Clusteroverview, Simple walkthrough...


Im in the process of making a few trivial changes to clusteroverview and so this page will change shortly but the following should help set up a simple scalability benchmark with swingbench 2.2.

In this particular scenario we have a four node cluster with hostnames node1, node2, node3 and node4. The database is called orcl and has 4 instances orcl1, orcl2, orcl3 and orcl4. We have a server that will generate the load called loadgen1 and a nodebook to display the results and coordinate the load called notebook1.

On nodebook1 we start the coordinator in graphical mode. This will allow you to see whats going on.
C:/> coodinator -g &
This is a one off event that does not need to be repeated.

An optional step is to start a cpumonitor on each of the cluster nodes (swingbench needs to be installed on each of them). On Unix/Linux it possible to start these remotely, in this scenario from notebook1. 
$ > ssh -f node1 'cd swingbench/bin;swingbench/bin/cpumonitor';
$ > ssh -f node2 'cd swingbench/bin;swingbench/bin/cpumonitor';
$ > ssh -f node3 'cd swingbench/bin;swingbench/bin/cpumonitor';
$ > ssh -f node4 'cd swingbench/bin;swingbench/bin/cpumonitor;
The next step is to start the loadgenerators. In another window start minibench/charbench or swingbench (same thing different interface).

NOTE : Whilst the examples use ssh to perform this functionality it is also possible to manually log onto the loadgenerator/node and start the service manually. 
$ > ssh -f loadgen1 'cd swingbench/bin;swingbench/bin/minibench -cs //node1:1521/orcl1
-dt thin -co //notebook1/CoordinatorServer
This will start start minibench running against node1 using a thin jdbc driver and attach to our local coordinator. Do this for the other nodes. 
$ > ssh -f loadgen1 'cd swingbench/bin;swingbench/bin/minibench -cs //node2:1521/orcl2
-dt thin -co //notebook1/CoordinatorServer
$ > ssh -f loadgen1 'cd swingbench/bin;swingbench/bin/minibench -cs //node3:1521/orcl3
-dt thin -co //notebook1/CoordinatorServer
$ > ssh -f loadgen1 'cd swingbench/bin;swingbench/bin/minibench -cs //node4:1521/orcl4
-dt thin -co //notebook1/CoordinatorServer
In the coordinator you should now see 4 loadgenerators attached.

We'll need to modify the local clusteroverview.xml on notebook1 (the location we intend to launch clusteroverview).

There'll be a section called Coordinator and MonitoredDatabaseList.

Change the coordinator section to reflect your environment. In this scenario to notebook1 where started the coordinator 
<Coordinator>
   <Location>//notebook1/CoordinatorServer</Location>
</Coordinator>
And the MonitoredDatabaseList. 
<MonitoredDatabaseList>
       <MonitoredDatabase DisplayName="//node1:1521/orcl1" Username="system" Password="manager" DriverType="thin" ConnectString="node1:1521:orcl1" MonitoredUsers="CC,SOE"/>
       <MonitoredDatabase DisplayName="//node2:1521/orcl2" Username="system" Password="manager" DriverType="thin" ConnectString="node2:1521:orcl2" MonitoredUsers="CC,SOE"/>
       <MonitoredDatabase DisplayName="//node3:1521/orcl3" Username="system" Password="manager" DriverType="thin" ConnectString="node3:1521:orcl3" MonitoredUsers="CC,SOE"/>
       <MonitoredDatabase DisplayName="//node4:1521/orcl4" Username="system" Password="manager" DriverType="thin" ConnectString="node4:1521:orcl4" MonitoredUsers="CC,SOE"/>
</MonitoredDatabaseList>
NOTE : It is currently important that the display name matches the connectstring of its associated loadgenerator(s).

Finally start clusteroverview.
C:\> clusteroverview

You can shut all of the load generators down (annd coordinator) with the command
C:\> coordinator -stop