what happens when single and one and only management node goes down in NDB Cluster

Question

I have set up the MySQL NDB Cluster (mysql-cluster-gpl-7.3.5-linux-glibc2.5-x86_64) with 5 Nodes as described below:

Node A: multithreaded data node1, SQL node1

Node B: multithreaded data node2, SQL node2

Node C: management node1

So I have kept one and only one management node which handles other nodes.

When transactions are going on, I suddenly kill the process of management node and still other nodes are kept running. Even the response times from both DBs(SQL Nodes) are not fluctuating.

Can you explain me what happens at this moment? Does the SQL Nodes sync in this scenario?

OR

Do they need management node to keep them in sync?

Thanks in advance.

Answer 1

Management node acts as arbitrator of the data nodes. If mgmt node goes down and both data nodes see each other, cluster has quorum and operates normally, syncing data. Mgmt node role is simply keeping cluster configuration as arbitrate, and it's not involved in data sync.