MaxDB on VMWare ESX Server ????????

Hello Everybody !

 

Sounds strange, but nevertheless:

We have a vmware cluster with n esx servers

We use a virtal "SUSE Linux" server on one of this esx servers as maxdb
database server.

Everthing works as expected.

Even if possible, we don't want to share the hardware resources with
serval virtual servers.

(IO problems when serveral io intensive servers have to share 1 network
card)

The main reason for us is, that we want to use the vmware ability to
restart the virtual server

automatically on another esx maschine when the main esx maschine is
down. 

Server downtime = max 5 minutes     

 

First question:

In the lists i found, that maxcpu 2 is a good maxdb parameter for a
maschine with 1 dual core cpu.

The esx server has 2 dual core cpus and there is only 1 virtual linux
server with 1 maxdb instance on the maschine.

The virtual server is configered with 2 cpu's!

So maxcpu for the maxdb database = 4 ??????????????

Has anyone experiences with installing a maxdb on an esx maschine ??

Someone told me, that the virtual server only knows 2 virtual processors
and nothing about dual core !?

So maxcpu = 2 for maxdb ????

 

Second question:

Until now i use maxcpu = 2, but i found, that calling an sql command
like update xxx on a table with 2 million records

extremly slows down the server. When looking at the linux maschine we
found, that 1 cpu = 100 % for n minutes,

the second has only 4% cpu but all user cry: what about performance!

 

Is there a way to limit a maxdb session ???

(something like: do not use more than 50 % of cpu or better do not use
more than 50 % of cpu if more than n sessions?)

 

Third question:

All application users connect via odbc with a standard username "xyz"

90 % of the users start the application exe from a terminal server,

so when looking at the maxdb sessions(500) i found:

user      node

xyz                   terminalserver1

xyz                   terminalserver1

xyz                   terminalserver1

xyz                   terminalserver1

xyz                   terminalserver1

xyz                   terminalserver2

xyz                   terminalserver2

xyz                   terminalserver2

...

 

In our aplication we know the connection_handle returned from
sqlconnect(database,user,pass,...)

and all commands against the database work like
sqlexec(connection_handle,mycommand)

Is there a way to identify maxdb sessions with the connection_handle ???

So we can kill the session that started the never ending sql command or
has other problems??? 

 

Any help or idea welcomed

Best regards

Albert