Hi Jason!
The patch looks good, but I would suggest to add a call to start_bulk_delete()
and start_bulk_update() at the appropriate places. They don't take an argument,
so there is nothing to compute for that.
Just my few cents,
Mats Kindahl
He Zhenxing wrote:
> #At file:///media/sda3/work/mysql/bzrwork/b35843/5.1-rpl/
>
> 2666 He Zhenxing 2008-09-10
> BUG#35843 Slow replication slave when using partitioned myisam table
>
> In order to improve the performance when replicating to partitioned
> myisam tables with row-based format, the number of rows of current
> rows log event is estimated and used to setup storage engine for bulk
> inserts.
> modified:
> sql/log_event.cc
>
> === modified file 'sql/log_event.cc'
> --- a/sql/log_event.cc 2008-08-06 10:41:27 +0000
> +++ b/sql/log_event.cc 2008-09-10 09:13:58 +0000
> @@ -7488,7 +7488,6 @@ Write_rows_log_event::do_before_row_oper
> */
> }
>
> - m_table->file->ha_start_bulk_insert(0);
> /*
> We need TIMESTAMP_NO_AUTO_SET otherwise ha_write_row() will not use fill
> any TIMESTAMP column with data from the row but instead will use
> @@ -7627,7 +7626,16 @@ Rows_log_event::write_row(const Relay_lo
>
> /* unpack row into table->record[0] */
> error= unpack_current_row(rli); // TODO: how to handle errors?
> -
> + if (m_curr_row == m_rows_buf)
> + {
> + /* this is the first row to be inserted, we estimate the rows with
> + the size of the first row and use that value to initialize
> + storage engine for bulk insertion */
> + ulong estimated_rows= (m_rows_end - m_curr_row) / (m_curr_row_end -
> m_curr_row);
> + m_table->file->ha_start_bulk_insert(estimated_rows);
> + }
> +
> +
> #ifndef DBUG_OFF
> DBUG_DUMP("record[0]", table->record[0], table->s->reclength);
> DBUG_PRINT_BITSET("debug", "write_set = %s", table->write_set);
>
>
--
Mats Kindahl
Lead Software Developer
Replication Team
MySQL AB, www.mysql.com