mirror of
https://github.com/MariaDB/server.git
synced 2025-01-29 10:14:19 +01:00
3f59bbeeae
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). * 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_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>
405 lines
13 KiB
C++
405 lines
13 KiB
C++
/* Copyright 2018 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., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA */
|
|
|
|
#include "my_global.h"
|
|
#include "wsrep_server_service.h"
|
|
#include "wsrep_server_state.h"
|
|
#include "wsrep_client_state.h"
|
|
#include "wsrep_client_service.h"
|
|
#include "wsrep_storage_service.h"
|
|
#include "wsrep_high_priority_service.h"
|
|
|
|
#include "wsrep_sst.h"
|
|
#include "wsrep_xid.h"
|
|
#include "wsrep_mysqld.h"
|
|
#include "wsrep_schema.h"
|
|
#include "wsrep_utils.h"
|
|
#include "wsrep_thd.h"
|
|
|
|
#include "log.h" /* sql_print_xxx() */
|
|
#include "sql_class.h" /* system variables */
|
|
#include "transaction.h" /* trans_xxx */
|
|
#include "sql_base.h" /* close_thread_tables */
|
|
#include "debug_sync.h"
|
|
|
|
static void init_service_thd(THD* thd, char* thread_stack)
|
|
{
|
|
thd->thread_stack= thread_stack;
|
|
thd->real_id= pthread_self();
|
|
thd->prior_thr_create_utime= thd->start_utime= microsecond_interval_timer();
|
|
thd->set_command(COM_SLEEP);
|
|
thd->reset_for_next_command(true);
|
|
}
|
|
|
|
Wsrep_storage_service*
|
|
wsrep_create_storage_service(THD* orig_THD, const char* ctx)
|
|
{
|
|
THD* thd= new THD(true, true);
|
|
init_service_thd(thd, orig_THD->thread_stack);
|
|
WSREP_DEBUG("Created storage service in %s context with thread id %llu",
|
|
ctx, thd->thread_id);
|
|
/* Use variables from the current thd attached to client_service.
|
|
This is because we need to be able to BF abort storage access
|
|
operations. */
|
|
wsrep_assign_from_threadvars(thd);
|
|
return new Wsrep_storage_service(thd);
|
|
}
|
|
|
|
wsrep::storage_service* Wsrep_server_service::storage_service(
|
|
wsrep::client_service& client_service)
|
|
{
|
|
Wsrep_client_service& cs=
|
|
static_cast<Wsrep_client_service&>(client_service);
|
|
return wsrep_create_storage_service(cs.m_thd, "local");
|
|
}
|
|
|
|
wsrep::storage_service* Wsrep_server_service::storage_service(
|
|
wsrep::high_priority_service& high_priority_service)
|
|
{
|
|
Wsrep_high_priority_service& hps=
|
|
static_cast<Wsrep_high_priority_service&>(high_priority_service);
|
|
return wsrep_create_storage_service(hps.m_thd, "high priority");
|
|
}
|
|
|
|
void Wsrep_server_service::release_storage_service(
|
|
wsrep::storage_service* storage_service)
|
|
{
|
|
Wsrep_storage_service* ss=
|
|
static_cast<Wsrep_storage_service*>(storage_service);
|
|
THD* thd= ss->m_thd;
|
|
wsrep_reset_threadvars(thd);
|
|
delete ss;
|
|
delete thd;
|
|
}
|
|
|
|
Wsrep_applier_service*
|
|
wsrep_create_streaming_applier(THD *orig_thd, const char *ctx)
|
|
{
|
|
/* Reset variables to allow creating new variables in thread local
|
|
storage for new THD if needed. Note that reset must be done for
|
|
current_thd, as orig_thd may not be in effect. This may be the case when
|
|
streaming transaction is BF aborted and streaming applier
|
|
is created from BF aborter context. */
|
|
Wsrep_threadvars saved_threadvars(wsrep_save_threadvars());
|
|
wsrep_reset_threadvars(saved_threadvars.cur_thd);
|
|
THD *thd= 0;
|
|
Wsrep_applier_service *ret= 0;
|
|
if (!wsrep_create_threadvars() &&
|
|
(thd= new THD(next_thread_id(), true)))
|
|
{
|
|
init_service_thd(thd, orig_thd->thread_stack);
|
|
wsrep_assign_from_threadvars(thd);
|
|
WSREP_DEBUG("Created streaming applier service in %s context with "
|
|
"thread id %llu", ctx, thd->thread_id);
|
|
if (!(ret= new (std::nothrow) Wsrep_applier_service(thd)))
|
|
{
|
|
delete thd;
|
|
}
|
|
}
|
|
/* Restore original thread local storage state before returning. */
|
|
wsrep_restore_threadvars(saved_threadvars);
|
|
wsrep_store_threadvars(saved_threadvars.cur_thd);
|
|
return ret;
|
|
}
|
|
|
|
wsrep::high_priority_service*
|
|
Wsrep_server_service::streaming_applier_service(
|
|
wsrep::client_service& orig_client_service)
|
|
{
|
|
Wsrep_client_service& orig_cs=
|
|
static_cast<Wsrep_client_service&>(orig_client_service);
|
|
return wsrep_create_streaming_applier(orig_cs.m_thd, "local");
|
|
}
|
|
|
|
wsrep::high_priority_service*
|
|
Wsrep_server_service::streaming_applier_service(
|
|
wsrep::high_priority_service& orig_high_priority_service)
|
|
{
|
|
Wsrep_high_priority_service&
|
|
orig_hps(static_cast<Wsrep_high_priority_service&>(orig_high_priority_service));
|
|
return wsrep_create_streaming_applier(orig_hps.m_thd, "high priority");
|
|
}
|
|
|
|
void Wsrep_server_service::release_high_priority_service(wsrep::high_priority_service* high_priority_service)
|
|
{
|
|
Wsrep_high_priority_service* hps=
|
|
static_cast<Wsrep_high_priority_service*>(high_priority_service);
|
|
THD* thd= hps->m_thd;
|
|
delete hps;
|
|
wsrep_store_threadvars(thd);
|
|
delete thd;
|
|
wsrep_delete_threadvars();
|
|
}
|
|
|
|
void Wsrep_server_service::background_rollback(
|
|
wsrep::unique_lock<wsrep::mutex> &lock WSREP_UNUSED,
|
|
wsrep::client_state &client_state)
|
|
{
|
|
DBUG_ASSERT(lock.owns_lock());
|
|
Wsrep_client_state &cs= static_cast<Wsrep_client_state &>(client_state);
|
|
mysql_mutex_assert_owner(&cs.thd()->LOCK_thd_data);
|
|
wsrep_fire_rollbacker(cs.thd());
|
|
}
|
|
|
|
void Wsrep_server_service::bootstrap()
|
|
{
|
|
wsrep::log_info()
|
|
<< "Bootstrapping a new cluster, setting initial position to "
|
|
<< wsrep::gtid::undefined();
|
|
wsrep_set_SE_checkpoint(wsrep::gtid::undefined(), wsrep_gtid_server.undefined());
|
|
}
|
|
|
|
void Wsrep_server_service::log_message(enum wsrep::log::level level,
|
|
const char* message)
|
|
{
|
|
switch (level)
|
|
{
|
|
case wsrep::log::debug:
|
|
WSREP_DEBUG("%s", message);
|
|
break;
|
|
case wsrep::log::info:
|
|
WSREP_INFO("%s", message);
|
|
break;
|
|
case wsrep::log::warning:
|
|
WSREP_WARN("%s", message);
|
|
break;
|
|
case wsrep::log::error:
|
|
WSREP_ERROR("%s", message);
|
|
break;
|
|
case wsrep::log::unknown:
|
|
WSREP_UNKNOWN("%s", message);
|
|
break;
|
|
}
|
|
}
|
|
|
|
void Wsrep_server_service::log_view(
|
|
wsrep::high_priority_service* high_priority_service,
|
|
const wsrep::view& view)
|
|
{
|
|
Wsrep_high_priority_service* applier=
|
|
static_cast<Wsrep_high_priority_service*>(high_priority_service);
|
|
/* Update global system variables */
|
|
mysql_mutex_lock(&LOCK_global_system_variables);
|
|
if (wsrep_auto_increment_control && view.own_index() >= 0)
|
|
{
|
|
global_system_variables.auto_increment_offset= view.own_index() + 1;
|
|
global_system_variables.auto_increment_increment= view.members().size();
|
|
wsrep_protocol_version= view.protocol_version();
|
|
}
|
|
mysql_mutex_unlock(&LOCK_global_system_variables);
|
|
|
|
/* Update wsrep status variables */
|
|
mysql_mutex_lock(&LOCK_status);
|
|
wsrep_cluster_size= view.members().size();
|
|
wsrep_local_index= view.own_index();
|
|
std::ostringstream os;
|
|
os << view.state_id().id();
|
|
wsrep_update_cluster_state_uuid(os.str().c_str());
|
|
mysql_mutex_unlock(&LOCK_status);
|
|
wsrep_config_state->set(view);
|
|
wsrep_cluster_conf_id= view.view_seqno().get();
|
|
|
|
if (view.status() == wsrep::view::primary)
|
|
{
|
|
if (applier)
|
|
{
|
|
Wsrep_id id;
|
|
Wsrep_view prev_view= wsrep_schema->restore_view(applier->m_thd, id);
|
|
bool checkpoint_was_reset= false;
|
|
if (prev_view.state_id().id() != view.state_id().id())
|
|
{
|
|
WSREP_DEBUG("New cluster UUID was generated, resetting position info");
|
|
wsrep_set_SE_checkpoint(wsrep::gtid::undefined(), wsrep_gtid_server.undefined());
|
|
checkpoint_was_reset= true;
|
|
}
|
|
|
|
if (wsrep_debug)
|
|
{
|
|
std::ostringstream os;
|
|
os << "Storing cluster view:\n" << view;
|
|
WSREP_INFO("%s", os.str().c_str());
|
|
DBUG_ASSERT(prev_view.state_id().id() != view.state_id().id() ||
|
|
view.state_id().seqno().get() >= prev_view.state_id().seqno().get());
|
|
}
|
|
|
|
if (trans_begin(applier->m_thd, MYSQL_START_TRANS_OPT_READ_WRITE))
|
|
{
|
|
WSREP_WARN("Failed to start transaction for store view");
|
|
}
|
|
else
|
|
{
|
|
if (wsrep_schema->store_view(applier->m_thd, view))
|
|
{
|
|
WSREP_WARN("Failed to store view");
|
|
trans_rollback_stmt(applier->m_thd);
|
|
if (!trans_rollback(applier->m_thd))
|
|
{
|
|
close_thread_tables(applier->m_thd);
|
|
}
|
|
}
|
|
else
|
|
{
|
|
if (trans_commit(applier->m_thd))
|
|
{
|
|
WSREP_WARN("Failed to commit transaction for store view");
|
|
}
|
|
}
|
|
applier->m_thd->release_transactional_locks();
|
|
}
|
|
|
|
/*
|
|
Backwards compatibility: When running in mixed cluster with
|
|
Galera 3.x, the provider does not generate unique sequence numbers
|
|
for views. This condition can be checked by inspecting last
|
|
committed as returned by the provider. If the last_committed
|
|
matches to view state_id seqno, the cluster runs in backwards
|
|
compatibility mode and we skip setting the checkpoint for
|
|
view.
|
|
*/
|
|
wsrep::seqno last_committed=
|
|
Wsrep_server_state::instance().provider().last_committed_gtid().seqno();
|
|
if (checkpoint_was_reset || last_committed != view.state_id().seqno())
|
|
{
|
|
wsrep_set_SE_checkpoint(view.state_id(), wsrep_gtid_server.gtid());
|
|
}
|
|
DBUG_ASSERT(wsrep_get_SE_checkpoint<wsrep::gtid>().id() == view.state_id().id());
|
|
}
|
|
else
|
|
{
|
|
WSREP_DEBUG("No applier in Wsrep_server_service::log_view(), "
|
|
"skipping write to wsrep_schema");
|
|
}
|
|
}
|
|
}
|
|
|
|
void Wsrep_server_service::recover_streaming_appliers(wsrep::client_service& cs)
|
|
{
|
|
Wsrep_client_service& client_service= static_cast<Wsrep_client_service&>(cs);
|
|
wsrep_recover_sr_from_storage(client_service.m_thd);
|
|
}
|
|
|
|
void Wsrep_server_service::recover_streaming_appliers(
|
|
wsrep::high_priority_service& hs)
|
|
{
|
|
Wsrep_high_priority_service& high_priority_service=
|
|
static_cast<Wsrep_high_priority_service&>(hs);
|
|
wsrep_recover_sr_from_storage(high_priority_service.m_thd);
|
|
}
|
|
|
|
wsrep::view Wsrep_server_service::get_view(wsrep::client_service& c,
|
|
const wsrep::id& own_id)
|
|
{
|
|
Wsrep_client_service& cs(static_cast<Wsrep_client_service&>(c));
|
|
wsrep::view v(wsrep_schema->restore_view(cs.m_thd, own_id));
|
|
return v;
|
|
}
|
|
|
|
wsrep::gtid Wsrep_server_service::get_position(wsrep::client_service&)
|
|
{
|
|
return wsrep_get_SE_checkpoint<wsrep::gtid>();
|
|
}
|
|
|
|
void Wsrep_server_service::set_position(wsrep::client_service& c WSREP_UNUSED,
|
|
const wsrep::gtid& gtid)
|
|
{
|
|
Wsrep_client_service& cs WSREP_UNUSED (static_cast<Wsrep_client_service&>(c));
|
|
DBUG_ASSERT(cs.m_client_state.transaction().state()
|
|
== wsrep::transaction::s_aborted);
|
|
// Wait until all prior committers have finished.
|
|
wsrep::gtid wait_for(gtid.id(),
|
|
wsrep::seqno(gtid.seqno().get() - 1));
|
|
if (auto err = Wsrep_server_state::instance().provider()
|
|
.wait_for_gtid(wait_for, std::numeric_limits<int>::max()))
|
|
{
|
|
WSREP_WARN("Wait for gtid returned error %d while waiting for "
|
|
"prior transactions to commit before setting position", err);
|
|
}
|
|
wsrep_set_SE_checkpoint(gtid, wsrep_gtid_server.gtid());
|
|
}
|
|
|
|
void Wsrep_server_service::log_state_change(
|
|
enum Wsrep_server_state::state prev_state,
|
|
enum Wsrep_server_state::state current_state)
|
|
{
|
|
WSREP_INFO("Server status change %s -> %s",
|
|
wsrep::to_c_string(prev_state),
|
|
wsrep::to_c_string(current_state));
|
|
mysql_mutex_lock(&LOCK_status);
|
|
switch (current_state)
|
|
{
|
|
case Wsrep_server_state::s_synced:
|
|
wsrep_ready= TRUE;
|
|
WSREP_INFO("Synchronized with group, ready for connections");
|
|
/* fall through */
|
|
case Wsrep_server_state::s_joined:
|
|
case Wsrep_server_state::s_donor:
|
|
wsrep_cluster_status= "Primary";
|
|
break;
|
|
case Wsrep_server_state::s_connected:
|
|
wsrep_cluster_status= "non-Primary";
|
|
wsrep_ready= FALSE;
|
|
wsrep_connected= TRUE;
|
|
break;
|
|
case Wsrep_server_state::s_disconnected:
|
|
wsrep_ready= FALSE;
|
|
wsrep_connected= FALSE;
|
|
wsrep_cluster_status= "Disconnected";
|
|
break;
|
|
default:
|
|
wsrep_ready= FALSE;
|
|
wsrep_cluster_status= "non-Primary";
|
|
break;
|
|
}
|
|
mysql_mutex_unlock(&LOCK_status);
|
|
wsrep_config_state->set(current_state);
|
|
}
|
|
|
|
bool Wsrep_server_service::sst_before_init() const
|
|
{
|
|
return wsrep_before_SE();
|
|
}
|
|
|
|
std::string Wsrep_server_service::sst_request()
|
|
{
|
|
return wsrep_sst_prepare();
|
|
}
|
|
|
|
int Wsrep_server_service::start_sst(const std::string& sst_request,
|
|
const wsrep::gtid& gtid,
|
|
bool bypass)
|
|
{
|
|
return wsrep_sst_donate(sst_request, gtid, bypass);
|
|
}
|
|
|
|
int Wsrep_server_service::wait_committing_transactions(int timeout)
|
|
{
|
|
return wsrep_wait_committing_connections_close(timeout);
|
|
}
|
|
|
|
void Wsrep_server_service::debug_sync(const char* sync_point)
|
|
{
|
|
#ifdef ENABLED_DEBUG_SYNC
|
|
DBUG_EXECUTE_IF(sync_point, {
|
|
std::stringstream dbug_action;
|
|
dbug_action << "now "
|
|
<< "SIGNAL " << sync_point << "_reached "
|
|
<< "WAIT_FOR " << sync_point << "_continue";
|
|
const std::string& action(dbug_action.str());
|
|
DBUG_ASSERT(!debug_sync_set_action(current_thd,
|
|
action.c_str(),
|
|
action.length()));
|
|
};);
|
|
#endif
|
|
}
|