Sizing Help

[costanza2k1]

We're in the process of purchasing the suite of Nagios tools for our company and wanted some help with sizing of our production servers. We figure to expand to 1,000 hosts and about 12,000 checks. Most of the checks will be snmp, wmi based, with a handful of agents deployed as well. We are going to with 2 production servers (hardware) for HA. Will need a recommendation for CPU and RAM please.

[jdalrymple]

I wrote this same exact response yesterday. I feel like we should create a sizing guide with the following (semi useless) information.

There is no hard set rule on what works and what doesn't with regard to sizing. There are simply far too many variables to factor in. Your mentioning of what plugins you'll be using is quite helpful, but doesn't answer everything. One of the biggest factors is the stability of your environment. Nagios can be humming along nicely on the brink but then a switch in your datacenter fails, about 500 notifications and event handlers kick off, that is when things spiral downward. It's best to have a nice cushion. This is mostly exacerbated by memory starvation (due to swapping of notification scripts and/or the loss of disk cache) but can be a bit aggravated by lack of CPU as well.

The best advice is to build on a VM so that you can allocate resources as needed. That or if you can start modeling your environment on existing hardware prior to purchasing.

I did offer a shoot from the hip answer when I gave my advice yesterday and I'll do the same again today... again with absolutely no guarantee of accuracy or reliability.

If your environment is stable, you offload your database and you go with the approximate sizing figures you mentioned, I would plan on 2 hexacore CPUs being adequate and at least 24GB of memory. These are just a general possible starting point, double it and I doubt you'll ever be dissatisfied

As an aside, in my lab I've created a very smooth running HA environment (using corosync, pacemaker and drbd) where the DB resides on one node and the App on the other, each with the ability to run Nagios checks. This is a great setup IMO but would require that you plan on each node being able to handle the entire workload, so definitely way oversize if you do something the like.

[ranjitw]

Hi jdalrymplem,

Could you please let us know the steps followed for achieving HA(using corosync, pacemaker and drbd) .

It would be of great help if you provide step to step guide.

[jdalrymple]

Setting up Linux Clustering is not in scope for Nagios. One of our partners has published a paper that does outline the process though:

https://www.nagios.com/news/2015/10/pre ... nagios-xi/

[costanza2k1]

Thanks for your advice...

[rkennedy]

I will now close this thread, let us know if you have any other questions.