Client: After stopping and restarting the MySQL Cluster Manager Agent with a stopped cluster, all cluster nodes were shown as having added for their status rather than stopped. When the cluster was started subsequent to this, This caused MySQL Cluster Manager to perform an initial start of the data nodes, resulting in the loss of the cluster's data. (Bug #13854142, Bug #13823551)

Client: Changes in some node configuration attributes require data nodes to be restarted with the --initial option, but do not require a system restart. Setting any of these attributes while the cluster was stopped caused all data nodes to be started with the --initial option, resulting in the loss of MySQL Cluster data.

Now in such cases, the set command used to make the change is rejected with an error. The operator can then wait until the cluster is running before trying to change one of these attributes; this update can then be made safely using a rolling restart of the cluster.