[MDEV-5852] MariaDB 10.0 slave SQL thread slower than MariaDB 5.5 Created: 2014-03-13  Updated: 2022-11-01  Resolved: 2022-11-01

Status: Closed
Project: MariaDB Server
Component/s: Replication
Affects Version/s: 10.0.9
Fix Version/s: N/A

Type: Bug Priority: Major
Reporter: Kolbe Kegel (Inactive) Assignee: Kristian Nielsen
Resolution: Won't Fix Votes: 0
Labels: None
Environment:

CentOS 6.5 (tests done on EC2 m1.xlarge instance).


 Description   

MariaDB 10.0.9 is slower to execute a relay long than MariaDB 5.5. My tests were performed by performing millions of autocommitted single-statement inserts. I took the binary log from a MySQL 5.5.36 instance and used that as the relay log for a variety of different versions of various products.

CREATE TABLE test.`t1` (
 
  `id` bigint(20) unsigned NOT NULL AUTO_INCREMENT,
 
  `pid` bigint(20) unsigned DEFAULT NULL,
 
  `iter` bigint(20) unsigned DEFAULT NULL,
 
  `v` bigint(20) DEFAULT NULL,
 
  PRIMARY KEY (`id`)
 
) ENGINE=InnoDB;

I tested using both ROW and MIXED binlog formats, and I found that the MariaDB 10.0 instance was 10-20% slower than the MariaDB 5.5 instance.

I see approximately 9500 inserts/s for MariaDB 5.5.36 and approximately 8000 inserts/s for MariaDB 10.0.9.

The binary logs I've used can be found at ftp://ftp.skysql.com/downloads/binlogs_for_slave_comparison.tgz

I've started MariaDB like this for both versions:

mysqld_safe --no-defaults --datadir="$PWD/data/" --socket="$PWD/data/mysql.sock" --server-id=1 --log-bin --binlog-format=mixed --innodb-flush-log-at-trx-commit=2 --sync-binlog=0 --innodb-buffer-pool-size=12GB --innodb-log-file-size=2000M --skip-name-resolve --relay-log=relay --replicate-same-server-id --relay-log-purge=0 --skip-slave-start &

I place the binary log I want to use for testing in the datadir as relay.000001 and start replication like this:

stop slave; truncate t1; change master to master_host='null', master_user='null', relay_log_file='relay.000001', relay_log_pos=4; start slave sql_thread; select sleep(1); show slave status\G



 Comments   
Comment by Sergei Golubchik [ 2022-11-01 ]

10.0 has reached EOL long time ago

Generated at Thu Feb 08 07:07:32 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.