Temp Tables Break MySQL Replication

Question

I have a MySQL Replication Master/Slave set up.

Multiple times a day, Replication breaks as follows

mysql> SHOW SLAVE STATUS\G
*************************** 1. row ***************************
               Slave_IO_State: Waiting for master to send event
                  Master_Host: ....
                  Master_User: repl
                  Master_Port: 3306
                Connect_Retry: 60
              Master_Log_File: bin-log.000012
          Read_Master_Log_Pos: 10458316
               Relay_Log_File: relay-bin.000010
                Relay_Log_Pos: 1528552
        Relay_Master_Log_File: bin-log.000012
             Slave_IO_Running: Yes
            Slave_SQL_Running: No
              Replicate_Do_DB:
          Replicate_Ignore_DB: dba
           Replicate_Do_Table:
       Replicate_Ignore_Table:
      Replicate_Wild_Do_Table:
  Replicate_Wild_Ignore_Table: mydb%.%\_tmp
                   Last_Errno: 1146
                   Last_Error: Error 'Table 'mydb.catalog_category_product_index_tmp' doesn't exist' on query. Default database: 'mydb'. Query: 'DELETE `catalog_category_product_index`  FROM `catalog_category_product_index`
 LEFT JOIN `catalog_category_product_index_tmp` AS `t` ON catalog_category_product_index.category_id = t.category_id AND catalog_category_product_index.store_id = t.store_id AND catalog_category_product_index.product_id = t.product_id WHERE (t.category_id IS NULL) AND (catalog_category_product_index.product_id IN ('69911'))'
                 Skip_Counter: 0
          Exec_Master_Log_Pos: 9644906
              Relay_Log_Space: 3944233
              Until_Condition: None
               Until_Log_File:
                Until_Log_Pos: 0
           Master_SSL_Allowed: No
           Master_SSL_CA_File:
           Master_SSL_CA_Path:
              Master_SSL_Cert:
            Master_SSL_Cipher:
               Master_SSL_Key:
        Seconds_Behind_Master: NULL
Master_SSL_Verify_Server_Cert: No
                Last_IO_Errno: 0
                Last_IO_Error:
               Last_SQL_Errno: 1146
               Last_SQL_Error: Error 'Table 'mydb.catalog_category_product_index_tmp' doesn't exist' on query. Default database: 'mydb'. Query: 'DELETE `catalog_category_product_index`  FROM `catalog_category_product_index`
 LEFT JOIN `catalog_category_product_index_tmp` AS `t` ON catalog_category_product_index.category_id = t.category_id AND catalog_category_product_index.store_id = t.store_id AND catalog_category_product_index.product_id = t.product_id WHERE (t.category_id IS NULL) AND (catalog_category_product_index.product_id IN ('69911'))'
  Replicate_Ignore_Server_Ids:
             Master_Server_Id: 1000
1 row in set (0.00 sec)

I simply run

STOP SLAVE;
SET GLOBAL SQL_SLAVE_SKIP_COUNTER=1;
START SLAVE;

and Replication comes back to normal

The binary logs on the Master have events that say

• DROP TABLE catalog_category_product_index_tmp;
• CREATE catalog_category_product_index_tmp (...);
• Inserts into catalog_category_product_index_tmp

My client uses binlog_format = MIXED on Master and Slave. When DROP TABLE and CREATE TABLE are executed, MySQL shifts to binlog_format = STATEMENT to execute the DDL.

This replication issue of the temp table not existing on the Slave happens to all tables that end in tmp.

mysql> select table_name,engine from information_schema.tables where table_schema='mydb' and table_name like '%tmp';
+---------------------------------------------+--------+
| table_name                                  | engine |
+---------------------------------------------+--------+
| catalog_category_anc_categs_index_tmp       | MEMORY |
| catalog_category_anc_products_index_tmp     | MEMORY |
| catalog_category_product_index_enbl_tmp     | MEMORY |
| catalog_product_index_eav_decimal_tmp       | MEMORY |
| catalog_product_index_eav_tmp               | MEMORY |
| catalog_product_index_price_bundle_opt_tmp  | MEMORY |
| catalog_product_index_price_bundle_sel_tmp  | MEMORY |
| catalog_product_index_price_bundle_tmp      | MEMORY |
| catalog_product_index_price_cfg_opt_agr_tmp | MEMORY |
| catalog_product_index_price_cfg_opt_tmp     | MEMORY |
| catalog_product_index_price_downlod_tmp     | MEMORY |
| catalog_product_index_price_final_tmp       | MEMORY |
| catalog_product_index_price_opt_agr_tmp     | MEMORY |
| catalog_product_index_price_opt_tmp         | MEMORY |
| catalog_product_index_price_tmp             | MEMORY |
| cataloginventory_stock_status_tmp           | MEMORY |
+---------------------------------------------+--------+

QUESTIONS

• Is there any known issue with MySQL Replication and Magento ?
• Could this be an issue with Magento ?
• Any other ideas as to why ?

Answer 1

I finally found my problem.

This line in the SHOW SLAVE STATUS\G

Replicate_Wild_Ignore_Table: mydb%.%\_tmp

was actually preventing the DB user from dropping and creating the MEMORY tables. I made a change to puppet to remove Replicate_Wild_Ignore_Table from my.cnf and everything was back to normal.