Uploaded image for project: 'MariaDB Server'
  1. MariaDB Server
  2. MDEV-20087

Galera + SET SQL_LOG_BIN=0 on binlog on others nodes

    XMLWordPrintable

    Details

      Description

      On node 1 :

      MariaDB [(none)]> set sql_log_bin=0;
      Query OK, 0 rows affected (0.000 sec)
       
      MariaDB [(none)]> create database gg;
      Query OK, 1 row affected (0.002 sec)
      MariaDB [(none)]> use gg
      Database changed
      MariaDB [gg]> create table gg2 (a int(11));
      Query OK, 0 rows affected (0.030 sec)
      MariaDB [gg]> insert into gg2 values(1);
      Query OK, 1 row affected (0.001 sec)
       
      MariaDB [gg]> select * from gg2;
      +------+
      | a    |
      +------+
      |    1 |
      +------+
      1 row in set (0.000 sec)
      

      On node 2 :

      MariaDB [gg]> select * from gg2;
      +------+
      | a    |
      +------+
      |    1 |
      +------+
      1 row in set (0.000 sec)
      

      Ok that's normal behavior !

      But if I look in binlog I got this :

      On node 1 :

      mysqlbinlog --start-datetime="2019-07-18 11:50:00" --base64-output=decode-rows mariadb-bin.000027 > binlog.search
      cat binlog.search | grep gg
      (empty)
      

      That's normal behavior too

      On node 2 :

      mysqlbinlog --start-datetime="2019-07-18 11:50:00" --base64-output=decode-rows mariadb-bin.000031 > binlog.search
      cat binlog.search | grep gg
      create database gg
      use `gg`/*!*/;
      create table gg2 (a int(11))
      #Q> insert into gg2 values(1)
      #190718 11:54:43 server id 473657853  end_log_pos 605411511 CRC32 0x4da5b0f0 	Table_map: `gg`.`gg2` mapped to number 673301494
      

      WTF ? these lines should be not there.

      As i Remember I pushed a bug for this in 10.0 (or similar), and it's happen again ! (could you please add this in test case the same with 'SET @@skip_replication = ON;'.

      My goal was to load a database (I developed my own script to load) in a galera cluster of production (with other databases) but node 2 crashed because the size of binlog on other node took all disk space.

        Attachments

          Issue Links

            Activity

              People

              Assignee:
              jplindst Jan Lindström
              Reporter:
              Aurelien_LEQUOY Aurélien LEQUOY
              Votes:
              2 Vote for this issue
              Watchers:
              4 Start watching this issue

                Dates

                Created:
                Updated: