Host Ping vs Service Ping

[WillemDH]

Hello,

I have a few questions regarding availability.
1) What's the added value of having a host ping as well as a service ping? it seems a bit overhead to also add a service ping (with for example the windows server wizard)

2) What's best practice in the configuration of the pings or the host pings. it seems by default Nagios XI does a ping each 5 minutes. As a virtual server can reboot in 20 seconds, it seems a good idea to set the checks to less then 5 minutes but what impact will this have on the Nagios server when this is done for 600 servers..?

Thanks.

[abrist]

1) You are right, it is a bit redundant.
2) Every check will write to the database with ndo. You can definitely decrease the the ping interval, but do understand that with each check comes database overhead. Each environment should handle these differently. If you do not have many other small interval checks, you could decrease the interval more. ICMP will not add much load to the server itself, it is the queries and db writes that are expensive. You can tune the numbers progressively downward until the I/0 wait and/or load of the XI server nears unacceptable thresholds.

[WillemDH]

Thanks for confirming this. W'll monitor the load as we ass more servers. At the moment we got 97 hosts and 790 services. Can you tell me again what's approximately the amount of hosts I would be able to add? I know this depends on many things as check intervals etc, but it's just to get a grip on how far I can go.

[abrist]

Nagios XI hardware requirements below:
http://assets.nagios.com/downloads/nagi ... ements.pdf
Remember these rules are fast and loose, and that every environment is unique.

[Box293]

I beleive John Murphy raised this in his talk at the Nagios World Conference 2012.
Rational Configuration Design To Prevent Irrational Problem Solving

From what I understand, in relation to services, you cannot configure a service to have a dependancy on a host object. I beleive he suggested that by having a ping service, you can make all of the other services for that particular host have a dependancy on the ping service you create. When the ping service object goes into a critical state it changes the behaviour of all those services that depend on this ping service.

Thats a real simple explanation of what he talked about. There were some more compelling points about this disucssion which I cannot remember at this point in time.

Please correct me if I'm barking up the wrong tree

[abrist]

When the ping service object goes into a critical state it changes the behavior of all those services that depend on this ping service.

This is true and very useful, especially if you need to run separate event handlers for each service on a host. Event handlers and service dependencies are the best reason to have the extra checks. If you will never use any extended features for those checks, you can remove the duplicate ping checks without worry.

[WillemDH]

Well, for the moment I don't use event handlers, but once I get the time I will most certainly look into them. They seem interesting. But this will probably be in 2 years or so lol. At the moment I've got 100 hosts and 820 service, and I have about 800 more hosts to transfer from our SCOM 2007 R2 environment to Nagios XI... As monitoring is only one third of my job..., this will take some time.
You can close this thread though, thanks again for all the help. The support I got through this forum is so much better then the support I had from Microsoft in the past 4 years! It took ages before an issue got to the right MS engineer.. And the cost of SCOM is like 80 times higher.....

[slansing]

Haha, thanks for the great feedback! Closing as resolved (for now.)