This chapter provides a overview of MySQL Cluster Manager, as well as its architecture, purpose, and capabilities.
This section provides definitions of key terms used to describe MySQL Cluster Manager and its components in this manual and in other documentation relating to MySQL Cluster Manager and MySQL Cluster.
Site. A set of hosts on which MySQL Cluster processes to be managed by MySQL Cluster Manager are located. A site can include one or more clusters.
Cluster. A MySQL Cluster deployment. A cluster consists of a set of MySQL Cluster processes running on one or more hosts. A minimal cluster is usually considered to include one management node, two data nodes, and one SQL node. A typical production cluster may have one or two management nodes, several SQL nodes, and 4 or more data nodes. The exact numbers of data and SQL nodes can vary according to data size, type and rating of hardware used on the hosts, expected throughput, network characteristics, and other factors; the particulars are beyond the scope of this document, and you should consult MySQL NDB Cluster 7.3 and NDB Cluster 7.4, for more specific information and guidelines.
Host. A computer. The exact meaning depends on the context:
A computer where one or more MySQL Cluster processes are run. In this context, we sometimes refer more specifically to a cluster host.
The number of cluster processes and number of cluster hosts may be, but are not necessarily, the same.
A computer where an instance of the MySQL Cluster Manager agent runs.
In order to run a MySQL Cluster using MySQL Cluster Manager, the MySQL Cluster Manager agent must be running on each host where cluster processes are to be run. In other words, when using MySQL Cluster Manager, all cluster hosts must also be MySQL Cluster Manager agent hosts (although the reverse is not necessarily true). Therefore, you should understand that anytime we use the term host, we are referring to a host computer in both of the senses just given.
Process. In the context of MySQL Cluster, a process (more specifically, a cluster process) is a MySQL Cluster node, of one of the following 3 types: management node (ndb_mgmd), data node (ndbd or ndbmtd), or SQL node (mysqld). For more information about these node types and their functions in a cluster, see NDB Cluster Core Concepts, and NDB Cluster Nodes, Node Groups, Replicas, and Partitions.
Package. A copy of the MySQL cluster software. This should include the binary executables needed to run the cluster processes of the desired types on a given host. The simplest way to make sure that this is done is to place a copy of the entire MySQL Cluster distribution on each computer that you intend to use as a cluster host.
Configuration attribute. A value whose setting affects cluster operations in a clearly defined and measurable way. When running MySQL Cluster manually, configuration is accomplished using cluster configuration parameters, MySQL server options, and MySQL system and status variables; MySQL Cluster Manager masks the differences between these, provides a unified view of them; see Configuration attributes, for more information.
Agent. A MySQL Cluster Manager process that runs on each cluster host, responsible for managing the cluster processes running on that host.
Client. The MySQL Cluster Manager client is a software application that allows a user to connect to MySQL Cluster Manager and perform administrative tasks, such as (but not limited to): creating, starting, and stopping clusters; obtaining cluster and cluster process status reports; getting cluster configuration information and setting cluster configuration attributes.
This section provides an architectural overview of MySQL Cluster Manager, its components, and their deployment.
MySQL Cluster Manager is a distributed client-server application consisting of two main components. The MySQL Cluster Manager agent is a set of one or more agent processes that manage MySQL cluster nodes, and the MySQL Cluster Manager client provides a command-line interface to the agent's management functions.
Agent. The MySQL Cluster Manager agent is comprised of the set of all MySQL Cluster Manager agent processes running on the hosts making up a given management site. A MySQL Cluster Manager agent process is a daemon process which runs on each host to be used in the cluster. In MySQL Cluster Manager, there is no single central server or process; all agents collaborate in managing a cluster as a whole. This means that any connected agent can be used to carry out tasks that effect the entire cluster.
Each agent process is responsible for managing the MySQL Cluster nodes running on the host where the agent is located. MySQL Cluster management and SQL nodes are managed directly by the MySQL Cluster Manager agent; cluster data nodes are managed indirectly, using the cluster management nodes.
Management responsibilities handled by the MySQL Cluster Manager agent include the following:
Starting, stopping, and restarting cluster nodes
Cluster configuration changes
Cluster software upgrades
Host and node status reporting
Recovery of failed cluster nodes
Creating, performing initial configuration of, or starting a cluster, requires that agent processes be running on all cluster hosts. Once the cluster has been started, it continues to run even if one or more agent processes fail. However, any failed agent processes must be restarted before you can perform additional cluster management functions.
Client. A MySQL Cluster Manager client is a software application used to access an MySQL Cluster Manager agent. In MySQL Cluster Manager, the client is actually nothing more than the mysql command-line client, started with the options that are necessary for it to connect to an MySQL Cluster Manager agent. MySQL Cluster Manager 1.3.6 and later releases include an mcm client for ease of use; this client consists of a script that acts as a wrapper for the mysql client with these options.
By way of example, we show how MySQL Cluster Manager would be deployed for use with a MySQL Cluster running on 4 host computers. This is illustrated in the following diagram:
In this example cluster, 2 of the hosts each house a management server and an SQL node; the other 2 hosts each house 2 data nodes. However, regardless of the distribution of cluster nodes among the hosts, a MySQL Cluster Manager agent process must be runnings on each host.
A MySQL Cluster Manager client can be used to access the agent from any of the hosts making up the management site to which the cluster belongs. In addition, the client can be used on any computer that has a network connection to at least 1 of the hosts where an agent process is running. The computer where the client itself runs is not required to be one of these hosts. The client can connect to and use different agent processes on different hosts within the management site, at different times, to perform cluster management functions.