Re: Pointers about replication

2007-03-07 Thread Adrian Bruce

Hi

Multi-master replication is safely possible with MySQL 5.0 when they 
introduced auto_increment_increment and auto_increment_offset 
variables.  Before this it was possible to run into problems with auto 
increment columns generating non-unique numbers between servers.  Try 
the following link for more info:


http://www.onlamp.com/pub/a/onlamp/2006/04/20/advanced-mysql-replication.html

hope this helps

Adrian

Cabbar Duzayak wrote:

Hi All,

Would it be possible to provide some advanced pointers
(articles/books/tutorials/sites) for learning more about replication?

I am particularly interested in master-to-master replication (not even
sure if this is possible with mysql) and/or real-world usage
scenarios/examples as to how much load it can handle, how reliable it
is, etc?

Any help is appreciated.

Thank you.



--
MySQL General Mailing List
For list archives: http://lists.mysql.com/mysql
To unsubscribe:http://lists.mysql.com/[EMAIL PROTECTED]



Re: Pointers about replication

2007-03-07 Thread Marcus Bointon

On 7 Mar 2007, at 06:39, Cabbar Duzayak wrote:


I am particularly interested in master-to-master replication (not even
sure if this is possible with mysql) and/or real-world usage
scenarios/examples as to how much load it can handle, how reliable it
is, etc?


I've had some success with it.

I found these useful:

http://www.onlamp.com/pub/a/onlamp/2006/04/20/advanced-mysql- 
replication.html
http://www.imvu.com/blogs/index.php? 
blog=12title=using_master_master_replication_for_backmore=1c=1tb=1p 
b=1


New 5.1 stuff is mentioned here:
http://firestuff.org/wordpress/2006/01/24/finally-sane-mysql-clustering/

In practice, you do need to tread carefully with circular  
replication. In particular there is no transactional isolation  
between servers, so when doing a transaction, it helps to make sure  
all queries go to the same server within a single transaction. The  
other big thing to look out for is replication latency - as far as I  
can tell, transactions are not replicated until committed, but if you  
have a transaction that takes a few minutes, you will have to wait  
for the same time again after committing for that change to appear on  
the other master. So if you do a write immediately followed by a  
read, the read may be missing what was just written if the read goes  
to a different server. I'm not sure if the new 5.1 stuff helps here.


You should gain the same amount of performance for reads as with a  
normal master-slave setup, but write performance may be slightly  
slower because of the replication overhead, and all masters have to  
execute all queries anyway, so there's no benefit in distributing  
writes (unlike reads).


You do get easy failover - you just point your clients at the other  
server if the default master fails, no messing with reallocating  
masters required. That's probably the biggest win.


To take this a step further, I've been tempted to look at  
Continuent's Sequoia and libMySequoia:

http://sequoia.continuent.org/
http://www.devguy.com/bb/viewtopic.php?p=1874
http://phpthinktank.com/plugin/tag/mysql

Marcus
--
Marcus Bointon
Synchromedia Limited: Creators of http://www.smartmessages.net/
[EMAIL PROTECTED] | http://www.synchromedia.co.uk/



--
MySQL General Mailing List
For list archives: http://lists.mysql.com/mysql
To unsubscribe:http://lists.mysql.com/[EMAIL PROTECTED]



Pointers about replication

2007-03-06 Thread Cabbar Duzayak

Hi All,

Would it be possible to provide some advanced pointers
(articles/books/tutorials/sites) for learning more about replication?

I am particularly interested in master-to-master replication (not even
sure if this is possible with mysql) and/or real-world usage
scenarios/examples as to how much load it can handle, how reliable it
is, etc?

Any help is appreciated.

Thank you.

--
MySQL General Mailing List
For list archives: http://lists.mysql.com/mysql
To unsubscribe:http://lists.mysql.com/[EMAIL PROTECTED]