mariadb/sql/wsrep_thd.h

326 lines
10 KiB
C
Raw Normal View History

/* Copyright (C) 2013-2023 Codership Oy <info@codership.com>
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.,
2019-05-11 21:19:05 +02:00
51 Franklin Street, Fifth Floor, Boston, MA 02110-1335 USA. */
2014-09-25 23:00:45 +02:00
#ifndef WSREP_THD_H
#define WSREP_THD_H
2019-01-23 12:30:00 +01:00
#include <my_config.h>
2014-09-25 23:00:45 +02:00
2019-01-23 12:30:00 +01:00
#include "mysql/service_wsrep.h"
#include "wsrep/client_state.hpp"
#include "sql_class.h"
2019-01-23 12:30:00 +01:00
#include "wsrep_utils.h"
#include <deque>
class Wsrep_thd_queue
{
public:
Wsrep_thd_queue(THD* t) : thd(t)
{
mysql_mutex_init(key_LOCK_wsrep_thd_queue,
&LOCK_wsrep_thd_queue,
MY_MUTEX_INIT_FAST);
mysql_cond_init(key_COND_wsrep_thd_queue, &COND_wsrep_thd_queue, NULL);
}
~Wsrep_thd_queue()
{
mysql_mutex_destroy(&LOCK_wsrep_thd_queue);
mysql_cond_destroy(&COND_wsrep_thd_queue);
}
bool push_back(THD* thd)
{
DBUG_ASSERT(thd);
wsp::auto_lock lock(&LOCK_wsrep_thd_queue);
std::deque<THD*>::iterator it = queue.begin();
while (it != queue.end())
{
if (*it == thd)
{
return true;
}
it++;
}
queue.push_back(thd);
mysql_cond_signal(&COND_wsrep_thd_queue);
return false;
}
THD* pop_front()
{
wsp::auto_lock lock(&LOCK_wsrep_thd_queue);
while (queue.empty())
{
if (thd->killed != NOT_KILLED)
return NULL;
thd->mysys_var->current_mutex= &LOCK_wsrep_thd_queue;
thd->mysys_var->current_cond= &COND_wsrep_thd_queue;
mysql_cond_wait(&COND_wsrep_thd_queue, &LOCK_wsrep_thd_queue);
thd->mysys_var->current_mutex= 0;
thd->mysys_var->current_cond= 0;
}
THD* ret= queue.front();
queue.pop_front();
return ret;
}
private:
THD* thd;
std::deque<THD*> queue;
mysql_mutex_t LOCK_wsrep_thd_queue;
mysql_cond_t COND_wsrep_thd_queue;
};
int wsrep_show_bf_aborts (THD *thd, SHOW_VAR *var, char *buff,
enum enum_var_type scope);
bool wsrep_create_appliers(long threads, bool mutex_protected=false);
void wsrep_create_rollbacker();
bool wsrep_bf_abort(THD* bf_thd, THD* 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
/*
Abort transaction for victim_thd. This function is called from
MDL BF abort codepath.
*/
void wsrep_abort_thd(THD *bf_thd,
THD *victim_thd,
my_bool signal) __attribute__((nonnull(1,2)));
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
/**
Kill wsrep connection with kill_signal. Object thd is not
guaranteed to exist anymore when this function returns.
Asserts that the caller holds victim_thd->LOCK_thd_kill,
victim_thd->LOCK_thd_data.
@param thd THD object for connection that executes the KILL.
@param victim_thd THD object for connection to be killed.
@param kill_signal Kill signal.
@return Zero if the kill was successful, otherwise non-zero error code.
*/
uint wsrep_kill_thd(THD *thd, THD *victim_thd, killed_state kill_signal);
/*
Backup kill status for commit.
*/
void wsrep_backup_kill_for_commit(THD *);
/*
Restore KILL status after commit.
*/
void wsrep_restore_kill_after_commit(THD *);
/*
Helper methods to deal with thread local storage.
The purpose of these methods is to hide the details of thread
local storage handling when operating with wsrep storage access
and streaming applier THDs
With one-thread-per-connection thread handling thread specific
variables are allocated when the thread is started and deallocated
before thread exits (my_thread_init(), my_thread_end()). However,
with pool-of-threads thread handling new thread specific variables
are allocated for each THD separately (see threadpool_add_connection()),
and the variables in thread local storage are assigned from
currently active thread (see thread_attach()). This must be taken into
account when storing/resetting thread local storage and when creating
streaming applier THDs.
*/
/**
Create new variables for thread local storage. With
one-thread-per-connection thread handling this is a no op,
with pool-of-threads new variables are created via my_thread_init().
It is assumed that the caller has called wsrep_reset_threadvars() to clear
the thread local storage before this call.
@return Zero in case of success, non-zero otherwise.
*/
int wsrep_create_threadvars();
/**
Delete variables which were created by wsrep_create_threadvars().
The caller must store variables into thread local storage before
this call via wsrep_store_threadvars().
*/
void wsrep_delete_threadvars();
/**
Assign variables from current thread local storage into THD.
This should be called for THDs whose lifetime is limited to single
thread execution or which may share the operation context with some
parent THD (e.g. storage access) and thus don't require separately
allocated globals.
With one-thread-per-connection thread handling this is a no-op,
with pool-of-threads the variables which are currently stored into
thread local storage are assigned to THD.
*/
void wsrep_assign_from_threadvars(THD *);
/**
Helper struct to save variables from thread local storage.
*/
struct Wsrep_threadvars
{
THD* cur_thd;
st_my_thread_var* mysys_var;
};
/**
Save variables from thread local storage into Wsrep_threadvars struct.
*/
Wsrep_threadvars wsrep_save_threadvars();
/**
Restore variables into thread local storage from Wsrep_threadvars struct.
*/
void wsrep_restore_threadvars(const Wsrep_threadvars&);
/**
Store variables into thread local storage.
*/
void wsrep_store_threadvars(THD *);
/**
Reset thread local storage.
*/
void wsrep_reset_threadvars(THD *);
2019-01-23 12:30:00 +01:00
/**
Helper functions to override error status
In many contexts it is desirable to mask the original error status
set for THD or it is necessary to change OK status to error.
This function implements the common logic for the most
of the cases.
Rules:
* If the diagnostics are has OK or EOF status, override it unconditionally
* If the error is either ER_ERROR_DURING_COMMIT or ER_LOCK_DEADLOCK
it is usually the correct error status to be returned to client,
so don't override those by default
*/
static inline void wsrep_override_error(THD *thd, uint error, const char *format= 0, ...)
2019-01-23 12:30:00 +01:00
{
Diagnostics_area *da= thd->get_stmt_da();
if (da->is_ok() ||
da->is_eof() ||
!da->is_set() ||
(da->is_error() &&
da->sql_errno() != error &&
da->sql_errno() != ER_ERROR_DURING_COMMIT &&
da->sql_errno() != ER_LOCK_DEADLOCK))
{
da->reset_diagnostics_area();
va_list args;
va_start(args, format);
if (!format) format= ER_THD(thd, error);
my_printv_error(error, format, MYF(0), args);
va_end(args);
2019-01-23 12:30:00 +01:00
}
}
static inline void wsrep_override_error(THD* thd,
wsrep::client_error ce,
enum wsrep::provider::status status)
{
DBUG_ASSERT(ce != wsrep::e_success);
switch (ce)
{
case wsrep::e_error_during_commit:
if (status == wsrep::provider::error_size_exceeded)
wsrep_override_error(thd, ER_UNKNOWN_ERROR, "Maximum writeset size exceeded");
else
wsrep_override_error(thd, ER_ERROR_DURING_COMMIT, 0, status);
2019-01-23 12:30:00 +01:00
break;
case wsrep::e_deadlock_error:
wsrep_override_error(thd, ER_LOCK_DEADLOCK);
break;
case wsrep::e_interrupted_error:
wsrep_override_error(thd, ER_QUERY_INTERRUPTED);
break;
case wsrep::e_size_exceeded_error:
wsrep_override_error(thd, ER_UNKNOWN_ERROR, "Maximum writeset size exceeded");
2019-01-23 12:30:00 +01:00
break;
case wsrep::e_append_fragment_error:
/* TODO: Figure out better error number */
if (status)
wsrep_override_error(thd, ER_ERROR_DURING_COMMIT,
"Error while appending streaming replication fragment"
"(provider status: %s)",
wsrep::provider::to_string(status).c_str());
else
wsrep_override_error(thd, ER_ERROR_DURING_COMMIT,
"Error while appending streaming replication fragment");
2019-01-23 12:30:00 +01:00
break;
case wsrep::e_not_supported_error:
wsrep_override_error(thd, ER_NOT_SUPPORTED_YET);
break;
case wsrep::e_timeout_error:
wsrep_override_error(thd, ER_LOCK_WAIT_TIMEOUT);
break;
default:
wsrep_override_error(thd, ER_UNKNOWN_ERROR);
break;
}
}
2019-01-23 12:30:00 +01:00
/**
Helper function to log THD wsrep context.
2019-01-23 12:30:00 +01:00
@param thd Pointer to THD
@param message Optional message
@param function Function where the call was made from
*/
static inline void wsrep_log_thd(const THD *thd,
2019-01-23 12:30:00 +01:00
const char *message,
const char *function)
{
WSREP_DEBUG("%s %s\n"
" thd: %llu thd_ptr: %p client_mode: %s client_state: %s trx_state: %s\n"
" next_trx_id: %lld trx_id: %lld seqno: %lld\n"
" is_streaming: %d fragments: %zu\n"
" sql_errno: %u message: %s\n"
#define WSREP_THD_LOG_QUERIES
#ifdef WSREP_THD_LOG_QUERIES
" command: %d query: %.72s"
#endif /* WSREP_OBSERVER_LOG_QUERIES */
,
function,
message ? message : "",
thd->thread_id,
thd,
wsrep_thd_client_mode_str(thd),
wsrep_thd_client_state_str(thd),
wsrep_thd_transaction_state_str(thd),
(long long)thd->wsrep_next_trx_id(),
(long long)thd->wsrep_trx_id(),
(long long)wsrep_thd_trx_seqno(thd),
thd->wsrep_trx().is_streaming(),
thd->wsrep_sr().fragments().size(),
(thd->get_stmt_da()->is_error() ? thd->get_stmt_da()->sql_errno() : 0),
(thd->get_stmt_da()->is_error() ? thd->get_stmt_da()->message() : "")
#ifdef WSREP_THD_LOG_QUERIES
, thd->lex->sql_command,
wsrep_thd_query(thd)
2019-01-23 12:30:00 +01:00
#endif /* WSREP_OBSERVER_LOG_QUERIES */
);
}
2014-09-25 23:00:45 +02:00
2019-01-23 12:30:00 +01:00
#define WSREP_LOG_THD(thd_, message_) wsrep_log_thd(thd_, message_, __FUNCTION__)
2014-09-25 23:00:45 +02:00
#endif /* WSREP_THD_H */