Skip to content
  • Mauritz Sundell's avatar
    78036f10
    WL#7774 Make MTR sync_slave_sql_with_master.inc handle NDB · 78036f10
    Mauritz Sundell authored
      Make source include/sync_slave_sql_with_master.inc work as a
      replacement to sync_slave_with_master for NDB in MTR tests.
    
      The usage of mysqltest command sync_slave_with_master in MTR
      tests should in time be replaced.
    
      The replacement source include/sync_slave_sql_with_master.inc
      did not work in exactly the same way for NDB.
    
      In WL#7205 some uses of sync_slave_with_master was replaced in
      some rpl tests.  These changes make the corresponding NDB test
      to fail.
    
      This patch make changes needed for sync_slave_sql_with_master.inc
      to work with NDB.
    
      Committed transactions using the NDB engine are not directly binlogged
      but first packed in epochs.  This introduce an extra delay between
      commit and binlogging that must be taken into account when one wants
      to make sure a slave have replayed the committed transactions.
    
      In MySQL Cluster sync_slave_with_master is patched to make sure that
      current epoch is binlogged before taking master binlog position.
      See Bug#18375920.
    78036f10
    WL#7774 Make MTR sync_slave_sql_with_master.inc handle NDB
    Mauritz Sundell authored
      Make source include/sync_slave_sql_with_master.inc work as a
      replacement to sync_slave_with_master for NDB in MTR tests.
    
      The usage of mysqltest command sync_slave_with_master in MTR
      tests should in time be replaced.
    
      The replacement source include/sync_slave_sql_with_master.inc
      did not work in exactly the same way for NDB.
    
      In WL#7205 some uses of sync_slave_with_master was replaced in
      some rpl tests.  These changes make the corresponding NDB test
      to fail.
    
      This patch make changes needed for sync_slave_sql_with_master.inc
      to work with NDB.
    
      Committed transactions using the NDB engine are not directly binlogged
      but first packed in epochs.  This introduce an extra delay between
      commit and binlogging that must be taken into account when one wants
      to make sure a slave have replayed the committed transactions.
    
      In MySQL Cluster sync_slave_with_master is patched to make sure that
      current epoch is binlogged before taking master binlog position.
      See Bug#18375920.
Loading