How child copy nodes are added

The child_creation option of the SET MIRROR OPTION statement controls how child nodes are added to a read-only scale-out system. The following values are supported:

  • Automatic   The root database server authenticates the copy node once it starts, or creates a new copy node if the copy node is not known. This is the recommended setting because the root server creates the definition for unknown copy nodes so that you do not have to create them manually.

  • Off   You must connect to the root database server and execute a CREATE MIRROR SERVER statement to add a new copy node.

  • Manual   You can add copy nodes to the tree by connecting to a copy node and executing a CREATE MIRROR SERVER statement for that database server. This statement requires DBA authority. The copy server sends a request to the root database server to define the new copy node. Once the copy node is defined, the root database server allows the new copy node to request log pages.

The database stores the connection string that is associated with the primary database server in the system. When you start a new copy node that has not been previously defined, it connects to the root database server using this connection string. The root database server uses the value of the authentication_string option that is stored in the database to authenticate the copy node.

As part of the mirror connection request, the copy node sends the copy database server's name and a string containing the copy database server IP addresses and ports to the root database server. Once the copy node is authenticated, the root database server determines whether the copy node is known. If the copy node is not known, then the root database server executes a CREATE MIRROR SERVER statement to define the new copy and its connection string. Once the copy is known, the root database server can establish a connection to it. The copy then requests all the transaction log pages that it does not already have, and once the copy node has them, the root database server starts pushing new transaction log pages to the copy node.

All copy node servers must have unique server names.

 See also

Adding copy nodes
Copy node parent assignment