[Klug-general] MySQL Clusters

[Alan Buchel]

On 15/04/12 12:40, Kevin Groves wrote:the table though was instantly
available for query.
> Did that for all the small ones while watching the memory usage on
> each node. A few times I had to restart nodes when various messages
> came up because tweaks to the configures were required. Rolling
> restarts are very good on the cluster and having four nodes is the
> ONLY way for this to work, some have tried it with two but you would
> need to bring the whole cluster down for any changes. Not good.
This intimates that failure (or temporary disconnection) of more than 1
node of 4 means the cluster just stops working?
>
> As I then started to edge into the swap disk space hit issues as I
> knew I would, the whole point is it should all be in memory. Before I
> got there I did a few tests between my existing MySQL server and this
> cluster. Considering the main server was under load and far out spec
> the processor etc the cluster did still beat it by a factor of 10 most
> of the time. I say most of the time, the documents do say that certain
> features are not supported and its a long list.
A factor of 10 (wow), means that server hardware need not be very highly
specced, just memory I guess.

Thanks Kevin, valuable research!