suggestion about hardware specification

Support forum for Nagios Core, Nagios Plugins, NCPA, NRPE, NSCA, NDOUtils and more. Engage with the community of users including those using the open source solutions.
Locked
satheesh19870819
Posts: 43
Joined: Wed Mar 21, 2018 1:11 pm

suggestion about hardware specification

Post by satheesh19870819 »

Hello All,

I am going to built a nagios core server . I need your suggestion for the hardware specification . Lets take that i am going to monitor 5000 nodes and 50000 services in this servers . How many cores , RAM and hard disk needed for this setup
User avatar
mcapra
Posts: 3739
Joined: Thu May 05, 2016 3:54 pm

Re: suggestion about hardware specification

Post by mcapra »

The answer to this is almost always "it depends". It depends on what sorts of checks you are running, how frequently you're running them, what things in addition to running checks you're doing within Nagios Core (notifications? event handling?), what are the auditing requirements, just an awful lot of variables.

Nagios Core is a framework, it is not an application that performs in a consistent way with consistent throughput. It will perform exactly as well as it is configured by the administrator. If I feed a bunch of dirt simple compiled plugins through as my service checks, it's going to perform better than many complex interpreted plugins. Especially when we're talking about 55,000 combined checks; The differences between a ~500 line compiled binary and a ~5000 line interpreted Perl script are incredibly significant.

I've seen 4 core 16GB machines handle ~20,000 checks with no issues. I've also seen 8 core 64GB machines struggle to handle ~10,000. Especially at this scale, there are going to be significant differences to how one Core machine performs versus any other given one. Not all infrastructures are created equally.
Former Nagios employee
https://www.mcapra.com/
npolovenko
Support Tech
Posts: 3457
Joined: Mon May 15, 2017 5:00 pm

Re: suggestion about hardware specification

Post by npolovenko »

I definitely agree with @mcapra's response. We need to know more info on what kind of service checks you're going to run. I'd also add that not all CPUs and RAM are created equally, different service check intervals can negatively or positively affect the system load, passive checks take fewer resources.
I'd probably start with at least 16gb of ram, 140Gb of hard drive space and 8 CPU cores. Check out this requirements page:
https://assets.nagios.com/downloads/nag ... ements.pdf

Also, RAM disk is a must(this tutorial should work for Nagios Core as well).
https://assets.nagios.com/downloads/nag ... giosXI.pdf

I'd recommend adding plus minus a thousand checks at a time and run the config verification after each time.
Also, as an option, you could split hosts and service checks between 2 Core servers.
As of May 25th, 2018, all communications with Nagios Enterprises and its employees are covered under our new Privacy Policy.
satheesh19870819
Posts: 43
Joined: Wed Mar 21, 2018 1:11 pm

Re: suggestion about hardware specification

Post by satheesh19870819 »

Thanks a lot @mcapra and @npolovenko .

Thats really an awesome information for architecting Nagios.

From your Docs i understood that Virtual machines will not work out for larger monitoring services .

Let me go for Physical server .

Because i tried 8core with 8 GB ram in virtual machine where i found very worst performance that Forking issue earlier which is the reason for my new plan .

How about NIC Card . Am bit confused about whether to with 6 - 1GB NIC or 2 10GB NIC so that i can plan the port configuration accordingly .. Am planning to bond the NIC and use for handling these many checks .
scottwilkerson
DevOps Engineer
Posts: 19396
Joined: Tue Nov 15, 2011 3:11 pm
Location: Nagios Enterprises
Contact:

Re: suggestion about hardware specification

Post by scottwilkerson »

satheesh19870819 wrote:How about NIC Card . Am bit confused about whether to with 6 - 1GB NIC or 2 10GB NIC so that i can plan the port configuration accordingly .. Am planning to bond the NIC and use for handling these many checks .
Honestly, either should be fine. Actually, you shouldn't be getting anywhere even near 100mbps in bandwidth with 50,000 checks, unless your check results are abnormally large.

I would guess an average result size with performance data is about 256bytes or 2048 bits

2048 bits * 50,000 = ~102mb and if you are running checks every 5 minutes, that's only 341kbps
Former Nagios employee
Creator:
ahumandesign.com
enneagrams.com
satheesh19870819
Posts: 43
Joined: Wed Mar 21, 2018 1:11 pm

Re: suggestion about hardware specification

Post by satheesh19870819 »

So to reduce Service check latency and Host check latency what resource should i increase .
scottwilkerson
DevOps Engineer
Posts: 19396
Joined: Tue Nov 15, 2011 3:11 pm
Location: Nagios Enterprises
Contact:

Re: suggestion about hardware specification

Post by scottwilkerson »

satheesh19870819 wrote:So to reduce Service check latency and Host check latency what resource should i increase .
we really don't have enough information about how the system is setup, how many checks you are performing, what type of checks they are and what the frequency is.

However, as a blanket statement, FAST disks not utilized by other systems go a really long way, (think SSD raid arrays).
Former Nagios employee
Creator:
ahumandesign.com
enneagrams.com
Locked