Distributed Install

[BanditBBS]

Ok, I know I already asked this question in another post...but here is goes...

We were originally going to install about 6 separate XI servers across our environment. That way they only monitored a specific set of hosts, especially since our environment is quite large(5000+ hosts). We were going to use Fusion to roll everything up for our NOC and put in network gear with ping only checks so we could set up some parenting.

I think I've changed my mind. I think we'll setup gearman and have all the configuration in one main Nagios. This will give me a few benefits:

• Good excuse to not give admin rights to heads of departments, so less hands in the kitchen(Going to get a bunch of push-back but I'll fight it )
• True parenting and dependency setup(One of the biggest reason for redoing the monitoring)
• Less administration

So my main questions are...

• Will gearman continue to function properly with 2012, when released?
• Does this sound like a good idea?

Thanks

[mguthrie]

Gearman will work the same with 2012 as it will with the current version of XI. The key factor in that is the version of the Nagios Core engine underneath will be the same. (3.4.1). Sounds like a pretty common setup for larger environments. I'd recommend a nice piece of physical hardware to put it on if you haven't already got one, as well as offloading mysql and using a RAM disk, but that should be doable.

Also, if you're processing performance data, I would consider mounting the /usr/local/nagios/share/perfdata directory on a different drive than the rest of the system. The rrd writes will be a big deal with disk IO with that many hosts. Using rrdcached is also something to look into.

[BanditBBS]

They really love the virtual environment here, which is way under used. Do you really think it needs to be physical? I have no problem requesting addition CPU or Ram for the VM.

One last question....I already know the answer, but going to ask anyway...can we use a different database( Oracle or MSSQL)?

We'll follow the other recommendations as well.

[mguthrie]

They really love the virtual environment here, which is way under used. Do you really think it needs to be physical? I have no problem requesting addition CPU or Ram for the VM.

Take a look at this presentation, slide #6, given by a guy who oversees over 1.5 million service checks in his environment.
http://exchange.nagios.org/directory/Mu ... ny/details
I'm not saying it can't be done with VMware, but I would keep the hardware option on standby for the main XI box. The rest can be VM's, that part doesn't matter too much.

One last question....I already know the answer, but going to ask anyway...can we use a different database( Oracle or MSSQL)?

Nope, sorry.

[BanditBBS]

I'll be requesting hardware for the main server.

Searched the internet a little for my next question....What port does gearman use to send the checks to workers? Is 4730/tcp correct? I need this information so I can verify with our security and network teams that they will open the communication between two sides of the datacenter(side A can't see side B and one of the workers will be in side B while the main server will be in side A)

[mguthrie]

Yeah, that's the default port for both directions.

[BanditBBS]

Cool, thanks for all the answers, now to convince everyone here that I am correct

[mguthrie]

Nice, good luck!