5.6.4 The start process Command

5.6.4 The `start process` Command

start process {[--initial|-i] nodespec | --added} cluster_name

nodespec:
    {nodetype | process_id_list}
process_id_list:
    process_id[, process_id[, ...]]

This command starts the MySQL NDB Cluster processes specified by nodespec in the cluster named cluster_name. The status of the processes to be started, as shown by show status --process, must be added, stopped, or failed (only if the failed process has exited properly can it be restarted with the command).

This example demonstrates how to start the process having the process ID 1 belonging to the cluster mycluster:

mcm> start process 1 mycluster;
+------------------------------+
| Command result               |
+------------------------------+
| Process started successfully |
+------------------------------+
1 row in set (13.93 sec)

You can obtain process IDs for all processes in a given cluster using show status --process or list processes. These are the same as the node IDs for these processes as shown in the output of other mcm client commands such as get or in the output of ndb_mgm -e "show" (see ndb_mgm — The NDB Cluster Management Client).

Instead of a single node, you can also specify the type of nodes or a list of nodes to start :

mcm> start process mysqld mycluster;
+------------------------------+
| Command result               |
+------------------------------+
| Process started successfully |
+------------------------------+
1 row in set (15.72 sec)

mcm> start process 146,147 mycluster;
+------------------------------+
| Command result               |
+------------------------------+
| Process started successfully |
+------------------------------+
1 row in set (3.92 sec)

The following requirements must be fulfilled when you specify a list of nodes to start, or the command will fail:

All nodes in the list must be of the same process type.
The list should not include all managed nodes of the cluster.
After the command finishes running, there should be at least 1 running data node per node group, and more than half of all data nodes in the cluster should be running.
The general rules on process dependencies are satisfied (for example, a mysqld node depends on some data nodes running, a data node depends on some management nodes running, and so on).
StartPartitionedTimeout> 0 is needed to allow a single data node out of a total of two to be started alone.

When the --initial option (short form: -i) is used, the following happens:

For a data node, MySQL Cluster Manager starts it with the --initial option, causing the data node to rebuild its file system.
For an SQL node, MySQL Cluster Manager rebuilds the mysqld data directory with the mysqld --initialize-insecure command for MySQL 8.4, 8.0 and 5.7, and with the mysql_install_db command for MySQL 5.6. The node's data directory must be empty, or the reinitialization will not be attempted.

Invoking this command with the --added option rather than with a nodespec starts all nodes that were added previously to the cluster using add process but not yet started. For the added data and mysqld nodes, the use of the --added option also implies the use of the --initial option, meaning that mcmd will attempt to initialize the added nodes (see description for the --initial option above). Also, when the --added option is used, once all the added nodes are running, a CREATE NODEGROUP command is issued to the management node for the creation of new nodegroups.

You cannot use this command to start a mysqld process in a stopped or unavailable cluster—trying to do so will cause an error. This applies, for example, to the case in which a cluster has been created for a cluster import, but the import is not yet completed (see Section 5.4.1, “The create cluster Command”, and Section 4.5, “Importing MySQL NDB Clusters into MySQL Cluster Manager”).