Server specification

[lntinfo]

If we have 123359 services in the master server. In how many salve servers these can be distributed equally & what would be the configuration of the slave servers ?

[scottwilkerson]

I don't exactly understand the question? When you say "slave server" what type of setup are you referring?

[lntinfo]

Currently we have prospective client where there is master / slave architecture Running on Nagios Core. The salient points of the architecture is as follows:
a) There are three master servers spread across three regions A, B, C.
b) Each master server is linked to a number of slave servers
c) Region A has 126,000 services 6 slave servers, Region B has 15,000 services, 3 slave servers and region C has 9,000 services and 2 slave servers.
d) Each master server acts as a passive listener
e) In each slave server, there are a number of Nagios instances running in the system
f) Each instance of the slave is monitoring 3000 services.
g) In total there are 150,000 services and 5000+ hosts in the system

[mguthrie]

This appears to be a duplicate of your previous thread:
http://support.nagios.com/forum/viewtop ... f=6&t=6664

Was there something specific you were looking for that wasn't answered in the above thread?

[lntinfo]

Existing thread i was looking out for the best approach for Nagios.

Similarly, if we are going with suggested approach then what would be the server specification for Slave & Master.

[slansing]

What do you mean by specification? Are you speaking from a hardware perspective, or the software being used to monitor? Much like Mguthrie said in the previous thread XI has more features than Fusion, but in a large scale environment, Fusion would make it a lot easier to manage multiple Nagios XI server's with thousands of checks each.

[lntinfo]

Yes. Hardware perspective (Server configuration )

[mguthrie]

I might actually go bug an actual field user and see if he's willing to chime in on this one. We've never benchmarked anything internally past about 20,000 services, so it might be more useful to get input from another user with a large installation.

Region A is by far going to be the most challenging for hardware:
- Fastest disks money can buy
- Lots of CPU core, more cores are better than high quality cores
- Offload mysql to another server
- Utilize a RAM disk
- Have performance data on a separate physical drive if you're going to process performance data and have graphs

Region B and C
- 8-12 CPU cores
- Fast disks
- Mysql probably would not have to be offloaded

[jsmurphy]

Mike is pretty much on the money, I'm expecting you're sites should look something like this:

Site A. (Make sure there is A LOT of bandwidth between the XI server and the Database server)
1 x Nagios Fusion Server
1 x Nagios XI Server
1 x Nagios Database Server (MySQL for NDOUtils and NagiosQL)
We get a capacity max of about 12k services on a dual proc, 4gb Memory and 15K SAN disks box while running XI. We avoid using high resource utilization plugins such as check_esx3.pl. This is a pretty low spec box because it's a VM, I'm going to assume you are going to have a physical that can handle at least 25k services.
So assuming 25k services per server: 5~6 x Slave Nagios Core

Site B.
1 x Nagios XI Server
1 x Nagios Database Server
1 x Nagios Core Server with a lot of room for scaling

Site C.
1 x Nagios XI Server
1 x Nagios Database server
1 x Nagios Core Server (Optional, you could easily run this site purely from the XI server)

It's important to note that we use virtual hosts which is why the Resource allocation is so small... Our database server runs a measly dual proc, 1gb ram on 15K SAN disks. You could also trial leaving the database and Nagios XI server together but that in my experience has had some pretty severe IO repercussions.

Really the two most vital factors are fast disk, a decent CPU and a little bit of optimization... you could probably squeeze (at a guess) the above setup on quad core processors, 8gb RAM and 10~15K disks; but if you can get SSD's for your core processing servers then do it. Ultimately your mileage is going to vary depending on the efficiency of the plugins you are using and the overhead of the distribution of such a large number of workers.

[lntinfo]

Thanks jsmurphy & mguthrie .

If we are migrating from Core to XI, do we need to have Mysql Database server to each location e.g. A,B & C
And then will have one Nagios Fusion server installed centrally so that we can get centralized view of all Slaves & Masters