Mysql – Clarification on MySQL innodb_flush_method variable

innodbMySQL

Let me begin by admitting that I'm very ignorant of the inner workings of hard disks. So when I read over the manual for the variable innodb_flush_method, it confused me. Can I get an explanation in layman's terms on the difference in O_DSYNC and O_DIRECT, and how to know if it's a performance issue on a database server.

Some stats on my setup: Mac OSX 10.6 (32-bit kernel, since architecture is out of date) running MySQL 5.1.49-64bit (hoping it would allow me to use the memory). 8GB RAM, ~6GB of innodb data/indexes.

Best Answer

Here is an explanation on how fdatasync() works vs how fsync() works

fdatasync() flushes all data buffers of a file to disk (before the system call returns). It resembles fsync() but is not required to update the metadata, such as access time. Applications that access databases or log files often write a tiny data fragment (e.g., one line in a log file) and then call fsync() immediately in order to ensure that the written data is physically stored on the harddisk. Unfortunately, fsync() will always initiate two write operations

one write operation for the newly written data
one write operation in order to update the modification time stored in the inode

If the modification time is not a part of the transaction concept, then fdatasync() can be used to avoid unnecessary inode disk write operations.

In English, O_DSYNC is faster than O_DIRECT since O_DIRECT calls fsync() twice (one for logs and one for data) and fsync() verifies data writes via two write operations. Using O_DSYNC calls fdatsync() and fsync(). You can think of fdatasync() as doing an asynchronous fsync() (not verfying data).

Looking at the numbers, O_DSYNC does four write ops, two of which are verified, while fsync() does four write operations, all being verfied afterwards.

CONCLUSION

O_DSYNC
- faster than O_DIRECT
- Data may/may not be consistent due to latency or an outright crash
O_DIRECT
- more stable
- data consistent
- naturally slower

I hope this answer helps, and I hope I didn't make things worse for you.

UPDATE 2011-10-20 16:40 EDT

It cleanly page out all data in the InnoDB Buffer Pool prior to redoing the Log Files, you should set this option about 1 hour before shutdown:

SET GLOBAL innodb_max_dirty_pages_pct = 0;

By default, innodb_max_dirty_pages_pct is 75 (MySQL 5.5+) or 90 (prior to MySQL 5.5). Setting this to zero keeps the number of dirty pages under 1% of the InnoDB Buffer Pool. Performing service mysql stop does this anyway. In addition, a shutdown will finish up any remaining items in the redo log. To keep to this option just add it to /etc/my.cnf:

[mysqld]
innodb_max_dirty_pages_pct = 0

UPDATE 2013-04-19 16:16 EDT

I updated my answer a little more with innodb_fast_shutdown because I used to restart mysql and stop mysql to do this. Now, this one-step is vital because every transaction uncommitted may have other moving parts within and outside of the InnoDB Transaction Logs (See InnoDB Infrastructure).

Please note that setting innodb_fast_shutdown to 2 would clean the logs out as well but more moving parts still exist and gets picked on Crash Recovery during mysqld's startup. Setting of 0 is best.

Mysql – Clarification about master slave configuration in thesql

Yes, it will overwrite the data on the Slave. However, you can setup replication to pick up from the point-in-time of the dump and make it roll all changes since the dump was loaded on the Slave

For this Example, let's assume

IP of the Master is 10.1.1.20
IP of the Slave is 10.1.1.30

Here is what you do

STEP01 : Activate Binary Logging on the Old Server

Step01-a) Add this to /etc/my.cnf on the Master

[mysqld]
server-id=101120
log-bin=mysql-bin

Step01-b) # service mysql restart

After STEP01, you should see mysql-bin.000001 and mysql-bin.index in /var/lib/mysql

STEP02 : Perform mysqldump

On the Master, you can mysqldump the data and record from what point in time it happened.

# service mysql restart --skip-networking --skip-grants
# mysqldump --single-transaction --master-data=2 --all-databases --routines --triggers > MySQLData.sql
# service mysql restart

What this does is record the Master Log File and Position the moment the mysqldump started as a comment. You can visibly see it when you view line 22:

# head -22 MySQLData.sql | tail -1

STEP03 : Load the mysqldump into the Slave

Execute the mysql client loading the mysqldump into the Slave's mysql instance

# mysql -h10.1.1.30 -uroot -p < MySQLData.sql

STEP04 : Create MySQL Replication User on the Master

# mysql -uroot -p -e"GRANT REPLICATION SLAVE, REPLICATION CLIENT ON *.* TO 'replicator'@'10.64.51.%' IDENTIFIED BY 'replpassword'"

STEP05 : Setup the Slave with a Separate Server ID

Add this to /etc/my.cnf on the Slave

[mysqld]
server-id=101130

and # service mysql restart

STEP06 : Setup the Replication on the Slave

Goto the mysql client and run the following command

mysql> CHANGE MASTER TO
    MASTER_HOST='10.1.1.20',
    MASTER_PORT=3306,
    MASTER_USER='replicator',
    MASTER_PASSWORD='replpassword',
    MASTER_LOG_FILE='mysql-bin.000001',
    MASTER_LOG_POS=1;

STEP07 : Setup the point-in-time Master Log and Position

Back in STEP02, I mentioned viewing the point-in-time position using

# head -22 MySQLData.sql | tail -1

You should see something like this:

#CHANGE MASTER TO MASTER_LOG_FILE='mysql-bin.000015', MASTER_LOG_POS=122957100;

Run it as command in the mysql client on the Slave

mysql> CHANGE MASTER TO MASTER_LOG_FILE='mysql-bin.000015', MASTER_LOG_POS=122957100;

Afterwards, run this command

mysql> SHOW SLAVE STATUS\G

You should see something like this:

*************************** 1. row ***************************
               Slave_IO_State: Waiting for master to send event
                  Master_Host: 10.64.113.232
                  Master_User: replicant
                  Master_Port: 3306
                Connect_Retry: 60
              Master_Log_File: mysql-bin.000015
          Read_Master_Log_Pos: 122957100
               Relay_Log_File: relay-bin.003666
                Relay_Log_Pos: 122957100
        Relay_Master_Log_File: mysql-bin.000015
             Slave_IO_Running: No
            Slave_SQL_Running: No
              Replicate_Do_DB:
          Replicate_Ignore_DB:
           Replicate_Do_Table:
       Replicate_Ignore_Table:
      Replicate_Wild_Do_Table:
  Replicate_Wild_Ignore_Table:
                   Last_Errno: 0
                   Last_Error:
                 Skip_Counter: 0
          Exec_Master_Log_Pos: 106
              Relay_Log_Space: 106
              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: 0
Master_SSL_Verify_Server_Cert: No
                Last_IO_Errno: 0
                Last_IO_Error:
               Last_SQL_Errno: 0
               Last_SQL_Error:
1 row in set (0.00 sec)

STEP08 : Launch Replication

Start up replication with this:

mysql> START SLAVE;

Afterwards, run this command again

mysql> SHOW SLAVE STATUS\G

If you see this

             Slave_IO_Running: Yes
            Slave_SQL_Running: Yes

CONGRATULATIONS, MySQL Replication is Working !!!