Re: A question about arbitration.

[David Smith]

Gabriel Crespo Carmona Wrote:
-------------------------------------------------------
> Hi David,
> I think the point here is the sentence "in case a
> network partition happens". According to MySQL
> documentation:
>
> Management node: The role of this type of node is
> to manage the other nodes within the MySQL
> Cluster, performing such functions as providing
> configuration data, starting and stopping nodes,
> running backup, and so forth [...]
>
> When a rolling restart is performed, there is no
> need for any arbitration, so there should not be
> any problem.


Thanks for replying. In a rolling restart, when we stop the mgmd process from the mgm management client, the NDB kernel probably knows the mgmd process has lost contact NOT because of a network partition. But let's say we do it the brutal way--kill the mgmd process with the command 'kill'. It seems there is no difference between this situation and that the mgmd process got disconnected from the remaining nodes. But doesn't the latter case result in a network partition and requires an arbitration?


In fact, if you stop the management
> node while running the cluster, nothing happens to
> data nor sql nodes.
>
> I hope this helps.
>
> Regards,
> Gabriel