I am looking for anyone's experiences on running the polling engine on VMWare. Good, bad, whatever. Trying to decide if a physical box is needed or if I can stay virtual.
I would put the database on an external SQL server already existing... so this would house the polling engine and serve up the web page.
My QA Orion v8.0.2 enviroment is entirly on VM, including the SQL DB, it runs fine. But it only has One Orion server and one DB :
How many elements do you plan to monitor?
Once all is said and done, anywhere from 3000 to 5000 elements.
John,
This is a good start for the amount of items I will have, most likely with 30 or 40 more nodes and more volumes. What are the specs of your VM that is running this?
JA
Honestly with that many nodes I wouldn't consider using VMWare. That's just me. I've run Orion with SL-100 and SL-500 under VMWare and it worked pretty well, but SL-2000 (2000 or more elements) VMWare just didn't perform well. People were always complaining about the performance of the web interface. We now run on dedicated hardware with close to 1800 elements and a separate SQL server. Even with NTA 2.0 and the Application Monitoring Add-on, the performance is excellent.
I only run Orion on VM in QA for version, hot fix and patch validation, and the DB is intentionally small (msde).
The VM client is a "plain-jane" definition as follows:
Windows Server 2003 Enterprise Edition SP13.20GHz, 1 GB RAM
C:\OS - 7.2GB
E:\APPS - 4.0GB
F:\DATA - 10GB
Orion Network Performance Monitor V8.0 SLX
VM or not; It's not the overall number of nodes, interface, and volumes in Orion that take up space in the DB. It is the length of time that stats are retained that make the DB large. The larger the db, and the additional detailed information displayed, the "slower" Orion will be. Now this is also dependent on how you setup the views. For instance if you add all syslog messages to the Node Detail View, and the node has allot of messages, the load time for the node detail view will take a very long time, thus your clients will say Orion is slow... You, as the the Administrator, have the responsibility to set your clients expectation of application response time. A hierarchal approach to the overall setup of Orion works fairly well in doing so. Summary and Detail views should be just that, quick summary and slower detail view.
The web performance is acceptable, after removing the ALL NODES view from the default Summary page.
I am not recommending that a full blown production Orion deployment be ran on VM, but it is a very cost effective solution for application QA/test bed; which is a requirement for us. It's not that I don't trust VM's performance or stability, but Orion's role and importance deserves it's own dedicated hardware.
