Hello,

I've asked this a few years back through mail, but at that time there were no features that could solve my problem, but since we are 6 versions further now I hope there is a solution.

We have multiple high-availability/failover clusters here (not a PRTG failover setup, but failover servers). We can reach the currently active server with the name system-msq-pcls.

When I add system-msq-pcls to PRTG, it seems to look for the IP-address that is currently attached to that name.

When we move our services to the other nodes on the failovercluster, we can reach that server through the same name, but PRTG seem to be cashing the IP-address that was attached when I added it, which leads to all the services being down while they are not down. Is there a solution to make PRTG look for the hostname instead of the IP?

As a work-around I thought of adding both servers of the clusters separately. But then I need to auto-acknowledge/pause an alarm if the service is UP at the other server, but I don't know it how set this up, if that's even possible.

I hope someone can help me with my questions.

Thanks in advance!

-Brian

Article Comments

Hello,

thank you very much for your KB-Post. Which exact sensors are you referring to? And what kind of OS is running on this cluster?

best regards.


Aug, 2013 - Permalink

Hello Torsten,

Mostly it's about WMI Service Sensor. The OS on the Servers is Windows Server 2008 R2.

So the services run on the cluster (system-msq-pcls). When I go with rdp to that cluster, it always goes to the active server with the UP services, but PRTG seem to store the IP of the active server when it's added, which results in alarms for services being down when we switch the failovercluster to the other node.

- Brian


Aug, 2013 - Permalink

Brian, the only real alternative I see would be using SNMP to monitor the Windows Cluster. The "problem" with WMI Sensors lies in the mechanism, how WMI (not PRTG!) builds connections. These base on the MAC-Address, and that is the reason why the WMI Sensors then fail with a cluster switch.
SNMP should not have that problem, if you run the sensors against the Cluster-IP/Hostname.


Aug, 2013 - Permalink