mariadb/sql/wsrep_server_service.cc

413 lines
13 KiB
C++
Raw Normal View History

2019-01-23 15:30:00 +04:00
/* 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"
2019-01-23 15:30:00 +04:00
#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"
2019-01-23 15:30:00 +04:00
MDEV-34533 asan error about stack overflow when writing record in Aria The problem was that when using clang + asan, we do not get a correct value for the thread stack as some local variables are not allocated at the normal stack. It looks like that for example clang 18.1.3, when compiling with -O2 -fsanitize=addressan it puts local variables and things allocated by alloca() in other areas than on the stack. The following code shows the issue Thread 6 "mariadbd" hit Breakpoint 3, do_handle_one_connection (connect=0x5080000027b8, put_in_cache=<optimized out>) at sql/sql_connect.cc:1399 THD *thd; 1399 thd->thread_stack= (char*) &thd; (gdb) p &thd (THD **) 0x7fffedee7060 (gdb) p $sp (void *) 0x7fffef4e7bc0 The address of thd is 24M away from the stack pointer (gdb) info reg ... rsp 0x7fffef4e7bc0 0x7fffef4e7bc0 ... r13 0x7fffedee7060 140737185214560 r13 is pointing to the address of the thd. Probably some kind of "local stack" used by the sanitizer I have verified this with gdb on a recursive call that calls alloca() in a loop. In this case all objects was stored in a local heap, not on the stack. To solve this issue in a portable way, I have added two functions: my_get_stack_pointer() returns the address of the current stack pointer. The code is using asm instructions for intel 32/64 bit, powerpc, arm 32/64 bit and sparc 32/64 bit. Supported compilers are gcc, clang and MSVC. For MSVC 64 bit we are using _AddressOfReturnAddress() As a fallback for other compilers/arch we use the address of a local variable. my_get_stack_bounds() that will return the address of the base stack and stack size using pthread_attr_getstack() or NtCurrentTed() with fallback to using the address of a local variable and user provided stack size. Server changes are: - Moving setting of thread_stack to THD::store_globals() using my_get_stack_bounds(). - Removing setting of thd->thread_stack, except in functions that allocates a lot on the stack before calling store_globals(). When using estimates for stack start, we reduce stack_size with MY_STACK_SAFE_MARGIN (8192) to take into account the stack used before calling store_globals(). I also added a unittest, stack_allocation-t, to verify the new code. Reviewed-by: Sergei Golubchik <serg@mariadb.org>
2024-10-01 17:07:48 +03:00
static void init_service_thd(THD* thd, void* thread_stack)
2019-01-23 15:30:00 +04:00
{
thd->thread_stack= thread_stack;
thd->real_id= pthread_self();
thd->prior_thr_create_utime= thd->start_utime= microsecond_interval_timer();
thd->mark_connection_idle();
2019-01-23 15:30:00 +04:00
thd->reset_for_next_command(true);
}
Wsrep_storage_service*
wsrep_create_storage_service(THD* orig_THD, const char* ctx)
2019-01-23 15:30:00 +04:00
{
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);
2019-01-23 15:30:00 +04:00
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");
}
2019-01-23 15:30:00 +04:00
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");
2019-01-23 15:30:00 +04:00
}
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);
2019-01-23 15:30:00 +04:00
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;
}
2019-01-23 15:30:00 +04:00
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");
2019-01-23 15:30:00 +04:00
}
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");
2019-01-23 15:30:00 +04:00
}
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);
2019-01-23 15:30:00 +04:00
delete thd;
wsrep_delete_threadvars();
2019-01-23 15:30:00 +04:00
}
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). * 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>
2023-04-17 16:04:01 +03:00
void Wsrep_server_service::background_rollback(
wsrep::unique_lock<wsrep::mutex> &lock WSREP_UNUSED,
wsrep::client_state &client_state)
2019-01-23 15:30:00 +04:00
{
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). * 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>
2023-04-17 16:04:01 +03:00
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);
2019-01-23 15:30:00 +04:00
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());
2019-01-23 15:30:00 +04:00
}
static std::atomic<bool> suppress_logging{false};
void wsrep_suppress_error_logging() { suppress_logging= true; }
2019-01-23 15:30:00 +04:00
void Wsrep_server_service::log_message(enum wsrep::log::level level,
const char *message)
2019-01-23 15:30:00 +04:00
{
if (suppress_logging.load(std::memory_order_relaxed))
{
return;
}
2019-01-23 15:30:00 +04:00
switch (level)
{
case wsrep::log::debug:
WSREP_DEBUG("%s", message);
2019-01-23 15:30:00 +04:00
break;
case wsrep::log::info:
WSREP_INFO("%s", message);
2019-01-23 15:30:00 +04:00
break;
case wsrep::log::warning:
WSREP_WARN("%s", message);
2019-01-23 15:30:00 +04:00
break;
case wsrep::log::error:
WSREP_ERROR("%s", message);
break;
case wsrep::log::unknown:
WSREP_UNKNOWN("%s", message);
2019-01-23 15:30:00 +04:00
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();
2019-01-23 15:30:00 +04:00
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;
2019-01-23 15:30:00 +04:00
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;
2019-01-23 15:30:00 +04:00
}
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());
2019-01-23 15:30:00 +04:00
}
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();
2019-01-23 15:30:00 +04:00
}
/*
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());
2019-01-23 15:30:00 +04:00
}
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>();
2019-01-23 15:30:00 +04:00
}
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());
}
2019-01-23 15:30:00 +04:00
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_INFO("Synchronized with group, ready for connections");
wsrep_ready_set(true);
2019-01-23 15:30:00 +04:00
/* 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_set(false);
2019-01-23 15:30:00 +04:00
wsrep_connected= TRUE;
break;
case Wsrep_server_state::s_disconnected:
wsrep_ready_set(false);
2019-01-23 15:30:00 +04:00
wsrep_connected= FALSE;
wsrep_cluster_status= "Disconnected";
break;
default:
wsrep_ready_set(false);
2019-01-23 15:30:00 +04:00
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)
2019-01-23 15:30:00 +04:00
{
#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
2019-01-23 15:30:00 +04:00
}