Hi,
Jonas Ibsen Wrote:
-------------------------------------------------------
> Hi
>
> I have read most of the MySQL cluster
> documentation Chapter 17 but a few things is not
> clear to me.
>
> 1.
> This is a follow up to the following thread
>
>
http://forums.mysql.com/read.php?25,65435,65435#ms
> g-65435
>
> In the answer it is mentioned that having only one
> SQL node is a single point of failure, which i
> understand... Then i am wondering when you have
> two or more SQL nodes what IP address should your
> application contact to be sure to get a response?
> Or is it your own applications responsibility to
> check which SQL nodes are avalible before
> submitting the query?
>
As mentioned by Andrew you can use the JDBC driver and it will
ensure that you use a working one and failover otherwise. For
other MySQL API's you'll have to do it yourself. It's on the
roadmap to fix those as well but not done yet.
> 2.
> In the forum i have read that it is the SQL node
> that need CPU power the most. And thereby i assume
> that the SQL node in case of a select only gets
> the data from the data nodes and then perform the
> query on the data.
> So that made me wonder if two data nodes have the
> exact same data, which they do, would it not be
> the best aproache in case of a select statement to
> pass the statement on to one of the two data nodes
> and wait for the response of the statement, like a
> load balancer?
>
Only the MySQL Server has the logic to handle SQL statements.
The data nodes have a storage manager API so they can perform
certain parts of a query like conditions on a table and so
forth. As we move along it is likely that more and more
functionality can be pushed down to the data nodes.
Rgrds Mikael
> I hope you guys understand my questions :)
>
> Kind Regards
> Jonas
Mikael Ronstrom
Senior Software Architect, MySQL AB
My blog:
http://mikaelronstrom.blogspot.com