Re: some issues

[Thilo Raufeisen]

Hi,

> There is a reporting of 80% full, 85% full and so forth going to the cluster
> log. Check the manual for details on how to ensure those messages are put into
> the cluster log.


the memory-notice in the clusterlog is insufficient.
Assume the cluster fills each day with 0,5 %. I´ve no possibility to figure out this increase.
Now some new advertisement-tour starts from the marketing and the increase in memory-usage is 4% per day.
Then I´ve only 5 days max. And if you know murphy: this starts friday night and the new hardware is ordered monday. So only 2 days left. Way to short.
There must be some kind of statement, that can be executed at any time to keep track of this.


> Most likely your performance bottleneck is in the MySQL
> Server not, in the data nodes. Try to divide the data and
> insert from several MySQL Servers at the same time and
> you will decrease load time even more.

I´ve modified the testscript, so that it uses all MySQL-Server and it was faster. But if I keep in mind, that the whole cluster is memory-based, I think that it´s still too slow for such a hardware-investment.


> This is currently the "perfect" method.

hmm
There needs to be a way to add new nodes without any downtime.
I know, the docu says this, just wanted to clarify this position

Regards
Thilo