MySQL Cluster supports online backups using the ndb_mgm client tool with the START BACKUP statement. If you omit the ID, it will take automatically the next value.

Problem is that when you do a system restart, this ID is reset to 1. If you didn’t move your backups from your data nodes to another machine (which is of course something you should do), you might get errors.

A rather easy trick is to provide a timestamp as ID, for example using Unix time.

shell> ID=`date +%s` && ndb_mgm "START BACKUP $ID"

Now your the directories on the data nodes containing the backups will have a time in them, and it will survive system restarts. You can of use of course use any integer as ID: 2^32 as of MySQL 6.3.23 (earlier only 2^31 is supported).