Posted by s on 03/02/06 07:55

Hi all,

My employer is in the planning phases of a High Availability upgrade
to our server infrastructure. Among other things, we've been taking a
look at MySQL Cluster.

Our primary database servers run MSSQL; the only thing we're currently
storing in MySQL is the PHP session data for our various web
applications. In order to meet the goals of the High Availability
project, we've decided that we no longer want to rely on a single
machine running mysqld. At first glance, MySQL Cluster seems perfect.

However, there are concerns in my department that the management node
in the cluster - the one running ndb_mgmd - would introduce a single
point of failure and negate the benefit of having the cluster. That
can't be correct, but thus far I've been unable to gather enough
information to explain why.

From the documentation and HOWTOs I've read, I understand that while
it is recommended to keep ndb_mgmd running at all times, it is only
_required_ to be running when the cluster's data nodes are started.
This makes me somewhat unclear as to the management node's role in
managing the cluster.

Suppose that the management node goes down (e.g. hardware failure) in
the middle of the night; what happens? How does the cluster retain
state information, stay synchronized, etc?

I am acting under the impression that ndb_mgmd is responsible for
ensuring that the various data nodes are kept synchronized, and for
routing queries from the SQL node to the appropriate data node. Am I
way off base?

Thanks!

--
<s@guerril.la.ape> (to email, remove .ape)
*** Free account sponsored by SecureIX.com ***
*** Encrypt your Internet usage with a free VPN account from http://www.SecureIX.com ***

[Back to original message]