Hello,
We have two PRTG servers set up in a cluster (they are in separate data centres for resilience), but have now hit performance issues with the probes and they are complaining of too many open requests.
I would like to add two additional servers as probes for our cluster to distribute the load and keep the resilience of the cluster i.e. if one side fails, the other continues monitoring just as it is now. How is this done?
As a test I have added the two servers as simply additional probes, but I don't think this is what I needed to do, perhaps I should have enlarged the cluster?
Cheers, Fernando
Article Comments
Hi,
Thanks for the response. The two new probes I created and added to the cluster are set to send their information to all cluster nodes as recommended.
So if I am reading Cluster Support for Remote Probes correctly, by simply adding a new probe I have to manually move sensors off the cluster to the new one, but if I want them to remain resilient i.e. one of the new remote probes goes offline, I have to make copies of those probes on a second extra probe? That doesn't sound very good as it would mean having to remember to create duplicate new sensors in the future.
I don't know if this helps, but this is what our set up looks like now:
The 01 machines are the original cluster, the 02 machines are the two new probes. The System Health warning on the first probe device is simply the CPU being high.
Cheers, Fernando
Feb, 2016 - Permalink
Hello Fernando, thank you for your reply.
Using remote probes in a cluster will allow you to keep monitoring (and the web-interface reachable) even when one of the Core Servers is unavailable, but it will not provide "monitoring from two distinct probes". This functionality can only be achieved with the Cluster Probe at the moment.
In other words, if you encounter performance issues due to load on the Cluster Probe the only currently available options are to reduce the number of sensors or increase scanning intervals, you will need to deploy an additional separate PRTG cluster and scale horizontally(please note that most licenses are only for a single standalone or clustered PRTG, so please check your license before deploying additional Core Servers).
Best Regards,
Feb, 2016 - Permalink
Hello there Fernando,
thank you for your inquiry.
No, adding more remote probes is indeed the correct procedure. Please keep in mind that "monitoring requests" are performed by Probes, either via the "Local Probe" (standalone Core Server), "Cluster Probe"(a virtual probe which refers to multiple cluster nodes) or a "Remote Probe"(With or without Cluster support).
Once you've added a Remote Probe to your system it should show-up on both or a single cluster node, depending on the configuration.
Once the probe is online you can simply move groups or devices from the "Cluster Probe" to any available Remote Probe. Please be aware that a Remote Probe in a Cluster behaves a bit differently than the Cluster Probe because it is still a single system which performs monitoring requests and sends these results back to both cluster nodes, whereas the Cluster Probe will perform the same monitoring requests from two distinct systems and store them redundantly on two nodes.
If you need to reduce the sensor load (e.g. Probe open requests) then deploying one or more Remote Probes is the correct approach.
For further details on "moving" sensors:
Best Regards,
Feb, 2016 - Permalink