percona-server-5.7.36-39/sql/rpl_trx_tracking.cc

/* Copyright (c) 2018, 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 "rpl_trx_tracking.h"

#include "mysqld.h"
#include "binlog.h"

#include "debug_sync.h"

Logical_clock::Logical_clock()
  : state(SEQ_UNINIT), offset(0)
{}

/**
  Atomically fetch the current state.
  @return  not subtracted "absolute" value.
 */
inline int64 Logical_clock::get_timestamp()
{
  int64 retval= 0;
  DBUG_ENTER("Logical_clock::get_timestamp");
  retval= my_atomic_load64(&state);
  DBUG_RETURN(retval);
}

/**
  Steps the absolute value of the clock (state) to return
  an updated value.
  The caller must be sure to call the method in no concurrent
  execution context so either offset and state can't change.

  @return  incremented "absolute" value
 */
inline int64 Logical_clock::step()
{
  assert(SEQ_UNINIT == 0);
  DBUG_EXECUTE_IF("logical_clock_step_2", ++state;);
  return ++state;
}

/**
  To try setting the clock *forward*.
  The clock does not change when the new value is in the past
  which is reflected by the new value and by offset.
  In other words the function main effects is described as
    state= max(state, new_value).
  Offset that exceeds the new value indicates the binary log rotation
  to render such new value useless.

  @param  new_val  a new value (offset included)
  @return a (new) value of state member regardless whether it's changed or not.
 */
inline int64 Logical_clock::set_if_greater(int64 new_val)
{
  longlong old_val= new_val - 1;
  bool cas_rc;

  DBUG_ENTER("Logical_clock::set_if_greater");

  assert(new_val > 0);

  if (new_val <= offset)
  {
    /*
      This function's invocation can be separated from the
      transaction's flushing by few rotations. A late to log
      transaction does not change the clock, similarly to how
      its timestamps are handled at flushing.
    */
    DBUG_RETURN(SEQ_UNINIT);
  }

  assert(new_val > 0);

  while (!(cas_rc= my_atomic_cas64(&state, &old_val, new_val)) &&
         old_val < new_val)
  {}

  assert(state >= new_val); // setting can't be done to past

  assert(cas_rc || old_val >= new_val);

  DBUG_RETURN(cas_rc ? new_val : old_val);
}


/**
  Get the sequence_number for a transaction, and get the last_commit based
  on parallel committing transactions.

  @param[in]     thd             Current THD from which to extract trx context.
  @param[in,out] sequence_number Sequence number of current transaction.
  @param[in,out] commit_parent   Commit_parent of current transaction,
                                 pre-filled with the commit_parent calculated
                                 by the logical clock logic.
*/
void
Commit_order_trx_dependency_tracker::get_dependency(THD *thd,
                                                    int64 &sequence_number,
                                                    int64 &commit_parent)
{
  Transaction_ctx *trn_ctx= thd->get_transaction();

  assert(trn_ctx->sequence_number
         > m_max_committed_transaction.get_offset());
  /*
    Prepare sequence_number and commit_parent relative to the current
    binlog.  This is done by subtracting the binlog's clock offset
    from the values.

    A transaction that commits after the binlog is rotated, can have a
    commit parent in the previous binlog. In this case, subtracting
    the offset from the sequence number results in a negative
    number. The commit parent dependency gets lost in such
    case. Therefore, we log the value SEQ_UNINIT in this case.
  */
  sequence_number=
    trn_ctx->sequence_number - m_max_committed_transaction.get_offset();

  commit_parent=
    trn_ctx->last_committed <= m_max_committed_transaction.get_offset()
           ? SEQ_UNINIT
           : trn_ctx->last_committed - m_max_committed_transaction.get_offset();
}

int64
Commit_order_trx_dependency_tracker::step()
{
  return m_transaction_counter.step();
}

void
Commit_order_trx_dependency_tracker::rotate()
{
  m_max_committed_transaction.
    update_offset(m_transaction_counter.get_timestamp());

  m_transaction_counter.
    update_offset(m_transaction_counter.get_timestamp());
}

void
Commit_order_trx_dependency_tracker::update_max_committed(int64 sequence_number)
{
  mysql_mutex_assert_owner(&LOCK_slave_trans_dep_tracker);
  m_max_committed_transaction.set_if_greater(sequence_number);
}

/**
  Get the writeset dependencies of a transaction.
  This takes the commit_parent that must be previously set using
  Commit_order_trx_dependency_tracker and tries to make the commit_parent as
  low as possible, using the writesets of each transaction.
  The commit_parent returned depends on how many row hashes are stored in the
  writeset_history, which is cleared once it reaches the user-defined maximum.

  @param[in]     thd             Current THD from which to extract trx context.
  @param[in,out] sequence_number Sequence number of current transaction.
  @param[in,out] commit_parent   Commit_parent of current transaction,
                                 pre-filled with the commit_parent calculated by
                                 Commit_order_trx_dependency_tracker to use when
                                 the writeset commit_parent is not valid.
*/
void
Writeset_trx_dependency_tracker::get_dependency(THD *thd,
                                                int64 &sequence_number,
                                                int64 &commit_parent)
{
  Rpl_transaction_write_set_ctx *write_set_ctx=
    thd->get_transaction()->get_transaction_write_set_ctx();
  std::set<uint64> *writeset= write_set_ctx->get_write_set();

#ifndef NDEBUG
  /* The writeset of an empty transaction must be empty. */
  if (is_empty_transaction_in_binlog_cache(thd))
    assert(writeset->size() == 0);
#endif

  /*
    Check if this transaction has a writeset, if the writeset will overflow the
    history size, if the transaction_write_set_extraction is consistent
    between session and global or if changes in the tables referenced in this
    transaction cascade to other tables. If that happens revert to using the
    COMMIT_ORDER and clear the history to keep data consistent.
  */
  bool can_use_writesets=
    // empty writeset implies DDL or similar, except if there are missing keys
    (writeset->size() != 0 || write_set_ctx->get_has_missing_keys() ||
     /*
       The empty transactions do not need to clear the writeset history, since
       they can be executed in parallel.
     */
     is_empty_transaction_in_binlog_cache(thd)) &&
    // hashing algorithm for the session must be the same as used by other rows in history
    (global_system_variables.transaction_write_set_extraction ==
     thd->variables.transaction_write_set_extraction) &&
    // must not use foreign keys
    !write_set_ctx->get_has_related_foreign_keys() &&
    // it did not broke past the capacity already
    !write_set_ctx->was_write_set_limit_reached();
  bool exceeds_capacity= false;

  mysql_mutex_lock(&LOCK_slave_trans_dep_tracker);
  if (can_use_writesets)
  {
    /*
     Check if adding this transaction exceeds the capacity of the writeset
     history. If that happens, m_writeset_history will be cleared only after
     using its information for current transaction.
    */
    exceeds_capacity=
      m_writeset_history.size() + writeset->size() > get_opt_max_history_size();

    /*
     Compute the greatest sequence_number among all conflicts and add the
     transaction's row hashes to the history.
    */
    DEBUG_SYNC(thd, "wait_in_get_dependency");
    int64 last_parent= m_writeset_history_start;
    for (std::set<uint64>::iterator it= writeset->begin();
         it != writeset->end(); ++it)
    {
      Writeset_history::iterator hst= m_writeset_history.find(*it);
      if (hst != m_writeset_history.end())
      {
        if (hst->second > last_parent && hst->second < sequence_number)
          last_parent= hst->second;

        hst->second= sequence_number;
      }
      else
      {
        if (!exceeds_capacity)
          m_writeset_history.insert(std::pair<uint64, int64>(*it, sequence_number));
      }
    }

    /*
      If the transaction references tables with missing primary keys revert to
      COMMIT_ORDER, update and not reset history, as it is unnecessary because
      any transaction that refers this table will also revert to COMMIT_ORDER.
    */
    if (!write_set_ctx->get_has_missing_keys())
    {
      /*
       The WRITESET commit_parent then becomes the minimum of largest parent
       found using the hashes of the row touched by the transaction and the
       commit parent calculated with COMMIT_ORDER.
      */
      commit_parent= std::min(last_parent, commit_parent);
    }
  }

  if (exceeds_capacity || !can_use_writesets)
  {
    m_writeset_history_start= sequence_number;
    m_writeset_history.clear();
  }
  mysql_mutex_unlock(&LOCK_slave_trans_dep_tracker);
}

void
Writeset_trx_dependency_tracker::rotate(int64 start)
{
  m_writeset_history_start= start;
  m_writeset_history.clear();
}

/**
  Get the writeset commit parent of transactions using the session dependencies.

  @param[in]     thd             Current THD from which to extract trx context.
  @param[in,out] sequence_number Sequence number of current transaction.
  @param[in,out] commit_parent   Commit_parent of current transaction,
                                 pre-filled with the commit_parent calculated
                                 by the Write_set_trx_dependency_tracker as a
                                 fall-back.
*/
void
Writeset_session_trx_dependency_tracker::get_dependency(THD *thd,
                                                        int64 &sequence_number,
                                                        int64 &commit_parent)
{
  int64 session_parent= thd->rpl_thd_ctx.dependency_tracker_ctx().
                        get_last_session_sequence_number();

  if (session_parent != 0 && session_parent < sequence_number)
    commit_parent= std::max(commit_parent, session_parent);

  thd->rpl_thd_ctx.dependency_tracker_ctx().
    set_last_session_sequence_number(sequence_number);
}

/**
  Get the dependencies in a transaction, the main entry point for the
  dependency tracking work.
*/
void
Transaction_dependency_tracker::get_dependency(THD *thd,
                                               int64 &sequence_number,
                                               int64 &commit_parent)
{
  sequence_number= commit_parent= 0;

  switch(my_atomic_load64(&m_opt_tracking_mode))
  {
    case DEPENDENCY_TRACKING_COMMIT_ORDER:
      m_commit_order.get_dependency(thd, sequence_number, commit_parent);
      break;
    case DEPENDENCY_TRACKING_WRITESET:
      m_commit_order.get_dependency(thd, sequence_number, commit_parent);
      m_writeset.get_dependency(thd, sequence_number, commit_parent);
      break;
    case DEPENDENCY_TRACKING_WRITESET_SESSION:
      m_commit_order.get_dependency(thd, sequence_number, commit_parent);
      m_writeset.get_dependency(thd, sequence_number, commit_parent);
      m_writeset_session.get_dependency(thd, sequence_number, commit_parent);
      break;
    default:
      assert(0);  // blow up on debug
      /*
        Fallback to commit order on production builds.
       */
      m_commit_order.get_dependency(thd, sequence_number, commit_parent);
  }
}

void
Transaction_dependency_tracker::tracking_mode_changed()
{
  Logical_clock max_committed_transaction=
    m_commit_order.get_max_committed_transaction();
  int64 timestamp= max_committed_transaction.get_timestamp()
                   - max_committed_transaction.get_offset();

  m_writeset.rotate(timestamp);
}

/**
  The method is to be executed right before committing time.
  It must be invoked even if the transaction does not commit
  to engine being merely logged into the binary log.
  max_committed_transaction is updated with a greater timestamp
  value.
  As a side effect, the transaction context's sequence_number
  is reset.

  @param thd a pointer to THD instance
*/
void
Transaction_dependency_tracker::update_max_committed(THD *thd)
{
  Transaction_ctx *trn_ctx= thd->get_transaction();
  m_commit_order.update_max_committed(trn_ctx->sequence_number);
  /*
    sequence_number timestamp isn't needed anymore, so it's cleared off.
  */
  trn_ctx->sequence_number= SEQ_UNINIT;

  assert(trn_ctx->last_committed == SEQ_UNINIT ||
         thd->commit_error == THD::CE_FLUSH_ERROR ||
         thd->commit_error == THD::CE_FLUSH_GNO_EXHAUSTED_ERROR);
}

int64
Transaction_dependency_tracker::step()
{
  return m_commit_order.step();
}

void
Transaction_dependency_tracker::rotate()
{
  m_commit_order.rotate();
  /*
    To make slave appliers be able to execute transactions in parallel
    after rotation, set the minimum commit_parent to 1 after rotation.
  */
  m_writeset.rotate(1);
  if (current_thd)
    current_thd->get_transaction()->sequence_number= 2;
}

int64 Transaction_dependency_tracker::get_max_committed_timestamp()
{
  return m_commit_order.get_max_committed_transaction().get_timestamp();
}