Here's a question for everyone...

What hardware are you running for your Core Server/s?

What hardware are you using for your remote probes?

How many sensors and what types of sensors do you have on each probe (local and remote)?

How would you say that your setup is working?

I'm really looking for ideas on how people scale their PRTG monitoring.

Article Comments

My setup is as follows:

Core Server - 6 core Intel Xeon Processor E5-2420 - 24GB RAM Local Probe - 10,000 sensors - Ping, Uptime and SNMP only.

Remote Probe - 2 core Intel Pentium Processor E5300 - 4GB RAM 1,200 sensors (inc. 950 WMI sensors on 10 minute scanning intervals)

I plan on moving all sensors from my core server to a remote probe server with 4 cores and 8GB RAM. I also plan on splitting the existing remote probe sensors between two probes of the existing specs (2 core Intel Pentium Processor E5300 - 4GB RAM)

Does anyone have any suggestions/advice/feedback for my PRTG installation?


May, 2018 - Permalink

Hi Matthew,

Sounds good so far. One thing you might want to consider is that the Remote Probe software always runs in x86, so the machine does not necessarily need more than 4 GB of memory, if it's exclusively used by the Probe. Always try to distribute the sensors across probes if you get notified about a delay.

Best regards, Felix


May, 2018 - Permalink

Thank Felix that is good to know.

I did not see that anywhere in the system requirements for remote probes. https://www.paessler.com/manuals/prtg/system_requirements#remote_probes

Matthew


May, 2018 - Permalink

Also, In my situation is it possible to run a failover cluster for my Core Server?

Thanks,

Matthew


May, 2018 - Permalink

Hi Matthew,

For cluster environments, we do not recommend to go above 5,000 sensors in total. Ass a rule of thumb, the load on the core server doubles and sensor scans which are located below the cluster probe are also sent twice.

The requirements for remote probes depend on the type of sensors you want to deploy on the probe. If you for instance do heavy flow-based monitoring, the probe would require more / stronger CPUs as if you just configure lighter protocols like HTTP or SNMP. Every probe comes with its own Probe Health Sensor, if the probe reaches a limit, PRTG will inform you so that you can better scale.

Best regards, Felix


May, 2018 - Permalink

Hi Felix,

So even if I have all my sensors on remote probes (not on the local probe or cluster probe) is it still not recommended?

Thanks,

Matthew


May, 2018 - Permalink

Hi Matthew,

Correct, the load occurs on the PRTG core server host depending on:

  • Number of sensors
  • Scanning Intervals
  • Number of Users
  • Number of Remote Probes
  • CPU intensive features like huge sensor reports, scheduled auto-discoveries, API usage
  • Network Connection Quality

All these metrics effect the utilization of the core server.

Best regards, Felix


May, 2018 - Permalink