that are going to be written in IO, and in case of
manual SET gtid_seq_no=X we are ignoring value.
*/
if (!ret && wsrep_gtid_mode && !thd->slave_thread && !wsrep_thd_is_applying(thd))
{
thd->variables.server_id= wsrep_gtid_server.server_id;
thd->variables.gtid_seq_no= 0;
}
2019-01-23 12:30:00 +01:00
return ret;
}
10.4 wsrep group commit fixes (#1224) * MDEV-16509 Improve wsrep commit performance with binlog disabled Release commit order critical section early after trx_commit_low() if binlog is not transaction coordinator. In order to avoid two phase commit, binlog_hton is not registered for THD during IO_CACHE population. Implemented a test which verifies that the transactions release commit order early. This optimization will change behavior during recovery as the commit is not two phase when binlog is off. Fixed and recorded wsrep-recover-v25 and wsrep-recover to match the behavior. * MDEV-18730 Ordering for wsrep binlog group commit Previously out of order execution was allowed for wsrep commits. Established proper ordering by populating wait_for_commit for every wsrep THD and making group commit leader to wait for prior commits before proceeding to trx_group_commit_leader(). * MDEV-18730 Added a test case to verify correct commit ordering * MDEV-16509, MDEV-18730 Review fixes Use WSREP_EMULATE_BINLOG() macro to decide if the binlog_hton should be registered. Whitespace/syntax fixes and cleanups. * MDEV-16509 Require binlog for galera_var_innodb_disallow_writes test If the commit to InnoDB is done in one phase, the native InnoDB behavior is that the transaction is committed in memory before it is persisted to disk. This means that the innodb_disallow_writes=ON may not prevent transaction to become visible to other readers before commit is completely over. On the other hand, if the commit is two phase (as it is with binlog), the transaction will be blocked in prepare phase. Fixed the test to use binlog, which enforces two phase commit, which in turn makes commit to block before the changes become visible to other connections. This guarantees that the test produces expected result.
2019-03-15 06:09:13 +01:00
extern "C" void wsrep_commit_ordered(THD *thd)
{
if (wsrep_is_active(thd) &&
(thd->wsrep_trx().state() == wsrep::transaction::s_committing ||
thd->wsrep_trx().state() == wsrep::transaction::s_ordered_commit))
10.4 wsrep group commit fixes (#1224) * MDEV-16509 Improve wsrep commit performance with binlog disabled Release commit order critical section early after trx_commit_low() if binlog is not transaction coordinator. In order to avoid two phase commit, binlog_hton is not registered for THD during IO_CACHE population. Implemented a test which verifies that the transactions release commit order early. This optimization will change behavior during recovery as the commit is not two phase when binlog is off. Fixed and recorded wsrep-recover-v25 and wsrep-recover to match the behavior. * MDEV-18730 Ordering for wsrep binlog group commit Previously out of order execution was allowed for wsrep commits. Established proper ordering by populating wait_for_commit for every wsrep THD and making group commit leader to wait for prior commits before proceeding to trx_group_commit_leader(). * MDEV-18730 Added a test case to verify correct commit ordering * MDEV-16509, MDEV-18730 Review fixes Use WSREP_EMULATE_BINLOG() macro to decide if the binlog_hton should be registered. Whitespace/syntax fixes and cleanups. * MDEV-16509 Require binlog for galera_var_innodb_disallow_writes test If the commit to InnoDB is done in one phase, the native InnoDB behavior is that the transaction is committed in memory before it is persisted to disk. This means that the innodb_disallow_writes=ON may not prevent transaction to become visible to other readers before commit is completely over. On the other hand, if the commit is two phase (as it is with binlog), the transaction will be blocked in prepare phase. Fixed the test to use binlog, which enforces two phase commit, which in turn makes commit to block before the changes become visible to other connections. This guarantees that the test produces expected result.
2019-03-15 06:09:13 +01:00
{
wsrep_gtid_server.signal_waiters(thd->wsrep_current_gtid_seqno, false);
if (wsrep_thd_is_local(thd))
{
thd->wsrep_last_written_gtid_seqno= thd->wsrep_current_gtid_seqno;
}
if (thd->wsrep_trx().state() != wsrep::transaction::s_ordered_commit &&
!wsrep_commit_will_write_binlog(thd))
{
2020-05-31 09:28:59 +02:00
DEBUG_SYNC(thd, "before_wsrep_ordered_commit");
thd->wsrep_cs().ordered_commit();
}
10.4 wsrep group commit fixes (#1224) * MDEV-16509 Improve wsrep commit performance with binlog disabled Release commit order critical section early after trx_commit_low() if binlog is not transaction coordinator. In order to avoid two phase commit, binlog_hton is not registered for THD during IO_CACHE population. Implemented a test which verifies that the transactions release commit order early. This optimization will change behavior during recovery as the commit is not two phase when binlog is off. Fixed and recorded wsrep-recover-v25 and wsrep-recover to match the behavior. * MDEV-18730 Ordering for wsrep binlog group commit Previously out of order execution was allowed for wsrep commits. Established proper ordering by populating wait_for_commit for every wsrep THD and making group commit leader to wait for prior commits before proceeding to trx_group_commit_leader(). * MDEV-18730 Added a test case to verify correct commit ordering * MDEV-16509, MDEV-18730 Review fixes Use WSREP_EMULATE_BINLOG() macro to decide if the binlog_hton should be registered. Whitespace/syntax fixes and cleanups. * MDEV-16509 Require binlog for galera_var_innodb_disallow_writes test If the commit to InnoDB is done in one phase, the native InnoDB behavior is that the transaction is committed in memory before it is persisted to disk. This means that the innodb_disallow_writes=ON may not prevent transaction to become visible to other readers before commit is completely over. On the other hand, if the commit is two phase (as it is with binlog), the transaction will be blocked in prepare phase. Fixed the test to use binlog, which enforces two phase commit, which in turn makes commit to block before the changes become visible to other connections. This guarantees that the test produces expected result.
2019-03-15 06:09:13 +01:00
}
}
extern "C" my_bool wsrep_thd_has_ignored_error(const THD *thd)
{
return thd->wsrep_has_ignored_error;
}
extern "C" void wsrep_thd_set_ignored_error(THD *thd, my_bool val)
{
thd->wsrep_has_ignored_error= val;
}
extern "C" ulong wsrep_OSU_method_get(const MYSQL_THD thd)
{
if (thd)
return(thd->variables.wsrep_OSU_method);
else
return(global_system_variables.wsrep_OSU_method);
}
MDEV-21910 Deadlock between BF abort and manual KILL command 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
2020-05-19 10:12:26 +02:00
2020-09-21 11:29:00 +02:00
extern "C" void wsrep_report_bf_lock_wait(const THD *thd,
unsigned long long trx_id)
{
if (thd)
{
WSREP_ERROR("Thread %s trx_id: %llu thread: %ld "
"seqno: %lld client_state: %s client_mode: %s transaction_mode: %s "
"applier: %d toi: %d local: %d "
"query: %s",
wsrep_thd_is_BF(thd, false) ? "BF" : "normal",
trx_id,
thd_get_thread_id(thd),
wsrep_thd_trx_seqno(thd),
wsrep_thd_client_state_str(thd),
wsrep_thd_client_mode_str(thd),
wsrep_thd_transaction_state_str(thd),
wsrep_thd_is_applying(thd),
wsrep_thd_is_toi(thd),
wsrep_thd_is_local(thd),
wsrep_thd_query(thd));
}
}
extern "C" void wsrep_thd_set_PA_unsafe(THD *thd)
{
if (thd && thd->wsrep_cs().mark_transaction_pa_unsafe())
{
WSREP_DEBUG("session does not have active transaction, can not mark as PA unsafe");
}
}
extern "C" my_bool wsrep_thd_is_local_transaction(const THD *thd)
{
return (wsrep_thd_is_local(thd) &&
thd->wsrep_cs().transaction().active());
}