group_replication/src/certifier.cc

/* Copyright (c) 2014, 2021, 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 also distributed 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 included with MySQL.

   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,
   51 Franklin Street, Suite 500, Boston, MA 02110-1335 USA */

#include <signal.h>
#include <map>

#include "certifier.h"
#include "plugin.h"
#include "plugin_log.h"
#include "observer_trans.h"

#include "sql_service_command.h"

const std::string Certifier::GTID_EXTRACTED_NAME= "gtid_extracted";
const std::string Certifier::CERTIFICATION_INFO_ERROR_NAME =
    "certification_info_error";

static void *launch_broadcast_thread(void* arg)
{
  Certifier_broadcast_thread *handler= (Certifier_broadcast_thread*) arg;
  handler->dispatcher();
  return 0;
}

Certifier_broadcast_thread::Certifier_broadcast_thread()
  :aborted(false), broadcast_thd_running(false), broadcast_counter(0),
   broadcast_gtid_executed_period(BROADCAST_GTID_EXECUTED_PERIOD)
{
  DBUG_EXECUTE_IF("group_replication_certifier_broadcast_thread_big_period",
                  { broadcast_gtid_executed_period= 600; });

  mysql_mutex_init(key_GR_LOCK_cert_broadcast_run,
                   &broadcast_run_lock,
                   MY_MUTEX_INIT_FAST);
  mysql_cond_init(key_GR_COND_cert_broadcast_run,
                  &broadcast_run_cond);
  mysql_mutex_init(key_GR_LOCK_cert_broadcast_dispatcher_run,
                   &broadcast_dispatcher_lock,
                   MY_MUTEX_INIT_FAST);
  mysql_cond_init(key_GR_COND_cert_broadcast_dispatcher_run,
                  &broadcast_dispatcher_cond);
 }

Certifier_broadcast_thread::~Certifier_broadcast_thread()
{
  mysql_mutex_destroy(&broadcast_run_lock);
  mysql_cond_destroy(&broadcast_run_cond);
  mysql_mutex_destroy(&broadcast_dispatcher_lock);
  mysql_cond_destroy(&broadcast_dispatcher_cond);
}

int Certifier_broadcast_thread::initialize()
{
  DBUG_ENTER("Certifier_broadcast_thread::initialize");

  mysql_mutex_lock(&broadcast_run_lock);
  if (broadcast_thd_running)
  {
    mysql_mutex_unlock(&broadcast_run_lock); /* purecov: inspected */
    DBUG_RETURN(0); /* purecov: inspected */
  }

  aborted= false;

  if ((mysql_thread_create(key_GR_THD_cert_broadcast,
                           &broadcast_pthd,
                           get_connection_attrib(),
                           launch_broadcast_thread,
                           (void*)this)))
  {
    mysql_mutex_unlock(&broadcast_run_lock); /* purecov: inspected */
    DBUG_RETURN(1); /* purecov: inspected */
  }

  while (!broadcast_thd_running)
  {
    DBUG_PRINT("sleep",("Waiting for certifier broadcast thread to start"));
    mysql_cond_wait(&broadcast_run_cond, &broadcast_run_lock);
  }
  mysql_mutex_unlock(&broadcast_run_lock);

  DBUG_RETURN(0);
}


int Certifier_broadcast_thread::terminate()
{
  DBUG_ENTER("Certifier_broadcast_thread::terminate");

  mysql_mutex_lock(&broadcast_run_lock);
  if (!broadcast_thd_running)
  {
    mysql_mutex_unlock(&broadcast_run_lock);
    DBUG_RETURN(0);
  }

  aborted= true;
  while (broadcast_thd_running)
  {
    DBUG_PRINT("loop", ("killing certifier broadcast thread"));
    mysql_mutex_lock(&broadcast_thd->LOCK_thd_data);

    //awake the cycle
    mysql_mutex_lock(&broadcast_dispatcher_lock);
    mysql_cond_broadcast(&broadcast_dispatcher_cond);
    mysql_mutex_unlock(&broadcast_dispatcher_lock);

    broadcast_thd->awake(THD::NOT_KILLED);
    mysql_mutex_unlock(&broadcast_thd->LOCK_thd_data);
    mysql_cond_wait(&broadcast_run_cond, &broadcast_run_lock);
  }
  mysql_mutex_unlock(&broadcast_run_lock);

  DBUG_RETURN(0);
}


void Certifier_broadcast_thread::dispatcher()
{
  DBUG_ENTER("Certifier_broadcast_thread::dispatcher");

  //Thread context operations
  my_thread_init();
  THD *thd= new THD;
  thd->set_new_thread_id();
  thd->thread_stack= (char*) &thd;
  thd->store_globals();
  global_thd_manager_add_thd(thd);
  broadcast_thd= thd;

  mysql_mutex_lock(&broadcast_run_lock);
  broadcast_thd_running= true;
  mysql_cond_broadcast(&broadcast_run_cond);
  mysql_mutex_unlock(&broadcast_run_lock);

  struct timespec abstime;
  while (!aborted)
  {
    broadcast_counter++;

    applier_module->get_pipeline_stats_member_collector()
        ->send_stats_member_message();

    applier_module->get_flow_control_module()->flow_control_step();

    if (broadcast_counter % broadcast_gtid_executed_period == 0)
      broadcast_gtid_executed();

    mysql_mutex_lock(&broadcast_dispatcher_lock);
    if (aborted)
    {
      mysql_mutex_unlock(&broadcast_dispatcher_lock); /* purecov: inspected */
      break; /* purecov: inspected */
    }
    set_timespec(&abstime, 1);
    mysql_cond_timedwait(&broadcast_dispatcher_cond,
                         &broadcast_dispatcher_lock, &abstime);
    mysql_mutex_unlock(&broadcast_dispatcher_lock);

    /*
      Clear server sessions open caches on transactions observer.
      TODO: move this to a global scheduler.
    */
    if (broadcast_counter % 300 == 0)
      observer_trans_clear_io_cache_unused_list(); /* purecov: inspected */
  }

  Gcs_interface_factory::cleanup(Gcs_operations::get_gcs_engine());

  thd->release_resources();
  global_thd_manager_remove_thd(thd);
  delete thd;

  mysql_mutex_lock(&broadcast_run_lock);
  broadcast_thd_running= false;
  mysql_cond_broadcast(&broadcast_run_cond);
  mysql_mutex_unlock(&broadcast_run_lock);

  my_thread_end();

  DBUG_VOID_RETURN;
}

int Certifier_broadcast_thread::broadcast_gtid_executed()
{
  DBUG_ENTER("Certifier_broadcast_thread::broadcast_gtid_executed");

  /*
    Member may be still joining group so we need to check if:
      1) communication interfaces are ready to be used;
      2) member is ONLINE, that is, distributed recovery is complete.
  */
  if (local_member_info == NULL)
    DBUG_RETURN(0); /* purecov: inspected */
  Group_member_info::Group_member_status member_status=
      local_member_info->get_recovery_status();
  if (member_status != Group_member_info::MEMBER_ONLINE &&
      member_status != Group_member_info::MEMBER_IN_RECOVERY)
    DBUG_RETURN(0);

  int error= 0;
  uchar *encoded_gtid_executed= NULL;
  size_t length;
  get_server_encoded_gtid_executed(&encoded_gtid_executed, &length);

  Gtid_Executed_Message gtid_executed_message;
  std::vector<uchar> encoded_gtid_executed_message;
  gtid_executed_message.append_gtid_executed(encoded_gtid_executed, length);

  enum enum_gcs_error send_err=
      gcs_module->send_message(gtid_executed_message, true);
  if (send_err == GCS_MESSAGE_TOO_BIG)
  {
    log_message(MY_ERROR_LEVEL, "Broadcast of committed transactions message "
                                "failed. Message is too big."); /* purecov: inspected */
    error= 1; /* purecov: inspected */
  }
  else if (send_err == GCS_NOK)
  {
    log_message(MY_INFORMATION_LEVEL,
                "Broadcast of committed transactions message failed."); /* purecov: inspected */
    error= 1; /* purecov: inspected */
  }


#if !defined(NDEBUG)
  char *encoded_gtid_executed_string=
      encoded_gtid_set_to_string(encoded_gtid_executed, length);
  DBUG_PRINT("info", ("Certifier broadcast executed_set: %s", encoded_gtid_executed_string));
  my_free(encoded_gtid_executed_string);
#endif

  my_free(encoded_gtid_executed);
  DBUG_RETURN(error);
}


Certifier::Certifier()
  :initialized(false),
   positive_cert(0), negative_cert(0),
   parallel_applier_last_committed_global(1),
   parallel_applier_sequence_number(2),
   certifying_already_applied_transactions(false),
   gtid_assignment_block_size(1),
   gtids_assigned_in_blocks_counter(1),
   conflict_detection_enable(!local_member_info->in_primary_mode())
{
   last_conflict_free_transaction.clear();

#if !defined(NDEBUG)
  certifier_garbage_collection_block= false;
  /*
    Debug flag to block the garbage collection and discard incoming stable
    set messages while garbage collection is on going.
  */
  DBUG_EXECUTE_IF("certifier_garbage_collection_block",
                  certifier_garbage_collection_block= true;);

  same_member_message_discarded= false;
  /*
    Debug flag to check for similar member sending multiple messages.
  */
  DBUG_EXECUTE_IF("certifier_inject_duplicate_certifier_data_message",
                  same_member_message_discarded= true;);
#endif

  certification_info_sid_map= new Sid_map(NULL);
  incoming= new Synchronized_queue<Data_packet*>();

  stable_gtid_set_lock= new Checkable_rwlock(
#ifdef HAVE_PSI_INTERFACE
                                             key_GR_RWLOCK_cert_stable_gtid_set
#endif
                                            );
  stable_sid_map= new Sid_map(stable_gtid_set_lock);
  stable_gtid_set= new Gtid_set(stable_sid_map, stable_gtid_set_lock);
  broadcast_thread= new Certifier_broadcast_thread();

  group_gtid_sid_map= new Sid_map(NULL);
  group_gtid_executed= new Gtid_set(group_gtid_sid_map, NULL);
  group_gtid_extracted= new Gtid_set(group_gtid_sid_map, NULL);

  last_local_gtid.clear();

  mysql_mutex_init(key_GR_LOCK_certification_info, &LOCK_certification_info,
                   MY_MUTEX_INIT_FAST);
  mysql_mutex_init(key_GR_LOCK_cert_members, &LOCK_members,
                   MY_MUTEX_INIT_FAST);
}


Certifier::~Certifier()
{
  clear_certification_info();
  delete certification_info_sid_map;

  delete stable_gtid_set;
  delete stable_sid_map;
  delete stable_gtid_set_lock;
  delete broadcast_thread;
  delete group_gtid_executed;
  delete group_gtid_extracted;
  delete group_gtid_sid_map;

  clear_incoming();
  delete incoming;

  clear_members();
  mysql_mutex_destroy(&LOCK_certification_info);
  mysql_mutex_destroy(&LOCK_members);
}

int Certifier::initialize_server_gtid_set(bool get_server_gtid_retrieved)
{
  DBUG_ENTER("initialize_server_gtid_set");
  mysql_mutex_assert_owner(&LOCK_certification_info);
  int error= 0;
  Sql_service_command_interface *sql_command_interface= NULL;
  std::string gtid_executed;
  std::string applier_retrieved_gtids;

  rpl_sid group_sid;
  if (group_sid.parse(group_name_var) != RETURN_STATUS_OK)
  {
    log_message(MY_ERROR_LEVEL,
                "Unable to parse the group name during"
                " the Certification module initialization"); /* purecov: inspected */
    error= 1; /* purecov: inspected */
    goto end; /* purecov: inspected */
  }

  group_gtid_sid_map_group_sidno= group_gtid_sid_map->add_sid(group_sid);
  if (group_gtid_sid_map_group_sidno < 0)
  {
    log_message(MY_ERROR_LEVEL,
                "Unable to add the group_sid in the group_gtid_sid_map during"
                " the Certification module initialization"); /* purecov: inspected */
    error= 1; /* purecov: inspected */
    goto end; /* purecov: inspected */
  }

  if (group_gtid_executed->ensure_sidno(group_gtid_sid_map_group_sidno) != RETURN_STATUS_OK)
  {
    log_message(MY_ERROR_LEVEL,
                "Error updating group_gtid_executed GITD set during"
                " the Certification module initialization"); /* purecov: inspected */
    error= 1; /* purecov: inspected */
    goto end; /* purecov: inspected */
  }

  if (group_gtid_extracted->ensure_sidno(group_gtid_sid_map_group_sidno) != RETURN_STATUS_OK)
  {
    log_message(MY_ERROR_LEVEL,
                "Unable to handle the donor's transaction information"
                " when initializing the conflict detection component."
                " Possible out of memory error."); /* purecov: inspected */
    error= 1; /* purecov: inspected */
    goto end; /* purecov: inspected */
  }

  sql_command_interface= new Sql_service_command_interface();
  if (sql_command_interface->establish_session_connection(PSESSION_USE_THREAD) ||
      sql_command_interface->set_interface_user(GROUPREPL_USER))
  {
    log_message(MY_ERROR_LEVEL,
                "Error when establishing a server connection during"
                " the Certification module initialization"); /* purecov: inspected */
    error= 1; /* purecov: inspected */
    goto end; /* purecov: inspected */
  }

  error= sql_command_interface->get_server_gtid_executed(gtid_executed);
  DBUG_EXECUTE_IF("gr_server_gtid_executed_extraction_error", error=1;);
  if (error)
  {
    log_message(MY_WARNING_LEVEL,
                "Error when extracting this member GTID executed set."
                " Certification module can't be properly initialized");
    goto end;
  }

  if (group_gtid_executed->add_gtid_text(gtid_executed.c_str()) != RETURN_STATUS_OK)
  {
    log_message(MY_ERROR_LEVEL,
                "Error while adding the server GTID EXECUTED set to the"
                " group_gtid_execute during the Certification module"
                " initialization"); /* purecov: inspected */
    error= 1; /* purecov: inspected */
    goto end; /* purecov: inspected */
  }

  if (get_server_gtid_retrieved)
  {
    Replication_thread_api applier_channel("group_replication_applier");
    if (applier_channel.get_retrieved_gtid_set(applier_retrieved_gtids))
    {
      log_message(MY_WARNING_LEVEL,
                  "Error when extracting this member retrieved set for its applier."
                  " Certification module can't be properly initialized"); /* purecov: inspected */
      error= 1; /* purecov: inspected */
      goto end; /* purecov: inspected */
    }

    if (group_gtid_executed->add_gtid_text(applier_retrieved_gtids.c_str()) != RETURN_STATUS_OK)
    {
      log_message(MY_ERROR_LEVEL,
                  "Error while adding the member retrieved set to the"
                  " group_gtid_executed during the Certification module"
                  " initialization"); /* purecov: inspected */
      error= 1; /* purecov: inspected */
      goto end; /* purecov: inspected */
    }
  }

  compute_group_available_gtid_intervals();

end:
  delete sql_command_interface;

  DBUG_RETURN(error);
}

void Certifier::compute_group_available_gtid_intervals()
{
  DBUG_ENTER("Certifier::compute_group_available_gtid_intervals");
  mysql_mutex_assert_owner(&LOCK_certification_info);

  gtids_assigned_in_blocks_counter= 1;
  member_gtids.clear();
  group_available_gtid_intervals.clear();

  /*
    Compute the GTID intervals that are available by inverting the
    group_gtid_executed or group_gtid_extracted intervals.
  */
  Gtid_set::Const_interval_iterator ivit(certifying_already_applied_transactions
                                             ? group_gtid_extracted
                                             : group_gtid_executed,
                                         group_gtid_sid_map_group_sidno);
#ifndef NDEBUG
  if (certifying_already_applied_transactions)
    DBUG_PRINT("Certifier::compute_group_available_gtid_intervals()",
               ("Generating group transaction intervals from group_gtid_extracted"));
#endif

  const Gtid_set::Interval *iv= NULL, *iv_next= NULL;

  // The fist interval: UUID:100 -> we have the interval 1-99
  if ((iv= ivit.get()) != NULL)
  {
    if (iv->start > 1)
    {
      Gtid_set::Interval interval= {1, iv->start - 1, NULL};
      group_available_gtid_intervals.push_back(interval);
    }
  }

  // For each used interval find the upper bound and from there
  // add the free GTIDs up to the next interval or GNO_END.
  while ((iv= ivit.get()) != NULL)
  {
    ivit.next();
    iv_next= ivit.get();

    rpl_gno start= iv->end;
    rpl_gno end= GNO_END;
    if (iv_next != NULL)
      end= iv_next->start - 1;

    assert(start <= end);
    Gtid_set::Interval interval= {start, end, NULL};
    group_available_gtid_intervals.push_back(interval);
  }

  // No GTIDs used, so the available interval is the complete set.
  if (group_available_gtid_intervals.size() == 0)
  {
    Gtid_set::Interval interval= {1, GNO_END, NULL};
    group_available_gtid_intervals.push_back(interval);
  }

  DBUG_VOID_RETURN;
}

Gtid_set::Interval Certifier::reserve_gtid_block(longlong block_size)
{
  DBUG_ENTER("Certifier::reserve_gtid_block");
  assert(block_size > 1);
  mysql_mutex_assert_owner(&LOCK_certification_info);

  Gtid_set::Interval result;

  // We are out of intervals, we need to force intervals computation.
  if (group_available_gtid_intervals.size() == 0)
    compute_group_available_gtid_intervals();

  std::list<Gtid_set::Interval>::iterator it= group_available_gtid_intervals.begin();
  assert(it != group_available_gtid_intervals.end());

  /*
    We always have one or more intervals, the only thing to check
    is if the first interval is exhausted, if so we need to purge
    it to avoid future use.
  */
  if (block_size > it->end - it->start)
  {
    result= *it;
    group_available_gtid_intervals.erase(it);
  }
  else
  {
    result.start= it->start;
    result.end= it->start + block_size - 1;
    it->start= result.end + 1;
    assert(result.start <= result.end);
    assert(result.start < it->start);
  }

  DBUG_RETURN(result);
}

void Certifier::add_to_group_gtid_executed_internal(rpl_sidno sidno,
                                                    rpl_gno gno,
                                                    bool local)
{
  DBUG_ENTER("Certifier::add_to_group_gtid_executed_internal");
  mysql_mutex_assert_owner(&LOCK_certification_info);
  group_gtid_executed->_add_gtid(sidno, gno);
  if (local)
  {
    assert(sidno >0 && gno >0);
    last_local_gtid.set(sidno, gno);
  }
  /*
    We only need to track certified transactions on
    group_gtid_extracted while:
     1) certifier is handling already applied transactions
        on distributed recovery procedure;
     2) the transaction does have a group GTID.
  */
  if (certifying_already_applied_transactions &&
      sidno == group_gtid_sid_map_group_sidno)
    group_gtid_extracted->_add_gtid(sidno, gno);

  DBUG_VOID_RETURN;
}

void Certifier::clear_certification_info()
{
  for (Certification_info::iterator it= certification_info.begin();
       it != certification_info.end();
       ++it)
  {
    // We can only delete the last reference.
    if (it->second->unlink() == 0)
      delete it->second;
  }

  certification_info.clear();
}


void Certifier::clear_incoming()
{
  DBUG_ENTER("Certifier::clear_incoming");
  while (!this->incoming->empty())
  {
    Data_packet *packet= NULL;
    this->incoming->pop(&packet);
    delete packet;
  }
  DBUG_VOID_RETURN;
}

void Certifier::clear_members()
{
  DBUG_ENTER("Certifier::clear_members");
  mysql_mutex_lock(&LOCK_members);
  members.clear();
  mysql_mutex_unlock(&LOCK_members);
  DBUG_VOID_RETURN;
}

int Certifier::initialize(ulonglong gtid_assignment_block_size)
{
  DBUG_ENTER("Certifier::initialize");
  int error= 0;
  mysql_mutex_lock(&LOCK_certification_info);

  if (is_initialized())
  {
    error= 1; /* purecov: inspected */
    goto end; /* purecov: inspected */
  }

  assert(gtid_assignment_block_size >= 1);
  this->gtid_assignment_block_size= gtid_assignment_block_size;

  /*
    We need to initialize group_gtid_executed from both GTID_EXECUTED
    and applier retrieved GTID set to consider the already certified
    but not yet applied GTIDs, that may exist on applier relay log when
    this member is the one bootstrapping the group.
  */
  if (initialize_server_gtid_set(true))
  {
    log_message(MY_ERROR_LEVEL,
                "Error during Certification module initialization.");
    error= 1;
    goto end;
  }

  error= broadcast_thread->initialize();
  initialized= !error;

end:
  mysql_mutex_unlock(&LOCK_certification_info);
  DBUG_RETURN(error);
}


int Certifier::terminate()
{
  DBUG_ENTER("Certifier::terminate");
  int error= 0;

  if (is_initialized())
    error= broadcast_thread->terminate();

  DBUG_RETURN(error);
}


void Certifier::increment_parallel_applier_sequence_number(
    bool update_parallel_applier_last_committed_global)
{
  DBUG_ENTER("Certifier::increment_parallel_applier_sequence_number");
  mysql_mutex_assert_owner(&LOCK_certification_info);

  assert(parallel_applier_last_committed_global <
         parallel_applier_sequence_number);
  if (update_parallel_applier_last_committed_global)
    parallel_applier_last_committed_global= parallel_applier_sequence_number;

  parallel_applier_sequence_number++;

  DBUG_VOID_RETURN;
}


rpl_gno Certifier::certify(Gtid_set *snapshot_version,
                           std::list<const char*> *write_set,
                           bool generate_group_id,
                           const char *member_uuid,
                           Gtid_log_event *gle,
                           bool local_transaction)
{
  DBUG_ENTER("Certifier::certify");
  rpl_gno result= 0;
  const bool has_write_set= !write_set->empty();

  if (!is_initialized())
    DBUG_RETURN(-1); /* purecov: inspected */

  mysql_mutex_lock(&LOCK_certification_info);
  int64 transaction_last_committed= parallel_applier_last_committed_global;

  DBUG_EXECUTE_IF("certifier_force_1_negative_certification", {
                  DBUG_SET("-d,certifier_force_1_negative_certification");
                  goto end;});

  if (conflict_detection_enable)
  {
    for (std::list<const char*>::iterator it= write_set->begin();
         it != write_set->end();
         ++it)
    {
      Gtid_set *certified_write_set_snapshot_version=
          get_certified_write_set_snapshot_version(*it);

      /*
        If the previous certified transaction snapshot version is not
        a subset of the incoming transaction snapshot version, the current
        transaction was executed on top of outdated data, so it will be
        negatively certified. Otherwise, this transaction is marked
        certified and goes into applier.
      */
      if (certified_write_set_snapshot_version != NULL &&
          !certified_write_set_snapshot_version->is_subset(snapshot_version))
        goto end;
    }
  }

  if (certifying_already_applied_transactions &&
      !group_gtid_extracted->is_subset_not_equals(group_gtid_executed))
  {
    certifying_already_applied_transactions= false;

#ifndef NDEBUG
    char *group_gtid_executed_string= NULL;
    char *group_gtid_extracted_string= NULL;
    group_gtid_executed->to_string(&group_gtid_executed_string, true);
    group_gtid_extracted->to_string(&group_gtid_extracted_string, true);
    DBUG_PRINT("Certifier::certify()",
               ("Set certifying_already_applied_transactions to false. "
                "group_gtid_executed: \"%s\"; group_gtid_extracted_string: \"%s\"",
                group_gtid_executed_string, group_gtid_extracted_string));
    my_free(group_gtid_executed_string);
    my_free(group_gtid_extracted_string);
#endif
  }

  /*
    If the current transaction doesn't have a specified GTID, one
    for group UUID will be generated.
    This situation happens when transactions are executed with
    GTID_NEXT equal to AUTOMATIC_GROUP (the default case).
  */
  if (generate_group_id)
  {
    /*
      We need to ensure that group sidno does exist on snapshot
      version due to the following scenario:
        1) Member joins the group.
        2) Goes through recovery procedure, view change is queued to
           apply, member is marked ONLINE. This requires
             --group_replication_recovery_complete_at=TRANSACTIONS_CERTIFIED
           to happen.
        3) Despite the view change log event is still being applied,
           since the member is already ONLINE it can execute
           transactions. The first transaction from this member will
           not include any group GTID, since no group transaction is
           yet applied.
        4) As a result of this sequence snapshot_version will not
           contain any group GTID and the below instruction
             snapshot_version->_add_gtid(group_sidno, result);
           would fail because of that.
    */
    if (snapshot_version->ensure_sidno(group_sidno) != RETURN_STATUS_OK)
    {
      log_message(MY_ERROR_LEVEL,
                  "Error updating transaction snapshot version after"
                  " transaction being positively certified"); /* purecov: inspected */
      goto end; /* purecov: inspected */
    }

    result= get_group_next_available_gtid(member_uuid);
    if (result < 0)
      goto end;

    /*
      Add generated transaction GTID to transaction snapshot version.
    */
    snapshot_version->_add_gtid(group_sidno, result);

    /*
      Store last conflict free transaction identification.
      sidno must be relative to group_gtid_sid_map.
    */
    last_conflict_free_transaction.set(group_gtid_sid_map_group_sidno,
                                       result);

    DBUG_PRINT("info",
               ("Group replication Certifier: generated transaction "
                "identifier: %llu", result));
  }
  else
  {
    /*
      Check if it is an already used GTID
    */
    rpl_sidno sidno_for_group_gtid_sid_map= gle->get_sidno(group_gtid_sid_map);
    if (sidno_for_group_gtid_sid_map < 1)
    {
      log_message(MY_ERROR_LEVEL,
                  "Error fetching transaction sidno after transaction"
                  " being positively certified"); /* purecov: inspected */
      goto end; /* purecov: inspected */
    }
    if (group_gtid_executed->contains_gtid(sidno_for_group_gtid_sid_map, gle->get_gno()))
    {
      char buf[rpl_sid::TEXT_LENGTH + 1];
      gle->get_sid()->to_string(buf);

      log_message(MY_ERROR_LEVEL,
                  "The requested GTID '%s:%lld' was already used, the transaction will rollback"
                  , buf, gle->get_gno());
      goto end;
    }
    /*
      Add received transaction GTID to transaction snapshot version.
    */
    rpl_sidno sidno= gle->get_sidno(snapshot_version->get_sid_map());
    if (sidno < 1)
    {
      log_message(MY_ERROR_LEVEL,
                  "Error fetching transaction sidno after transaction"
                  " being positively certified"); /* purecov: inspected */
      goto end; /* purecov: inspected */
    }

    if (snapshot_version->ensure_sidno(sidno) != RETURN_STATUS_OK)
    {
      log_message(MY_ERROR_LEVEL,
                  "Error updating transaction snapshot version after"
                  " transaction being positively certified"); /* purecov: inspected */
      goto end; /* purecov: inspected */
    }
    snapshot_version->_add_gtid(sidno, gle->get_gno());

    /*
      Store last conflict free transaction identification.
      sidno must be relative to group_gtid_sid_map.
    */
    rpl_sidno last_conflict_free_transaction_sidno= gle->get_sidno(group_gtid_sid_map);
    if (last_conflict_free_transaction_sidno < 1)
    {
      log_message(MY_WARNING_LEVEL,
                  "Unable to update last conflict free transaction, "
                  "this transaction will not be tracked on "
                  "performance_schema.replication_group_member_stats.last_conflict_free_transaction"); /* purecov: inspected */
    }
    else
    {
      last_conflict_free_transaction.set(last_conflict_free_transaction_sidno,
                                         gle->get_gno());
    }

    result= 1;
    DBUG_PRINT("info",
               ("Group replication Certifier: there was no transaction identifier "
                "generated since transaction already had a GTID specified"));
  }

  /*
    Add the transaction's write set to certification info.
  */
  if (has_write_set)
  {
    // Only consider remote transactions for parallel applier indexes.
    int64 transaction_sequence_number=
        local_transaction ? -1 : parallel_applier_sequence_number;
    Gtid_set_ref *snapshot_version_value=
        new Gtid_set_ref(certification_info_sid_map, transaction_sequence_number);
    if (snapshot_version_value->add_gtid_set(snapshot_version) != RETURN_STATUS_OK)
    {
      result= 0; /* purecov: inspected */
      delete snapshot_version_value; /* purecov: inspected */
      log_message(MY_ERROR_LEVEL,
                  "Error updating transaction snapshot version reference "
                  "for internal storage"); /* purecov: inspected */
      goto end; /* purecov: inspected */
    }

    for(std::list<const char*>::iterator it= write_set->begin();
        it != write_set->end();
        ++it)
    {
      int64 item_previous_sequence_number= -1;

      add_item(*it, snapshot_version_value,
               &item_previous_sequence_number);

      /*
        Exclude previous sequence number that are smaller than global
        last committed and that are the current sequence number.
        transaction_last_committed is initialized with
        parallel_applier_last_committed_global on the beginning of
        this method.
      */
      if (item_previous_sequence_number > transaction_last_committed &&
          item_previous_sequence_number != parallel_applier_sequence_number)
        transaction_last_committed= item_previous_sequence_number;
    }
  }

  /*
    Update parallel applier indexes.
  */
  if (!local_transaction)
  {
    if (!has_write_set)
    {
      /*
        DDL does not have write-set, so we need to ensure that it
        is applied without any other transaction in parallel.
      */
      transaction_last_committed= parallel_applier_sequence_number - 1;
    }

    gle->last_committed= transaction_last_committed;
    gle->sequence_number= parallel_applier_sequence_number;
    assert(gle->last_committed >= 0);
    assert(gle->sequence_number > 0);
    assert(gle->last_committed < gle->sequence_number);

    increment_parallel_applier_sequence_number(!has_write_set);
  }

end:
  update_certified_transaction_count(result>0);

  mysql_mutex_unlock(&LOCK_certification_info);
  DBUG_PRINT("info", ("Group replication Certifier: certification result: %llu",
                      result));
  DBUG_RETURN(result);
}


int Certifier::add_specified_gtid_to_group_gtid_executed(Gtid_log_event *gle,
                                                         bool local)
{
  DBUG_ENTER("Certifier::add_specified_gtid_to_group_gtid_executed");

  mysql_mutex_lock(&LOCK_certification_info);
  rpl_sidno sidno= gle->get_sidno(group_gtid_sid_map);

  if (sidno < 1)
  {
    log_message(MY_ERROR_LEVEL,
                "Error fetching transaction sidno while adding to the "
                "group_gtid_executed set."); /* purecov: inspected */
    mysql_mutex_unlock(&LOCK_certification_info); /* purecov: inspected */
    DBUG_RETURN(1); /* purecov: inspected */
  }

  if (group_gtid_executed->ensure_sidno(sidno) != RETURN_STATUS_OK)
  {
    log_message(MY_ERROR_LEVEL,
                "Error while ensuring the sidno be present in the "
                "group_gtid_executed"); /* purecov: inspected */
    mysql_mutex_unlock(&LOCK_certification_info); /* purecov: inspected */
    DBUG_RETURN(1); /* purecov: inspected */
  }

  add_to_group_gtid_executed_internal(sidno, gle->get_gno(), local);

  mysql_mutex_unlock(&LOCK_certification_info);
  DBUG_RETURN(0);
}

int Certifier::add_group_gtid_to_group_gtid_executed(rpl_gno gno, bool local)
{
  DBUG_ENTER("Certifier::add_group_gtid_to_group_gtid_executed");
  mysql_mutex_lock(&LOCK_certification_info);
  add_to_group_gtid_executed_internal(group_gtid_sid_map_group_sidno, gno, local);
  mysql_mutex_unlock(&LOCK_certification_info);
  DBUG_RETURN(0);
}

/*
  This method will return the next GNO for the current transaction, it
  will work with two behaviours:

  1) member_uuid == NULL || gtid_assignment_block_size <= 1
     View_change_log_events creation does call this method with
     member_uuid set to NULL to force it to be created with the
     first available GNO of the group. This will ensure that all
     members do use the same GNO for it.
     After a View_change_log_event is created we recompute available
     GNOs to ensure that all members do have the same available GNOs
     set.
     This branch is also used when gtid_assignment_block_size is
     set to 1, meaning that GNO will be assigned sequentially
     according with certification order.

  2) On the second branch we assign GNOs according to intervals
     assigned to each member.
     To avoid having eternal gaps when a member do use all of its
     assigned GNOs, periodically we recompute the intervals, this
     will make that GNOs available to other members.
     The GNO is generated within the interval of available GNOs for
     a given member.
     When a member exhaust its assigned GNOs we reserve more for it
     from the available GNOs set.
*/
rpl_gno Certifier::get_group_next_available_gtid(const char *member_uuid)
{
  DBUG_ENTER("Certifier::get_group_next_available_gtid");
  mysql_mutex_assert_owner(&LOCK_certification_info);
  rpl_gno result= 0;

  if (member_uuid == NULL || gtid_assignment_block_size <= 1)
  {
    result= get_group_next_available_gtid_candidate(1, GNO_END);
    if (result < 0)
    {
      assert(result == -1);
      DBUG_RETURN(result);
    }

    /*
      If we did log a view change event we need to recompute
      intervals, so that all members start from the same
      intervals.
    */
    if (member_uuid == NULL && gtid_assignment_block_size > 1)
      compute_group_available_gtid_intervals();
  }
  else
  {
    /*
      After a number of rounds equal to block size the blocks are
      collected back so that the GTID holes can be filled up by
      following transactions from other members.
    */
    if (gtids_assigned_in_blocks_counter % (gtid_assignment_block_size + 1) == 0)
      compute_group_available_gtid_intervals();

    /*
      GTID is assigned in blocks to each member and are consumed
      from that block unless a new block is needed.
    */
    std::string member(member_uuid);
    std::map<std::string, Gtid_set::Interval>::iterator it=
        member_gtids.find(member);

    if (it == member_gtids.end())
    {
      // There is no block assigned to this member so get one.
      std::pair<std::map<std::string, Gtid_set::Interval>::iterator, bool> insert_ret;
      std::pair<std::string, Gtid_set::Interval> member_pair(member,
          reserve_gtid_block(gtid_assignment_block_size));
      insert_ret= member_gtids.insert(member_pair);
      assert(insert_ret.second == true);
      it= insert_ret.first;
    }

    result= get_group_next_available_gtid_candidate(it->second.start,
                                                    it->second.end);
    while (result == -2)
    {
      // Block has no available GTIDs, reserve more.
      it->second= reserve_gtid_block(gtid_assignment_block_size);
      result= get_group_next_available_gtid_candidate(it->second.start,
                                                      it->second.end);
    }
    if (result < 0)
      DBUG_RETURN(result);

    it->second.start= result;
    gtids_assigned_in_blocks_counter++;
  }

  assert(result > 0);
  DBUG_RETURN(result);
}

rpl_gno
Certifier::get_group_next_available_gtid_candidate(rpl_gno start,
                                                   rpl_gno end) const
{
  DBUG_ENTER("Certifier::get_group_next_available_gtid_candidate");
  assert(start > 0);
  assert(start <= end);
  mysql_mutex_assert_owner(&LOCK_certification_info);

  rpl_gno candidate= start;
  Gtid_set::Const_interval_iterator ivit(certifying_already_applied_transactions
                                             ? group_gtid_extracted
                                             : group_gtid_executed,
                                         group_gtid_sid_map_group_sidno);
#ifndef NDEBUG
  if (certifying_already_applied_transactions)
    DBUG_PRINT("Certifier::get_group_next_available_gtid_candidate()",
               ("Generating group transaction id from group_gtid_extracted"));
#endif

  /*
    Walk through available intervals until we find the correct one
    or return GNO exhausted error.
  */
  while (true)
  {
    assert(candidate >= start);
    const Gtid_set::Interval *iv= ivit.get();
    rpl_gno next_interval_start= iv != NULL ? iv->start : GNO_END;

    // Correct interval.
    if (candidate < next_interval_start)
    {
      if (candidate <= end)
        DBUG_RETURN(candidate);
      else
        DBUG_RETURN(-2);
    }

    if (iv == NULL)
    {
      log_message(MY_ERROR_LEVEL,
                  "Impossible to generate Global Transaction Identifier: "
                  "the integer component reached the maximal value. Restart "
                  "the group with a new group_replication_group_name.");
      DBUG_RETURN(-1);
    }

    candidate= std::max(candidate, iv->end);
    ivit.next();
  }
}

bool Certifier::add_item(const char* item, Gtid_set_ref *snapshot_version,
                         int64 *item_previous_sequence_number)
{
  DBUG_ENTER("Certifier::add_item");
  mysql_mutex_assert_owner(&LOCK_certification_info);
  bool error= true;
  std::string key(item);
  Certification_info::iterator it= certification_info.find(key);
  snapshot_version->link();

  if (it == certification_info.end())
  {
    std::pair<Certification_info::iterator, bool> ret=
        certification_info.insert(std::pair<std::string, Gtid_set_ref*>
                                  (key, snapshot_version));
    error= !ret.second;
  }
  else
  {
    *item_previous_sequence_number=
        it->second->get_parallel_applier_sequence_number();

    if (it->second->unlink() == 0)
      delete it->second;

    it->second= snapshot_version;
    error= false;
  }

  DBUG_RETURN(error);
}


Gtid_set *Certifier::get_certified_write_set_snapshot_version(const char* item)
{
  DBUG_ENTER("Certifier::get_certified_write_set_snapshot_version");
  mysql_mutex_assert_owner(&LOCK_certification_info);

  if (!is_initialized())
    DBUG_RETURN(NULL); /* purecov: inspected */

  Certification_info::iterator it;
  std::string item_str(item);

  it= certification_info.find(item_str);

  if (it == certification_info.end())
    DBUG_RETURN(NULL);
  else
    DBUG_RETURN(it->second);
}


int
Certifier::get_group_stable_transactions_set_string(char **buffer,
                                                    size_t *length)
{
  DBUG_ENTER("Certifier::get_group_stable_transactions_set_string");
  int error= 1;

  char *m_buffer= NULL;
  int m_length= stable_gtid_set->to_string(&m_buffer, true);
  if (m_length >= 0)
  {
    *buffer= m_buffer;
    *length= static_cast<size_t>(m_length);
    error= 0;
  }
  else
    my_free(m_buffer); /* purecov: inspected */

  DBUG_RETURN(error);
}


bool Certifier::set_group_stable_transactions_set(Gtid_set* executed_gtid_set)
{
  DBUG_ENTER("Certifier::set_group_stable_transactions_set");

  if (!is_initialized())
    DBUG_RETURN(true); /* purecov: inspected */

  if (executed_gtid_set == NULL)
  {
    log_message(MY_ERROR_LEVEL, "Invalid stable transactions set"); /* purecov: inspected */
    DBUG_RETURN(true); /* purecov: inspected */
  }

  stable_gtid_set_lock->wrlock();
  if (stable_gtid_set->add_gtid_set(executed_gtid_set) != RETURN_STATUS_OK)
  {
    stable_gtid_set_lock->unlock(); /* purecov: inspected */
    log_message(MY_ERROR_LEVEL, "Error updating stable transactions set"); /* purecov: inspected */
    DBUG_RETURN(true); /* purecov: inspected */
  }
  stable_gtid_set_lock->unlock();

  garbage_collect();

  DBUG_RETURN(false);
}

void Certifier::garbage_collect()
{
  DBUG_ENTER("Certifier::garbage_collect");
  DBUG_EXECUTE_IF("group_replication_do_not_clear_certification_database",
                    { DBUG_VOID_RETURN; };);

  mysql_mutex_lock(&LOCK_certification_info);

  /*
    When a transaction "t" is applied to all group members and for all
    ongoing, i.e., not yet committed or aborted transactions,
    "t" was already committed when they executed (thus "t"
    precedes them), then "t" is stable and can be removed from
    the certification info.
  */
  Certification_info::iterator it= certification_info.begin();
  stable_gtid_set_lock->wrlock();
  while (it != certification_info.end())
  {
    if (it->second->is_subset_not_equals(stable_gtid_set))
    {
      if (it->second->unlink() == 0)
        delete it->second;
      certification_info.erase(it++);
    }
    else
      ++it;
  }
  stable_gtid_set_lock->unlock();

  /*
    We need to update parallel applier indexes since we do not know
    what write sets were purged, which may cause transactions
    last committed to be incorrectly computed.
  */
  increment_parallel_applier_sequence_number(true);

#if !defined(NDEBUG)
  /*
    This part blocks the garbage collection process for 300 sec in order to
    simulate the case that while garbage collection is going on, we should
    skip the stable set messages round in order to prevent simultaneous
    access to stable_gtid_set.
  */
  if (certifier_garbage_collection_block)
  {
    certifier_garbage_collection_block= false;
    // my_sleep expects a given number of microseconds.
    my_sleep(broadcast_thread->BROADCAST_GTID_EXECUTED_PERIOD * 1500000);
  }
#endif

  mysql_mutex_unlock(&LOCK_certification_info);

  /*
    Applier channel received set does only contain the GTIDs of the
    remote (committed by other members) transactions. On the long
    term, the gaps may create performance issues on the received
    set update. To avoid that, periodically, we update the received
    set with the full set of transactions committed on the group,
    closing the gaps.
  */
  if (channel_add_executed_gtids_to_received_gtids(applier_module_channel_name))
  {
    log_message(MY_WARNING_LEVEL,
                "There was an error when filling the missing GTIDs on "
                "the applier channel received set. Despite not critical, "
                "on the long run this may cause performance issues"); /* purecov: inspected */
  }

  DBUG_VOID_RETURN;
}


int Certifier::handle_certifier_data(const uchar *data, ulong len,
                                     const Gcs_member_identifier& gcs_member_id)
{
  DBUG_ENTER("Certifier::handle_certifier_data");
  bool member_message_received= false;

  if (!is_initialized())
    DBUG_RETURN(1); /* purecov: inspected */

  mysql_mutex_lock(&LOCK_members);
  std::string member_id= gcs_member_id.get_member_id();
#if !defined(NDEBUG)
  if (same_member_message_discarded)
  {
    /*
      Injecting the member_id in the member's vector to simulate the case of
      same member sending multiple messages.
    */
    this->members.push_back(member_id);
  }
#endif

  if (this->get_members_size() != plugin_get_group_members_number())
  {
    /*
      We check for the member_id of the current message if it is present in
      the member vector or not. If it is present, we will need to discard the
      message. If not we will add the message in the incoming message
      synchronized queue for stable set handling.
    */
    std::vector<std::string>::iterator it;
    it= std::find(members.begin(), members.end(), member_id);
    if (it != members.end())
      member_message_received= true;
    else
      this->members.push_back(member_id);

    /*
      Since member is not present we can queue this message.
    */
    if (!member_message_received)
    {
      this->incoming->push(new Data_packet(data, len));
    }
    else
    {
      /*
        As member is already received we can throw the necessary warning of the
        member message already received.
      */
      Group_member_info *member_info=
          group_member_mgr->get_group_member_info_by_member_id(gcs_member_id);
      if (member_info != NULL)
      {
        log_message(MY_WARNING_LEVEL, "The member with address %s:%u has "
                    "already sent the stable set. Therefore discarding the second "
                    "message.", member_info->get_hostname().c_str(),
                    member_info->get_port());
        delete member_info;
      }
    }

    mysql_mutex_unlock(&LOCK_members);

    /*
      If the incoming message queue size is equal to the number of the
      members in the group, we are sure that each member has sent their
      gtid_executed. So we can go ahead with the stable set handling.
    */
    if (plugin_get_group_members_number() == this->incoming->size())
    {
      int error= stable_set_handle();
      /*
        Clearing the members to proceed with the next round of garbage
        collection.
      */
      clear_members();
      DBUG_RETURN(error);
    }
  }
  else
  {
    log_message(MY_WARNING_LEVEL, "Skipping this round of stable set "
                "computation as certification garbage collection process is "
                "still running."); /* purecov: inspected */
    mysql_mutex_unlock(&LOCK_members); /* purecov: inspected */
  }

#if !defined(NDEBUG)
  if (same_member_message_discarded)
  {
    /*
      Clearing the flag here as the members vector is not cleaned above.
    */
    same_member_message_discarded= false;
    clear_members();
  }
#endif

  DBUG_RETURN(0);
}

int Certifier::stable_set_handle()
{
  DBUG_ENTER("Certifier:stable_set_handle");

  Data_packet *packet= NULL;
  int error= 0;

  Sid_map sid_map(NULL);
  Gtid_set executed_set(&sid_map, NULL);

  /*
    Compute intersection between all received sets.
  */
  while(!error && !this->incoming->empty())
  {
    this->incoming->pop(&packet);

    if (packet == NULL)
    {
      log_message(MY_ERROR_LEVEL, "Null packet on certifier's queue"); /* purecov: inspected */
      error= 1; /* purecov: inspected */
      break;    /* purecov: inspected */
    }

    uchar* payload= packet->payload;
    Gtid_set member_set(&sid_map, NULL);
    Gtid_set intersection_result(&sid_map, NULL);

    if (member_set.add_gtid_encoding(payload, packet->len) != RETURN_STATUS_OK)
    {
      log_message(MY_ERROR_LEVEL, "Error reading GTIDs from the message"); /* purecov: inspected */
      error= 1; /* purecov: inspected */
    }
    else
    {
      /*
        First member set? If so we only need to add it to executed set.
      */
      if (executed_set.is_empty())
      {
        if (executed_set.add_gtid_set(&member_set))
        {
          log_message(MY_ERROR_LEVEL, "Error processing stable transactions set"); /* purecov: inspected */
          error= 1; /* purecov: inspected */
        }
      }
      else
      {
        /*
          We have three sets:
            member_set:          the one sent from a given member;
            executed_set:        the one that contains the intersection of
                                 the computed sets until now;
            intersection_result: the intersection between set and
                                 intersection_result.
          So we compute the intersection between set and executed_set, and
          set that value to executed_set to be used on the next intersection.
        */
        if (member_set.intersection(&executed_set, &intersection_result) != RETURN_STATUS_OK)
        {
          log_message(MY_ERROR_LEVEL, "Error processing intersection of stable transactions set"); /* purecov: inspected */
          error= 1; /* purecov: inspected */
        }
        else
        {
          executed_set.clear();
          if (executed_set.add_gtid_set(&intersection_result) != RETURN_STATUS_OK)
          {
            log_message(MY_ERROR_LEVEL, "Error processing stable transactions set"); /* purecov: inspected */
            error= 1; /* purecov: inspected */
          }
        }
      }
    }

    delete packet;
  }

  if (!error && set_group_stable_transactions_set(&executed_set))
  {
    log_message(MY_ERROR_LEVEL, "Error setting stable transactions set"); /* purecov: inspected */
    error= 1; /* purecov: inspected */
  }

#if !defined(NDEBUG)
  char *executed_set_string;
  executed_set.to_string(&executed_set_string);
  DBUG_PRINT("info", ("Certifier stable_set_handle: executed_set: %s", executed_set_string));
  my_free(executed_set_string);
#endif

  DBUG_RETURN(error);
}

void Certifier::handle_view_change()
{
  DBUG_ENTER("Certifier::handle_view_change");
  clear_incoming();
  clear_members();
  DBUG_VOID_RETURN;
}


void Certifier::get_certification_info(std::map<std::string, std::string> *cert_info)
{
  DBUG_ENTER("Certifier::get_certification_info");
  mysql_mutex_lock(&LOCK_certification_info);

  for(Certification_info::iterator it = certification_info.begin();
      it != certification_info.end(); ++it)
  {
    std::string key= it->first;
    assert(key.compare(GTID_EXTRACTED_NAME) != 0);

    size_t len= it->second->get_encoded_length();
    uchar* buf= (uchar *)my_malloc(
                                   PSI_NOT_INSTRUMENTED,
                                   len, MYF(0));
    it->second->encode(buf);
    std::string value(reinterpret_cast<const char*>(buf), len);
    my_free(buf);

    (*cert_info).insert(std::pair<std::string, std::string>(key, value));
  }

  // Add the group_gtid_executed to certification info sent to joiners.
  size_t len= group_gtid_executed->get_encoded_length();
  uchar* buf= (uchar*) my_malloc(PSI_NOT_INSTRUMENTED, len, MYF(0));
  group_gtid_executed->encode(buf);
  std::string value(reinterpret_cast<const char*>(buf), len);
  my_free(buf);
  (*cert_info).insert(std::pair<std::string, std::string>(
      GTID_EXTRACTED_NAME, value));

  mysql_mutex_unlock(&LOCK_certification_info);
  DBUG_VOID_RETURN;
}

rpl_gno Certifier::generate_view_change_group_gno()
{
  DBUG_ENTER("Certifier::generate_view_change_group_gno");

  mysql_mutex_lock(&LOCK_certification_info);
  rpl_gno result= get_group_next_available_gtid(NULL);

  DBUG_EXECUTE_IF("certifier_assert_next_seqno_equal_5",
                  assert(result == 5););
  DBUG_EXECUTE_IF("certifier_assert_next_seqno_equal_7",
                  assert(result == 7););

  if (result > 0)
    add_to_group_gtid_executed_internal(group_gtid_sid_map_group_sidno, result, false);
  mysql_mutex_unlock(&LOCK_certification_info);

  DBUG_RETURN(result);
}


int Certifier::set_certification_info(std::map<std::string, std::string> *cert_info)
{
  DBUG_ENTER("Certifier::set_certification_info");
  assert(cert_info != NULL);

  if (cert_info->size() == 1) {
    std::map<std::string, std::string>::iterator it =
        cert_info->find(CERTIFICATION_INFO_ERROR_NAME);
    if (it != cert_info->end()) {
      log_message(MY_ERROR_LEVEL,
                  "The certification information could not be set in this server: '%s'",
                  it->second.c_str());
      DBUG_RETURN(1);
    }
  }

  mysql_mutex_lock(&LOCK_certification_info);

  clear_certification_info();
  for(std::map<std::string, std::string>::iterator it = cert_info->begin();
      it != cert_info->end(); ++it)
  {
    std::string key= it->first;

    /*
      Extract the donor group_gtid_executed so that it can be used to
      while member is applying transactions that were already applied
      by distrubuted recovery procedure.
    */
    if (it->first.compare(GTID_EXTRACTED_NAME) == 0)
    {
      if (group_gtid_extracted->add_gtid_encoding(
              reinterpret_cast<const uchar*>(it->second.c_str()), it->second.length())
            != RETURN_STATUS_OK)
      {
        log_message(MY_ERROR_LEVEL,
                    "Error reading group_gtid_extracted from the View_change_log_event"); /* purecov: inspected */
        mysql_mutex_unlock(&LOCK_certification_info); /* purecov: inspected */
        DBUG_RETURN(1); /* purecov: inspected */
      }
      continue;
    }

    Gtid_set_ref *value = new Gtid_set_ref(certification_info_sid_map, -1);
    if (value->add_gtid_encoding(
            reinterpret_cast<const uchar*>(it->second.c_str()), it->second.length())
          != RETURN_STATUS_OK)
    {
      log_message(MY_ERROR_LEVEL,
                  "Error reading the write set item '%s' from the View_change_log_event",
                  key.c_str()); /* purecov: inspected */
      mysql_mutex_unlock(&LOCK_certification_info); /* purecov: inspected */
      DBUG_RETURN(1); /* purecov: inspected */
    }
    value->link();
    certification_info.insert(std::pair<std::string, Gtid_set_ref*>(key, value));
  }

  if (initialize_server_gtid_set())
  {
    log_message(MY_ERROR_LEVEL, "Error during certfication_info"
                " initialization."); /* purecov: inspected */
    mysql_mutex_unlock(&LOCK_certification_info); /* purecov: inspected */
    DBUG_RETURN(1); /* purecov: inspected */
  }

  if (group_gtid_extracted->is_subset_not_equals(group_gtid_executed))
  {
    certifying_already_applied_transactions= true;
    compute_group_available_gtid_intervals();

#ifndef NDEBUG
    char *group_gtid_executed_string= NULL;
    char *group_gtid_extracted_string= NULL;
    group_gtid_executed->to_string(&group_gtid_executed_string, true);
    group_gtid_extracted->to_string(&group_gtid_extracted_string, true);
    DBUG_PRINT("Certifier::set_certification_info()",
               ("Set certifying_already_applied_transactions to true. "
                "group_gtid_executed: \"%s\"; group_gtid_extracted_string: \"%s\"",
                group_gtid_executed_string, group_gtid_extracted_string));
    my_free(group_gtid_executed_string);
    my_free(group_gtid_extracted_string);
#endif
  }

  mysql_mutex_unlock(&LOCK_certification_info);
  DBUG_RETURN(0);
}

void Certifier::update_certified_transaction_count(bool result)
{
  if(result)
    positive_cert++;
  else
    negative_cert++;

  if (local_member_info->get_recovery_status() == Group_member_info::MEMBER_ONLINE)
  {
    applier_module->get_pipeline_stats_member_collector()
        ->increment_transactions_certified();
  }
}

ulonglong Certifier::get_positive_certified()
{
  return positive_cert;
}

ulonglong Certifier::get_negative_certified()
{
  return negative_cert;
}

ulonglong Certifier::get_certification_info_size()
{
  return certification_info.size();
}

void Certifier::get_last_conflict_free_transaction(std::string* value)
{
  int length= 0;
  char buffer[Gtid::MAX_TEXT_LENGTH + 1];

  mysql_mutex_lock(&LOCK_certification_info);
  if (last_conflict_free_transaction.is_empty())
    goto end;

  length= last_conflict_free_transaction.to_string(group_gtid_sid_map, buffer);
  if (length > 0)
    value->assign(buffer);

end:
  mysql_mutex_unlock(&LOCK_certification_info);
}

size_t Certifier::get_members_size()
{
  return members.size();
}

size_t Certifier::get_local_certified_gtid(std::string& local_gtid_certified_string)
{
  if (last_local_gtid.is_empty())
      return 0;

  char buf[Gtid::MAX_TEXT_LENGTH + 1];
  last_local_gtid.to_string(group_gtid_sid_map, buf);
  local_gtid_certified_string.assign(buf);
  return local_gtid_certified_string.size();
}

void Certifier::enable_conflict_detection()
{
  DBUG_ENTER("Certifier::enable_conflict_detection");
  assert(local_member_info->in_primary_mode());

  mysql_mutex_lock(&LOCK_certification_info);
  conflict_detection_enable= true;
  local_member_info->enable_conflict_detection();
  mysql_mutex_unlock(&LOCK_certification_info);
  DBUG_VOID_RETURN;
}

void Certifier::disable_conflict_detection()
{
  DBUG_ENTER("Certifier::disable_conflict_detection");
  assert(local_member_info->in_primary_mode());

  mysql_mutex_lock(&LOCK_certification_info);
  conflict_detection_enable= false;
  local_member_info->disable_conflict_detection();
  mysql_mutex_unlock(&LOCK_certification_info);

  log_message(MY_INFORMATION_LEVEL,
              "Primary had applied all relay logs, disabled conflict "
              "detection");

  DBUG_VOID_RETURN;
}

bool Certifier::is_conflict_detection_enable()
{
  DBUG_ENTER("Certifier::is_conflict_detection_enable");

  mysql_mutex_lock(&LOCK_certification_info);
  bool result= conflict_detection_enable;
  mysql_mutex_unlock(&LOCK_certification_info);

  DBUG_RETURN(result);
}

/*
  Gtid_Executed_Message implementation
 */

Gtid_Executed_Message::Gtid_Executed_Message()
  :Plugin_gcs_message(CT_CERTIFICATION_MESSAGE)
{
}

Gtid_Executed_Message::~Gtid_Executed_Message()
{
}

void
Gtid_Executed_Message::append_gtid_executed(uchar* gtid_data, size_t len)
{
  data.insert(data.end(), gtid_data, gtid_data+len);
}

void
Gtid_Executed_Message::encode_payload(std::vector<unsigned char>* buffer) const
{
  DBUG_ENTER("Gtid_Executed_Message::encode_payload");

  encode_payload_item_type_and_length(buffer, PIT_GTID_EXECUTED, data.size());
  buffer->insert(buffer->end(), data.begin(), data.end());

  DBUG_VOID_RETURN;
}

void
Gtid_Executed_Message::decode_payload(const unsigned char* buffer,
                                      const unsigned char* length)
{
  DBUG_ENTER("Gtid_Executed_Message::decode_payload");
  const unsigned char *slider= buffer;
  uint16 payload_item_type= 0;
  unsigned long long payload_item_length= 0;

  decode_payload_item_type_and_length(&slider,
                                      &payload_item_type,
                                      &payload_item_length);
  data.clear();
  data.insert(data.end(), slider, slider + payload_item_length);

  DBUG_VOID_RETURN;
}