dbs/mysql-server
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
mysql-server/sql/rpl_channel_service_interface.cc
Ho Sy Tan 4a107fd10e 2025-03-05 14:31:08
2025-03-05 14:31:37 +07:00

1529 lines
44 KiB
C++
Raw Permalink Blame History

/* Copyright (c) 2015, 2024, Oracle and/or its affiliates.
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License, version 2.0,
as published by the Free Software Foundation.
This program is designed to work with certain software (including
but not limited to OpenSSL) that is licensed under separate terms,
as designated in a particular file or component or in included license
documentation. The authors of MySQL hereby grant you an additional
permission to link the program and your derivative works with the
separately licensed software that they have either included with
the program or referenced in the documentation.
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, version 2.0, 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 "sql/rpl_channel_service_interface.h"
#include <stdio.h>
#include <string.h>
#include <sys/types.h>
#include <atomic>
#include <map>
#include <sstream>
#include <utility>
#include "mutex_lock.h" // MUTEX_LOCK
#include "my_compiler.h"
#include "my_dbug.h"
#include "my_inttypes.h"
#include "my_sys.h"
#include "my_thread.h"
#include "mysql/components/services/bits/psi_bits.h"
#include "mysql/components/services/bits/psi_stage_bits.h"
#include "mysql/components/services/log_builtins.h"
#include "mysql/my_loglevel.h"
#include "mysql/psi/mysql_cond.h"
#include "mysql/psi/mysql_mutex.h"
#include "mysql/service_mysql_alloc.h"
#include "mysql_com.h"
#include "mysqld_error.h"
#include "sql/auth/sql_security_ctx.h"
#include "sql/binlog.h"
#include "sql/changestreams/apply/replication_thread_status.h"
#include "sql/current_thd.h"
#include "sql/log.h"
#include "sql/log_event.h"
#include "sql/mysqld.h" // opt_mts_replica_parallel_workers
#include "sql/mysqld_thd_manager.h" // Global_THD_manager
#include "sql/protocol_classic.h"
#include "sql/raii/sentry.h"
#include "sql/rpl_async_conn_failover_configuration_propagation.h"
#include "sql/rpl_channel_credentials.h"
#include "sql/rpl_gtid.h"
#include "sql/rpl_info_factory.h"
#include "sql/rpl_info_handler.h"
#include "sql/rpl_mi.h"
#include "sql/rpl_msr.h" /* Multisource replication */
#include "sql/rpl_mta_submode.h"
#include "sql/rpl_replica.h"
#include "sql/rpl_rli.h"
#include "sql/rpl_rli_pdb.h"
#include "sql/rpl_trx_boundary_parser.h"
#include "sql/sql_class.h"
#include "sql/sql_lex.h"
#include "sql/transaction.h" // trans_begin
/**
Auxiliary function to stop all the running channel threads according to the
given mask.
@note: The caller shall possess channel_map lock before calling this function,
and unlock after returning from this function.
@param mi The pointer to Master_info instance
@param threads_to_stop The types of threads to be stopped
@param timeout The expected time in which the thread should stop
@return the operation status
@retval 0 OK
@retval !=0 Error
*/
int channel_stop(Master_info *mi, int threads_to_stop, long timeout);
int initialize_channel_service_interface() {
DBUG_TRACE;
// server id must be different from 0
if (server_id == 0) {
LogErr(ERROR_LEVEL, ER_RPL_CHANNELS_REQUIRE_NON_ZERO_SERVER_ID);
return 1;
}
return 0;
}
static void set_mi_settings(Master_info *mi,
Channel_creation_info *channel_info) {
mysql_mutex_lock(mi->rli->relay_log.get_log_lock());
mysql_mutex_lock(&mi->data_lock);
mi->rli->set_thd_tx_priority(channel_info->thd_tx_priority);
mi->rli->set_ignore_write_set_memory_limit(
channel_info->m_ignore_write_set_memory_limit);
mi->rli->set_allow_drop_write_set(channel_info->m_allow_drop_write_set);
mi->rli->replicate_same_server_id =
(channel_info->replicate_same_server_id == RPL_SERVICE_SERVER_DEFAULT)
? replicate_same_server_id
: channel_info->replicate_same_server_id;
mi->rli->opt_replica_parallel_workers =
(channel_info->channel_mts_parallel_workers == RPL_SERVICE_SERVER_DEFAULT)
? opt_mts_replica_parallel_workers
: channel_info->channel_mts_parallel_workers;
if (channel_info->channel_mts_parallel_type == RPL_SERVICE_SERVER_DEFAULT) {
if (mts_parallel_option == MTS_PARALLEL_TYPE_DB_NAME)
mi->rli->channel_mts_submode = MTS_PARALLEL_TYPE_DB_NAME;
else
mi->rli->channel_mts_submode = MTS_PARALLEL_TYPE_LOGICAL_CLOCK;
} else {
if (channel_info->channel_mts_parallel_type ==
CHANNEL_MTS_PARALLEL_TYPE_DB_NAME)
mi->rli->channel_mts_submode = MTS_PARALLEL_TYPE_DB_NAME;
else
mi->rli->channel_mts_submode = MTS_PARALLEL_TYPE_LOGICAL_CLOCK;
}
mi->rli->checkpoint_group =
(channel_info->channel_mta_checkpoint_group == RPL_SERVICE_SERVER_DEFAULT)
? opt_mta_checkpoint_group
: channel_info->channel_mta_checkpoint_group;
Format_description_log_event *fde = new Format_description_log_event();
/*
Group replication applier channel shall not use checksum on its relay log
files.
*/
if (channel_map.is_group_replication_applier_channel_name(
mi->get_channel())) {
fde->footer()->checksum_alg = mysql::binlog::event::BINLOG_CHECKSUM_ALG_OFF;
/*
When the receiver thread connects to the master, it gets its current
binlog checksum algorithm, but as GR applier channel has no receiver
thread (and also does not connect to a master), we will set the variable
here to BINLOG_CHECKSUM_ALG_OFF as events queued after certification have
no checksum information.
*/
mi->checksum_alg_before_fd = mysql::binlog::event::BINLOG_CHECKSUM_ALG_OFF;
}
mi->set_mi_description_event(fde);
mysql_mutex_unlock(&mi->data_lock);
mysql_mutex_unlock(mi->rli->relay_log.get_log_lock());
}
static bool init_thread_context() { return my_thread_init(); }
static void clean_thread_context() { my_thread_end(); }
static THD *create_surrogate_thread() {
THD *thd = nullptr;
thd = new THD;
thd->thread_stack = (char *)&thd;
thd->store_globals();
thd->security_context()->skip_grants();
return (thd);
}
static void delete_surrogate_thread(THD *thd) {
thd->release_resources();
delete thd;
current_thd = nullptr;
}
void initialize_channel_creation_info(Channel_creation_info *channel_info) {
channel_info->type = SLAVE_REPLICATION_CHANNEL;
channel_info->hostname = nullptr;
channel_info->port = 0;
channel_info->user = nullptr;
channel_info->password = nullptr;
channel_info->ssl_info = nullptr;
channel_info->auto_position = RPL_SERVICE_SERVER_DEFAULT;
channel_info->channel_mts_parallel_type = RPL_SERVICE_SERVER_DEFAULT;
channel_info->channel_mts_parallel_workers = RPL_SERVICE_SERVER_DEFAULT;
channel_info->channel_mta_checkpoint_group = RPL_SERVICE_SERVER_DEFAULT;
channel_info->replicate_same_server_id = RPL_SERVICE_SERVER_DEFAULT;
channel_info->thd_tx_priority = 0;
channel_info->sql_delay = RPL_SERVICE_SERVER_DEFAULT;
channel_info->preserve_relay_logs = false;
channel_info->retry_count = 0;
channel_info->connect_retry = 0;
channel_info->public_key_path = nullptr;
channel_info->get_public_key = 0;
channel_info->compression_algorithm = nullptr;
channel_info->zstd_compression_level = 0;
channel_info->m_ignore_write_set_memory_limit = false;
channel_info->m_allow_drop_write_set = false;
}
void initialize_channel_ssl_info(Channel_ssl_info *channel_ssl_info) {
channel_ssl_info->use_ssl = 0;
channel_ssl_info->ssl_ca_file_name = nullptr;
channel_ssl_info->ssl_ca_directory = nullptr;
channel_ssl_info->ssl_cert_file_name = nullptr;
channel_ssl_info->ssl_crl_file_name = nullptr;
channel_ssl_info->ssl_crl_directory = nullptr;
channel_ssl_info->ssl_key = nullptr;
channel_ssl_info->ssl_cipher = nullptr;
channel_ssl_info->tls_version = nullptr;
channel_ssl_info->ssl_verify_server_cert = 0;
channel_ssl_info->tls_ciphersuites = nullptr;
}
void initialize_channel_connection_info(Channel_connection_info *channel_info) {
channel_info->until_condition = CHANNEL_NO_UNTIL_CONDITION;
channel_info->gtid = nullptr;
channel_info->view_id = nullptr;
}
static void set_mi_ssl_options(LEX_SOURCE_INFO *lex_mi,
Channel_ssl_info *channel_ssl_info) {
lex_mi->ssl = (channel_ssl_info->use_ssl) ? LEX_SOURCE_INFO::LEX_MI_ENABLE
: LEX_SOURCE_INFO::LEX_MI_DISABLE;
if (channel_ssl_info->ssl_ca_file_name != nullptr) {
lex_mi->ssl_ca = channel_ssl_info->ssl_ca_file_name;
}
if (channel_ssl_info->ssl_ca_directory != nullptr) {
lex_mi->ssl_capath = channel_ssl_info->ssl_ca_directory;
}
if (channel_ssl_info->tls_version != nullptr) {
lex_mi->tls_version = channel_ssl_info->tls_version;
}
if (channel_ssl_info->ssl_cert_file_name != nullptr) {
lex_mi->ssl_cert = channel_ssl_info->ssl_cert_file_name;
}
if (channel_ssl_info->ssl_crl_file_name != nullptr) {
lex_mi->ssl_crl = channel_ssl_info->ssl_crl_file_name;
}
if (channel_ssl_info->ssl_crl_directory != nullptr) {
lex_mi->ssl_crlpath = channel_ssl_info->ssl_crl_directory;
}
if (channel_ssl_info->ssl_key != nullptr) {
lex_mi->ssl_key = channel_ssl_info->ssl_key;
}
if (channel_ssl_info->ssl_cipher != nullptr) {
lex_mi->ssl_cipher = channel_ssl_info->ssl_cipher;
}
if (channel_ssl_info->tls_ciphersuites != nullptr) {
lex_mi->tls_ciphersuites = LEX_SOURCE_INFO::SPECIFIED_STRING;
lex_mi->tls_ciphersuites_string = channel_ssl_info->tls_ciphersuites;
} else {
lex_mi->tls_ciphersuites = LEX_SOURCE_INFO::SPECIFIED_NULL;
}
lex_mi->ssl_verify_server_cert = (channel_ssl_info->ssl_verify_server_cert)
? LEX_SOURCE_INFO::LEX_MI_ENABLE
: LEX_SOURCE_INFO::LEX_MI_DISABLE;
}
int channel_create(const char *channel, Channel_creation_info *channel_info) {
DBUG_TRACE;
Master_info *mi = nullptr;
int error = 0;
LEX_SOURCE_INFO *lex_mi = nullptr;
bool thd_created = false;
THD *thd = current_thd;
// Don't create default channels
if (!strcmp(channel_map.get_default_channel(), channel))
return RPL_CHANNEL_SERVICE_DEFAULT_CHANNEL_CREATION_ERROR;
/* Service channels are not supposed to use sql_replica_skip_counter */
mysql_mutex_lock(&LOCK_sql_replica_skip_counter);
if (sql_replica_skip_counter > 0)
error = RPL_CHANNEL_SERVICE_SLAVE_SKIP_COUNTER_ACTIVE;
mysql_mutex_unlock(&LOCK_sql_replica_skip_counter);
if (error) return error;
channel_map.wrlock();
/* Get the Master_info of the channel */
mi = channel_map.get_mi(channel);
/* create a new channel if doesn't exist */
if (!mi) {
if ((error = add_new_channel(&mi, channel))) goto err;
}
lex_mi = new LEX_SOURCE_INFO();
lex_mi->channel = channel;
lex_mi->host = channel_info->hostname;
/*
'group_replication_recovery' channel (*after recovery is done*)
or 'group_replication_applier' channel wants to set the port number
to '0' as there is no actual network usage on these channels.
*/
lex_mi->port_opt = LEX_SOURCE_INFO::LEX_MI_ENABLE;
lex_mi->port = channel_info->port;
lex_mi->user = channel_info->user;
lex_mi->password = channel_info->password;
lex_mi->sql_delay = channel_info->sql_delay;
lex_mi->connect_retry = channel_info->connect_retry;
if (channel_info->retry_count) {
lex_mi->retry_count_opt = LEX_SOURCE_INFO::LEX_MI_ENABLE;
lex_mi->retry_count = channel_info->retry_count;
}
if (channel_info->auto_position) {
lex_mi->auto_position = LEX_SOURCE_INFO::LEX_MI_ENABLE;
if ((mi && mi->is_auto_position()) ||
channel_info->auto_position == RPL_SERVICE_SERVER_DEFAULT) {
// So change replication source allows new configurations with a running
// SQL thread
lex_mi->auto_position = LEX_SOURCE_INFO::LEX_MI_UNCHANGED;
}
}
if (channel_info->public_key_path) {
lex_mi->public_key_path = channel_info->public_key_path;
}
if (channel_info->get_public_key) {
lex_mi->get_public_key = LEX_SOURCE_INFO::LEX_MI_ENABLE;
if (mi && mi->get_public_key) {
// So change replication source allows new configurations with a running
// SQL thread
lex_mi->get_public_key = LEX_SOURCE_INFO::LEX_MI_UNCHANGED;
}
} else {
lex_mi->get_public_key = LEX_SOURCE_INFO::LEX_MI_DISABLE;
if (mi && !mi->get_public_key) {
// So change replication source allows new configurations with a running
// SQL thread
lex_mi->get_public_key = LEX_SOURCE_INFO::LEX_MI_UNCHANGED;
}
}
if (channel_info->compression_algorithm != nullptr) {
lex_mi->compression_algorithm = channel_info->compression_algorithm;
}
if (channel_info->zstd_compression_level) {
lex_mi->zstd_compression_level = channel_info->zstd_compression_level;
}
lex_mi->m_source_connection_auto_failover = LEX_SOURCE_INFO::LEX_MI_UNCHANGED;
if (channel_info->m_source_connection_auto_failover) {
if (mi && !mi->is_source_connection_auto_failover()) {
lex_mi->m_source_connection_auto_failover =
LEX_SOURCE_INFO::LEX_MI_ENABLE;
}
} else {
if (mi && mi->is_source_connection_auto_failover()) {
lex_mi->m_source_connection_auto_failover =
LEX_SOURCE_INFO::LEX_MI_DISABLE;
}
}
if (channel_info->ssl_info != nullptr) {
set_mi_ssl_options(lex_mi, channel_info->ssl_info);
}
if (mi) {
if (!thd) {
thd_created = true;
thd = create_surrogate_thread();
}
if ((error = change_master(thd, mi, lex_mi,
channel_info->preserve_relay_logs))) {
goto err;
}
}
set_mi_settings(mi, channel_info);
if (channel_map.is_group_replication_channel_name(mi->get_channel())) {
thd->variables.max_allowed_packet = replica_max_allowed_packet;
thd->get_protocol_classic()->set_max_packet_size(
replica_max_allowed_packet + MAX_LOG_EVENT_HEADER);
}
err:
channel_map.unlock();
if (thd_created) {
delete_surrogate_thread(thd);
}
delete lex_mi;
return error;
}
int channel_start(const char *channel, Channel_connection_info *connection_info,
int threads_to_start, int wait_for_connection,
bool use_server_mta_configuration,
bool channel_map_already_locked) {
DBUG_TRACE;
int error = 0;
int thread_mask = 0;
LEX_SOURCE_INFO lex_mi;
ulong thread_start_id = 0;
bool thd_created = false;
THD *thd = current_thd;
String_set user, pass, auth;
/* Service channels are not supposed to use sql_replica_skip_counter */
mysql_mutex_lock(&LOCK_sql_replica_skip_counter);
if (sql_replica_skip_counter > 0)
error = RPL_CHANNEL_SERVICE_SLAVE_SKIP_COUNTER_ACTIVE;
mysql_mutex_unlock(&LOCK_sql_replica_skip_counter);
if (error) return error;
if (channel_map_already_locked) {
channel_map.assert_some_wrlock();
} else {
channel_map.wrlock();
}
Master_info *mi = channel_map.get_mi(channel);
if (mi == nullptr) {
error = RPL_CHANNEL_SERVICE_CHANNEL_DOES_NOT_EXISTS_ERROR;
goto err;
}
if (threads_to_start & CHANNEL_APPLIER_THREAD) {
thread_mask |= REPLICA_SQL;
}
if (threads_to_start & CHANNEL_RECEIVER_THREAD) {
thread_mask |= REPLICA_IO;
}
// Nothing to be done here
if (!thread_mask) goto err;
LEX_REPLICA_CONNECTION lex_connection;
lex_connection.reset();
if (!Rpl_channel_credentials::get_instance().get_credentials(channel, user,
pass, auth)) {
lex_connection.user =
(user.first ? const_cast<char *>(user.second.c_str()) : nullptr);
lex_connection.password =
(pass.first ? const_cast<char *>(pass.second.c_str()) : nullptr);
lex_connection.plugin_auth =
(auth.first ? const_cast<char *>(auth.second.c_str()) : nullptr);
}
if (connection_info->until_condition != CHANNEL_NO_UNTIL_CONDITION) {
switch (connection_info->until_condition) {
case CHANNEL_UNTIL_APPLIER_AFTER_GTIDS:
lex_mi.gtid_until_condition = LEX_SOURCE_INFO::UNTIL_SQL_AFTER_GTIDS;
lex_mi.gtid = connection_info->gtid;
break;
case CHANNEL_UNTIL_APPLIER_BEFORE_GTIDS:
lex_mi.gtid_until_condition = LEX_SOURCE_INFO::UNTIL_SQL_BEFORE_GTIDS;
lex_mi.gtid = connection_info->gtid;
break;
case CHANNEL_UNTIL_APPLIER_AFTER_GAPS:
lex_mi.until_after_gaps = true;
break;
case CHANNEL_UNTIL_VIEW_ID:
assert((thread_mask & REPLICA_SQL) && connection_info->view_id);
lex_mi.view_id = connection_info->view_id;
break;
default:
assert(0);
}
}
if (wait_for_connection && (thread_mask & REPLICA_IO))
thread_start_id = mi->slave_run_id;
if (!thd) {
thd_created = true;
thd = create_surrogate_thread();
}
error = start_slave(thd, &lex_connection, &lex_mi, thread_mask, mi,
use_server_mta_configuration);
if (wait_for_connection && (thread_mask & REPLICA_IO) && !error) {
mysql_mutex_lock(&mi->run_lock);
/*
If the ids are still equal this means the start thread method did not
wait for the thread to start
*/
while (thread_start_id == mi->slave_run_id) {
mysql_cond_wait(&mi->start_cond, &mi->run_lock);
}
mysql_mutex_unlock(&mi->run_lock);
while (mi->slave_running != MYSQL_SLAVE_RUN_CONNECT) {
// If there is such a state change then there was an error on connection
if (mi->slave_running == MYSQL_SLAVE_NOT_RUN) {
error = RPL_CHANNEL_SERVICE_RECEIVER_CONNECTION_ERROR;
break;
}
my_sleep(100);
}
}
err:
if (!channel_map_already_locked) {
channel_map.unlock();
}
if (thd_created) {
delete_surrogate_thread(thd);
}
return error;
}
int channel_stop(Master_info *mi, int threads_to_stop, long timeout) {
channel_map.assert_some_lock();
if (mi == nullptr) {
return RPL_CHANNEL_SERVICE_CHANNEL_DOES_NOT_EXISTS_ERROR;
}
int thread_mask = 0;
int server_thd_mask = 0;
int error = 0;
bool thd_init = false;
mi->channel_wrlock();
lock_slave_threads(mi);
init_thread_mask(&server_thd_mask, mi, false /* not inverse*/);
if ((threads_to_stop & CHANNEL_APPLIER_THREAD) &&
(server_thd_mask & REPLICA_SQL)) {
thread_mask |= REPLICA_SQL;
}
if ((threads_to_stop & CHANNEL_RECEIVER_THREAD) &&
(server_thd_mask & REPLICA_IO)) {
thread_mask |= REPLICA_IO;
}
if ((threads_to_stop & CHANNEL_RECEIVER_THREAD) &&
(server_thd_mask & SLAVE_MONITOR)) {
thread_mask |= SLAVE_MONITOR;
}
if (thread_mask == 0) {
goto end;
}
thd_init = init_thread_context();
if (current_thd) current_thd->set_skip_readonly_check();
error = terminate_slave_threads(mi, thread_mask, timeout, false);
if (current_thd) current_thd->reset_skip_readonly_check();
end:
unlock_slave_threads(mi);
mi->channel_unlock();
if (thd_init) {
clean_thread_context();
}
return error;
}
int channel_stop(const char *channel, int threads_to_stop, long timeout) {
DBUG_TRACE;
channel_map.rdlock();
Master_info *mi = channel_map.get_mi(channel);
int error = channel_stop(mi, threads_to_stop, timeout);
channel_map.unlock();
return error;
}
int channel_stop_all(int threads_to_stop, long timeout,
std::string *error_message) {
Master_info *mi = nullptr;
/* Error related variables */
int error = 0;
std::stringstream err_msg_ss;
err_msg_ss << "Error stopping channel(s): ";
channel_map.rdlock();
for (mi_map::iterator it = channel_map.begin(); it != channel_map.end();
it++) {
mi = it->second;
if (mi) {
int channel_error = channel_stop(mi, threads_to_stop, timeout);
DBUG_EXECUTE_IF("group_replication_stop_all_channels_failure",
{ channel_error = 1; });
if (channel_error &&
channel_error != RPL_CHANNEL_SERVICE_CHANNEL_DOES_NOT_EXISTS_ERROR) {
error = channel_error;
mi->report(ERROR_LEVEL, error,
"Error stopping channel: %s. Got error: %d",
mi->get_channel(), error);
err_msg_ss << " '" << mi->get_channel() << "' [error number: " << error
<< "],";
}
}
}
if (error) {
*error_message = err_msg_ss.str();
(*error_message)[error_message->length() - 1] = '.';
error_message->append(
" Please check the error log for additional details.");
}
channel_map.unlock();
return error;
}
class Kill_binlog_dump : public Do_THD_Impl {
public:
Kill_binlog_dump() = default;
void operator()(THD *thd_to_kill) override {
if (thd_to_kill->get_command() == COM_BINLOG_DUMP ||
thd_to_kill->get_command() == COM_BINLOG_DUMP_GTID) {
assert(thd_to_kill != current_thd);
MUTEX_LOCK(thd_data_lock, &thd_to_kill->LOCK_thd_data);
thd_to_kill->duplicate_slave_id = true;
thd_to_kill->awake(THD::KILL_CONNECTION);
}
}
};
int binlog_dump_thread_kill() {
DBUG_TRACE;
Global_THD_manager *thd_manager = Global_THD_manager::get_instance();
Kill_binlog_dump kill_binlog_dump;
thd_manager->do_for_all_thd(&kill_binlog_dump);
return 0;
}
int channel_purge_queue(const char *channel, bool reset_all) {
DBUG_TRACE;
channel_map.wrlock();
Master_info *mi = channel_map.get_mi(channel);
if (mi == nullptr) {
channel_map.unlock();
return RPL_CHANNEL_SERVICE_CHANNEL_DOES_NOT_EXISTS_ERROR;
}
bool thd_init = init_thread_context();
int error = reset_slave(current_thd, mi, reset_all);
channel_map.unlock();
if (thd_init) {
clean_thread_context();
}
return error;
}
bool channel_is_active(const char *channel,
enum_channel_thread_types thd_type) {
int thread_mask = 0;
DBUG_TRACE;
channel_map.rdlock();
Master_info *mi = channel_map.get_mi(channel);
if (mi == nullptr) {
channel_map.unlock();
return false;
}
init_thread_mask(&thread_mask, mi, false /* not inverse*/);
channel_map.unlock();
switch (thd_type) {
case CHANNEL_NO_THD:
return true; // return true as the channel exists
case CHANNEL_RECEIVER_THREAD:
return thread_mask & REPLICA_IO;
case CHANNEL_APPLIER_THREAD:
return thread_mask & REPLICA_SQL;
default:
assert(0);
}
return false;
}
int channel_get_thread_id(const char *channel,
enum_channel_thread_types thd_type,
unsigned long **thread_id, bool need_lock) {
DBUG_TRACE;
int number_threads = -1;
if (need_lock) channel_map.rdlock();
channel_map.assert_some_lock();
Master_info *mi = channel_map.get_mi(channel);
if (mi == nullptr) {
if (need_lock) channel_map.unlock();
return RPL_CHANNEL_SERVICE_CHANNEL_DOES_NOT_EXISTS_ERROR;
}
switch (thd_type) {
case CHANNEL_RECEIVER_THREAD:
mysql_mutex_lock(&mi->info_thd_lock);
if (mi->info_thd != nullptr) {
*thread_id = (unsigned long *)my_malloc(
PSI_NOT_INSTRUMENTED, sizeof(unsigned long), MYF(MY_WME));
**thread_id = mi->info_thd->thread_id();
number_threads = 1;
}
mysql_mutex_unlock(&mi->info_thd_lock);
break;
case CHANNEL_APPLIER_THREAD:
if (mi->rli != nullptr) {
mysql_mutex_lock(&mi->rli->run_lock);
if (mi->rli->replica_parallel_workers > 0) {
// Parallel applier.
size_t num_workers = mi->rli->get_worker_count();
number_threads = 1 + num_workers;
*thread_id = (unsigned long *)my_malloc(
PSI_NOT_INSTRUMENTED, number_threads * sizeof(unsigned long),
MYF(MY_WME));
unsigned long *thread_id_pointer = *thread_id;
// Set default values on thread_id array.
for (int i = 0; i < number_threads; i++, thread_id_pointer++)
*thread_id_pointer = -1;
thread_id_pointer = *thread_id;
// Coordinator thread id.
if (mi->rli->info_thd != nullptr) {
mysql_mutex_lock(&mi->rli->info_thd_lock);
*thread_id_pointer = mi->rli->info_thd->thread_id();
mysql_mutex_unlock(&mi->rli->info_thd_lock);
thread_id_pointer++;
}
// Workers thread id.
if (mi->rli->workers_array_initialized) {
for (size_t i = 0; i < num_workers; i++, thread_id_pointer++) {
Slave_worker *worker = mi->rli->get_worker(i);
if (worker != nullptr) {
mysql_mutex_lock(&worker->jobs_lock);
if (worker->info_thd != nullptr &&
worker->running_status != Slave_worker::NOT_RUNNING) {
mysql_mutex_lock(&worker->info_thd_lock);
*thread_id_pointer = worker->info_thd->thread_id();
mysql_mutex_unlock(&worker->info_thd_lock);
}
mysql_mutex_unlock(&worker->jobs_lock);
}
}
}
} else {
// Sequential applier.
if (mi->rli->info_thd != nullptr) {
*thread_id = (unsigned long *)my_malloc(
PSI_NOT_INSTRUMENTED, sizeof(unsigned long), MYF(MY_WME));
mysql_mutex_lock(&mi->rli->info_thd_lock);
**thread_id = mi->rli->info_thd->thread_id();
mysql_mutex_unlock(&mi->rli->info_thd_lock);
number_threads = 1;
}
}
mysql_mutex_unlock(&mi->rli->run_lock);
}
break;
default:
break;
}
if (need_lock) channel_map.unlock();
return number_threads;
}
long long channel_get_last_delivered_gno(const char *channel, int sidno) {
DBUG_TRACE;
channel_map.rdlock();
Master_info *mi = channel_map.get_mi(channel);
if (mi == nullptr) {
channel_map.unlock();
return RPL_CHANNEL_SERVICE_CHANNEL_DOES_NOT_EXISTS_ERROR;
}
rpl_gno last_gno = 0;
Checkable_rwlock *tsid_lock = mi->rli->get_tsid_lock();
tsid_lock->rdlock();
last_gno = mi->rli->get_gtid_set()->get_last_gno(sidno);
tsid_lock->unlock();
#if !defined(NDEBUG)
const Gtid_set *retrieved_gtid_set = mi->rli->get_gtid_set();
char *retrieved_gtid_set_string = nullptr;
tsid_lock->wrlock();
retrieved_gtid_set->to_string(&retrieved_gtid_set_string);
tsid_lock->unlock();
DBUG_PRINT("info", ("get_last_delivered_gno retrieved_set_string: %s",
retrieved_gtid_set_string));
my_free(retrieved_gtid_set_string);
#endif
channel_map.unlock();
return last_gno;
}
int channel_add_executed_gtids_to_received_gtids(const char *channel) {
DBUG_TRACE;
channel_map.rdlock();
Master_info *mi = channel_map.get_mi(channel);
if (mi == nullptr) {
channel_map.unlock();
return RPL_CHANNEL_SERVICE_CHANNEL_DOES_NOT_EXISTS_ERROR;
}
global_tsid_lock->wrlock();
enum_return_status return_status =
mi->rli->add_gtid_set(gtid_state->get_executed_gtids());
global_tsid_lock->unlock();
channel_map.unlock();
return return_status != RETURN_STATUS_OK;
}
int channel_queue_packet(const char *channel, const char *buf,
unsigned long event_len) {
int result;
DBUG_TRACE;
channel_map.rdlock();
Master_info *mi = channel_map.get_mi(channel);
if (mi == nullptr) {
channel_map.unlock();
return RPL_CHANNEL_SERVICE_CHANNEL_DOES_NOT_EXISTS_ERROR;
}
channel_map.unlock();
result = queue_event(mi, buf, event_len, false /*flush_master_info*/);
return result;
}
int channel_wait_until_apply_queue_applied(const char *channel,
double timeout) {
DBUG_TRACE;
channel_map.rdlock();
Master_info *mi = channel_map.get_mi(channel);
if (mi == nullptr) {
channel_map.unlock();
return RPL_CHANNEL_SERVICE_CHANNEL_DOES_NOT_EXISTS_ERROR;
}
mi->inc_reference();
channel_map.unlock();
/*
The retrieved_gtid_set (rli->get_gtid_set) has its own tsid_map/tsid_lock
and do not use global_tsid_map/global_tsid_lock. Instead of blocking both
sid locks on each wait iteration at rli->wait_for_gtid_set(Gtid_set), it
would be better to use rli->wait_for_gtid_set(char *) that will create a
new Gtid_set based on global_tsid_map.
*/
char *retrieved_gtid_set_buf;
mi->rli->get_tsid_lock()->wrlock();
mi->rli->get_gtid_set()->to_string(&retrieved_gtid_set_buf);
mi->rli->get_tsid_lock()->unlock();
int error = mi->rli->wait_for_gtid_set(current_thd, retrieved_gtid_set_buf,
timeout, false);
my_free(retrieved_gtid_set_buf);
mi->dec_reference();
if (error == -1) return REPLICATION_THREAD_WAIT_TIMEOUT_ERROR;
if (error == -2) return REPLICATION_THREAD_WAIT_NO_INFO_ERROR;
return error;
}
int channel_wait_until_transactions_applied(const char *channel,
const char *gtid_set,
double timeout,
bool update_THD_status) {
DBUG_TRACE;
channel_map.rdlock();
Master_info *mi = channel_map.get_mi(channel);
if (mi == nullptr) {
channel_map.unlock(); /* purecov: inspected */
return RPL_CHANNEL_SERVICE_CHANNEL_DOES_NOT_EXISTS_ERROR; /* purecov:
inspected */
}
mi->inc_reference();
channel_map.unlock();
int error = mi->rli->wait_for_gtid_set(current_thd, gtid_set, timeout,
update_THD_status);
mi->dec_reference();
if (error == -1) return REPLICATION_THREAD_WAIT_TIMEOUT_ERROR;
if (error == -2) return REPLICATION_THREAD_WAIT_NO_INFO_ERROR;
return error;
}
int channel_is_applier_waiting(const char *channel) {
DBUG_TRACE;
int result = RPL_CHANNEL_SERVICE_CHANNEL_DOES_NOT_EXISTS_ERROR;
channel_map.rdlock();
Master_info *mi = channel_map.get_mi(channel);
if (mi == nullptr) {
channel_map.unlock();
return result;
}
unsigned long *thread_ids = nullptr;
int number_appliers = channel_get_thread_id(channel, CHANNEL_APPLIER_THREAD,
&thread_ids, false);
if (number_appliers <= 0) {
goto end;
}
if (number_appliers == 1) {
result = channel_is_applier_thread_waiting(*thread_ids);
} else if (number_appliers > 1) {
int waiting = 0;
// Check if coordinator is waiting.
waiting += channel_is_applier_thread_waiting(thread_ids[0]);
// Check if workers are waiting.
for (int i = 1; i < number_appliers; i++)
waiting += channel_is_applier_thread_waiting(thread_ids[i], true);
// Check if all are waiting.
if (waiting == number_appliers)
result = 1;
else
result = 0;
}
end:
channel_map.unlock();
my_free(thread_ids);
return result;
}
int channel_is_applier_thread_waiting(unsigned long thread_id, bool worker) {
DBUG_TRACE;
int result = -1;
Find_thd_with_id find_thd_with_id(thread_id);
THD_ptr thd_ptr =
Global_THD_manager::get_instance()->find_thd(&find_thd_with_id);
if (thd_ptr) {
if (thd_ptr->get_current_stage_key() ==
(worker ? stage_replica_waiting_event_from_coordinator
: stage_replica_has_read_all_relay_log)
.m_key)
result = 1;
else
result = 0;
}
return result;
}
int channel_flush(const char *channel) {
DBUG_TRACE;
channel_map.rdlock();
Master_info *mi = channel_map.get_mi(channel);
if (mi == nullptr) {
channel_map.unlock();
return RPL_CHANNEL_SERVICE_CHANNEL_DOES_NOT_EXISTS_ERROR;
}
bool error = (flush_relay_logs(mi, mi->info_thd) == 1);
channel_map.unlock();
return error ? 1 : 0;
}
int channel_get_retrieved_gtid_set(const char *channel, char **retrieved_set) {
DBUG_TRACE;
channel_map.rdlock();
Master_info *mi = channel_map.get_mi(channel);
if (mi == nullptr) {
channel_map.unlock();
return RPL_CHANNEL_SERVICE_CHANNEL_DOES_NOT_EXISTS_ERROR;
}
mi->inc_reference();
channel_map.unlock();
int error = 0;
const Gtid_set *receiver_gtid_set = mi->rli->get_gtid_set();
if (receiver_gtid_set->to_string(retrieved_set, true /*need_lock*/) == -1)
error = ER_OUTOFMEMORY;
mi->dec_reference();
return error;
}
int channel_get_credentials(const char *channel, std::string &username,
std::string &password) {
DBUG_TRACE;
String_set user_store, pass_store, auth_store;
if (!Rpl_channel_credentials::get_instance().get_credentials(
channel, user_store, pass_store, auth_store)) {
if (user_store.first) username = user_store.second;
if (pass_store.first) password = pass_store.second;
return 0;
}
channel_map.rdlock();
Master_info *mi = channel_map.get_mi(channel);
if (mi == nullptr) {
channel_map.unlock();
return RPL_CHANNEL_SERVICE_CHANNEL_DOES_NOT_EXISTS_ERROR;
}
mi->inc_reference();
channel_map.unlock();
char pass[MAX_PASSWORD_LENGTH + 1];
size_t pass_size;
mi->get_password(pass, &pass_size);
username.assign(mi->get_user());
password.assign(pass, pass_size);
mi->dec_reference();
return 0;
}
int channel_get_network_namespace(const char *channel, std::string &net_ns) {
DBUG_TRACE;
channel_map.rdlock();
Master_info *mi = channel_map.get_mi(channel);
if (mi == nullptr) {
channel_map.unlock();
return RPL_CHANNEL_SERVICE_CHANNEL_DOES_NOT_EXISTS_ERROR;
}
mi->inc_reference();
channel_map.unlock();
net_ns.assign(mi->network_namespace_str());
mi->dec_reference();
return 0;
}
int channel_get_gtid_set_to_apply(const char *channel,
std::string &gtid_set_to_apply) {
DBUG_TRACE;
int error = 0;
channel_map.rdlock();
Master_info *mi = channel_map.get_mi(channel);
if (mi == nullptr) {
channel_map.unlock();
return RPL_CHANNEL_SERVICE_CHANNEL_DOES_NOT_EXISTS_ERROR;
}
mi->inc_reference();
channel_map.unlock();
Tsid_map tsid_map{nullptr};
Gtid_set gtids{&tsid_map, nullptr};
/* RECEIVED_TRANSACTION_SET */
mi->rli->get_tsid_lock()->wrlock();
if (RETURN_STATUS_OK != gtids.add_gtid_set(mi->rli->get_gtid_set())) {
error = 1;
}
mi->rli->get_tsid_lock()->unlock();
mi->dec_reference();
if (error) {
return error;
}
/* GTID_EXECUTED */
global_tsid_lock->wrlock();
gtids.remove_gtid_set(gtid_state->get_executed_gtids());
global_tsid_lock->unlock();
char *gtids_string = nullptr;
if (-1 != gtids.to_string(&gtids_string)) {
gtid_set_to_apply.assign(gtids_string);
} else {
error = ER_OUTOFMEMORY;
}
my_free(gtids_string);
return error;
}
bool channel_is_stopping(const char *channel,
enum_channel_thread_types thd_type) {
bool is_stopping = false;
DBUG_TRACE;
channel_map.rdlock();
Master_info *mi = channel_map.get_mi(channel);
if (mi == nullptr) {
channel_map.unlock();
return false;
}
switch (thd_type) {
case CHANNEL_NO_THD:
break;
case CHANNEL_RECEIVER_THREAD:
is_stopping = likely(mi->atomic_is_stopping);
break;
case CHANNEL_APPLIER_THREAD:
is_stopping = likely(mi->rli->atomic_is_stopping);
break;
default:
assert(0);
}
channel_map.unlock();
return is_stopping;
}
bool is_partial_transaction_on_channel_relay_log(const char *channel) {
DBUG_TRACE;
channel_map.rdlock();
Master_info *mi = channel_map.get_mi(channel);
if (mi == nullptr) {
channel_map.unlock();
return false;
}
bool ret = mi->transaction_parser.is_inside_transaction();
channel_map.unlock();
return ret;
}
bool channel_has_same_uuid_as_group_name(const char *group_name) {
DBUG_TRACE;
Master_info *mi = nullptr;
channel_map.rdlock();
raii::Sentry<> map_lock_sentry{[&]() -> void { channel_map.unlock(); }};
for (mi_map::iterator it = channel_map.begin(); it != channel_map.end();
it++) {
mi = it->second;
if (mi != nullptr &&
mi->rli->m_assign_gtids_to_anonymous_transactions_info.get_type() >
Assign_gtids_to_anonymous_transactions_info::enum_type::AGAT_OFF) {
if (!(strcmp((mi->rli->m_assign_gtids_to_anonymous_transactions_info
.get_value())
.data(),
group_name))) {
return true;
}
}
}
return false;
}
bool is_any_slave_channel_running(int thread_mask) {
DBUG_TRACE;
Master_info *mi = nullptr;
bool is_running;
channel_map.rdlock();
for (mi_map::iterator it = channel_map.begin(); it != channel_map.end();
it++) {
mi = it->second;
if (mi) {
if ((thread_mask & REPLICA_IO) != 0) {
mysql_mutex_lock(&mi->run_lock);
is_running = mi->slave_running;
mysql_mutex_unlock(&mi->run_lock);
if (is_running) {
channel_map.unlock();
return true;
}
}
if ((thread_mask & REPLICA_SQL) != 0) {
mysql_mutex_lock(&mi->rli->run_lock);
is_running = mi->rli->slave_running;
mysql_mutex_unlock(&mi->rli->run_lock);
if (is_running) {
channel_map.unlock();
return true;
}
}
}
}
channel_map.unlock();
return false;
}
bool is_any_slave_channel_running_with_failover_enabled(int thread_mask) {
DBUG_TRACE;
Master_info *mi = nullptr;
bool is_running;
channel_map.rdlock();
for (mi_map::iterator it = channel_map.begin(); it != channel_map.end();
it++) {
mi = it->second;
if (mi && Master_info::is_configured(mi) &&
mi->is_source_connection_auto_failover()) {
if ((thread_mask & REPLICA_IO) != 0) {
mysql_mutex_lock(&mi->run_lock);
is_running = mi->slave_running;
mysql_mutex_unlock(&mi->run_lock);
if (is_running) {
channel_map.unlock();
return true;
}
}
if ((thread_mask & REPLICA_SQL) != 0) {
mysql_mutex_lock(&mi->rli->run_lock);
is_running = mi->rli->slave_running;
mysql_mutex_unlock(&mi->rli->run_lock);
if (is_running) {
channel_map.unlock();
return true;
}
}
}
}
assert(!Source_IO_monitor::get_instance()->is_monitoring_process_running());
channel_map.unlock();
return false;
}
enum_slave_channel_status
has_any_slave_channel_open_temp_table_or_is_its_applier_running() {
DBUG_TRACE;
Master_info *mi = nullptr;
bool is_applier_running = false;
bool has_open_temp_tables = false;
mi_map::iterator it;
channel_map.rdlock();
mi_map::iterator it_end = channel_map.end();
for (it = channel_map.begin(); it != channel_map.end(); ++it) {
mi = it->second;
if (Master_info::is_configured(mi)) {
mysql_mutex_lock(&mi->rli->run_lock);
is_applier_running = mi->rli->slave_running;
if (mi->rli->atomic_channel_open_temp_tables > 0)
has_open_temp_tables = true;
if (is_applier_running || has_open_temp_tables) {
/*
Stop acquiring more run_locks and start to release the held
run_locks once finding that a slave channel applier thread
is running or a slave channel has open temporary table(s),
and record the stop position.
*/
it_end = ++it;
break;
}
}
}
/*
Release the held run_locks until the stop position recorded in above
or the end of the channel_map.
*/
for (it = channel_map.begin(); it != it_end; ++it) {
mi = it->second;
if (Master_info::is_configured(mi)) mysql_mutex_unlock(&mi->rli->run_lock);
}
channel_map.unlock();
if (has_open_temp_tables)
return SLAVE_CHANNEL_HAS_OPEN_TEMPORARY_TABLE;
else if (is_applier_running)
return SLAVE_CHANNEL_APPLIER_IS_RUNNING;
return SLAVE_CHANNEL_NO_APPLIER_RUNNING_AND_NO_OPEN_TEMPORARY_TABLE;
}
int channel_delete_credentials(const char *channel_name) {
DBUG_TRACE;
return Rpl_channel_credentials::get_instance().delete_credentials(
channel_name);
}
bool start_failover_channels() {
DBUG_TRACE;
bool error = false;
channel_map.wrlock();
for (mi_map::iterator it = channel_map.begin();
!error && it != channel_map.end(); it++) {
Master_info *mi = it->second;
if (Master_info::is_configured(mi) &&
mi->is_source_connection_auto_failover()) {
Channel_connection_info info;
initialize_channel_connection_info(&info);
int thread_mask = 0;
thread_mask |= CHANNEL_APPLIER_THREAD;
thread_mask |= CHANNEL_RECEIVER_THREAD;
DBUG_EXECUTE_IF("force_error_on_start_failover_channels", {
channel_map.unlock();
return 1;
});
error = channel_start(mi->get_channel(), &info, thread_mask, false,
true /* use_server_mta_configuration*/,
true /* channel_map_already_locked */);
}
}
channel_map.unlock();
return error;
}
bool channel_change_source_connection_auto_failover(const char *channel,
bool status) {
bool error = false;
bool thd_created = false;
THD *thd = current_thd;
channel_map.assert_some_wrlock();
Master_info *mi = channel_map.get_mi(channel);
if (!Master_info::is_configured(mi)) {
LogErr(ERROR_LEVEL, ER_GRP_RPL_FAILOVER_CONF_CHANNEL_DOES_NOT_EXIST,
channel);
return true;
}
if (!thd) {
thd_created = true;
thd = create_surrogate_thread();
thd->set_skip_readonly_check();
}
mi->channel_wrlock();
lock_slave_threads(mi);
/*
When autocommit= 0, we force a new transaction to prevent
table access deadlocks when a master and slave info repositories
transaction is executed after changing SOURCE_CONNECTION_AUTO_FAILOVER.
*/
if (thd->variables.option_bits & OPTION_NOT_AUTOCOMMIT) {
assert(!(thd->server_status & SERVER_STATUS_IN_TRANS));
if (trans_begin(thd)) {
unlock_slave_threads(mi);
mi->channel_unlock();
if (thd_created) delete_surrogate_thread(thd);
return true;
}
}
if (status && !mi->is_source_connection_auto_failover()) {
mi->set_source_connection_auto_failover();
error |= (flush_master_info(mi, true, true, false) != 0);
}
if (!status && mi->is_source_connection_auto_failover()) {
mi->unset_source_connection_auto_failover();
error |= (flush_master_info(mi, true, true, false) != 0);
}
/*
When autocommit= 0, we force a new transaction to prevent
table access deadlocks when a master and slave info repositories
transaction is executed after changing SOURCE_CONNECTION_AUTO_FAILOVER.
*/
if (thd->variables.option_bits & OPTION_NOT_AUTOCOMMIT) {
if (trans_commit(thd)) {
unlock_slave_threads(mi);
mi->channel_unlock();
if (thd_created) delete_surrogate_thread(thd);
return true;
}
}
unlock_slave_threads(mi);
mi->channel_unlock();
if (thd_created) delete_surrogate_thread(thd);
return error;
}
bool unset_source_connection_auto_failover_on_all_channels() {
channel_map.assert_some_wrlock();
bool error = false;
for (mi_map::iterator it = channel_map.begin();
!error && it != channel_map.end(); it++) {
Master_info *mi = it->second;
if (Master_info::is_configured(mi) &&
mi->is_source_connection_auto_failover()) {
error |= channel_change_source_connection_auto_failover(mi->get_channel(),
false);
}
}
return error;
}
void reload_failover_channels_status() {
DBUG_TRACE;
channel_map.rdlock();
rpl_acf_configuration_handler->reload_failover_channels_status();
channel_map.unlock();
}
bool get_replication_failover_channels_configuration(
std::string &serialized_configuration) {
DBUG_TRACE;
return rpl_acf_configuration_handler->get_configuration(
serialized_configuration);
}
bool set_replication_failover_channels_configuration(
const std::vector<std::string>
&exchanged_replication_failover_channels_serialized_configuration) {
DBUG_TRACE;
channel_map.wrlock();
bool error = rpl_acf_configuration_handler->set_configuration(
exchanged_replication_failover_channels_serialized_configuration);
channel_map.unlock();
return error;
}
bool force_my_replication_failover_channels_configuration_on_all_members() {
DBUG_TRACE;
return rpl_acf_configuration_handler
->force_my_replication_failover_channels_configuration_on_all_members();
}
Reference in a new issue View git blame Copy permalink