RE: Cluster without Replica - Jeff Sturm - com.mysql.lists.cluster

4 messages in com.mysql.lists.clusterRE: Cluster without Replica

From	Sent On	Attachments
Hayden Livingston	16 Feb 2008 03:31
Massimo	16 Feb 2008 03:51
Hayden Livingston	16 Feb 2008 03:57
Jeff Sturm	16 Feb 2008 06:32

Subject:	RE: Cluster without Replica
From:	Jeff Sturm (jeff...@eprize.com)
Date:	02/16/2008 06:32:03 AM
List:	com.mysql.lists.cluster

Master<->master replication may be a possiblity for you. You can configure each of two hosts to mutually slave each other's transactions, or it is even possible to configure 3 or more nodes into circular replication.

It is not completely safe for failover when the importance of transaction integrity is high, because replication is not synchronous with standard binlogs. It is possible to process the same transaction on two separate nodes, which will conflict on replication. Since you are using MyISAM which lacks transactional integrity, this may not be important to you.

If you go with master<->master replication, you'll want to strongly consider upgrading to 5.x so you can configure offsetting auto increment counters.

-----Original Message----- From: Hayden Livingston [mailto:hali...@gmail.com] Sent: Saturday, February 16, 2008 6:57 AM To: Massimo Cc: clus...@lists.mysql.com Subject: Re: Cluster without Replica

Actually, we tried Replication. It's not what we need. We would like node2 to be also able to add to the database, that then gets recorded into the master. So unless there is sort of "two way replication" that I don't know about, then I think it's clustering, but maybe there is a two way method. Is there?

On Feb 16, 2008 5:51 AM, Massimo <mass...@mysql.com> wrote:

Hi Hayden,

in your situation you don't need MySQL Cluster: is designed for near reall-time applications with requirements of very low latency, high availability and scalability.

What you need probably is the MySQL Replication: an asynchronous replication from master (your node1) to the slave (your node 2).

I suggest to upgrade your MySQL Server to version 5.0 and set up the replication: is easy and you will find all the infos in the replication chapter on the manual.

If you have some trouble setting up your replication, you can write at

the repl...@lists.mysql.com mailing list.

Massimo

Hayden Livingston ha scritto:

The situation:

100GB of MyISAM data 3 physical machines (node1, node2, node3) MySQL 4.1.20

right now, all our queries goto one machine, node1, which is a webserver, as well among other things. We basically want to have an exact replica of node1, so that say node2 can be accessed. Most of our code is mysql_connect("localhost" ...);

So for node1 it goes to node1's mysql and for node2 it goes to node2's mysql.

We really don't have any need for load balanacing / high availabilty, just a live replica. Generally if we go down, it's almost always because of some other problem that takes the entire machine down in some way.

So, my problem is I'm confused about 4.1 Clustering, the manual says

it's all in memory so I need 200GB of memory? Secondly, what's the difference between SQL nodes and data nodes? I would have thought they're the same, since we're hoping to run mysql and ndbd on node1 and node2, is this not possible?

Thanks a bunch!