Nagios 4 Load issues on OS X 10.9

[umd]

I have the same issue as in this post with two Nagios installation on OS X 10.9 with Nagios 4.0.6 & Nagios Plugins 2.0. The load nagios 4 is producing on these machines is really amazing . I just reverted today to 3.5.1. To get an impression have a look at the attached PNG

[tmcdonald]

Can you run it with Core 4 and provide the following:

Code: Select all

ps -ef
top
free -m

so that we can get an idea of what all is running? Then, do the same with Core 3.5. We want to compare the two. In addition, your /var/log/messages log would be helpful. It may be easiest to zip all this up and PM it.

[umd]

Hello tmcdonald,

thanks for your reply and sorry that it had taken some time for my response. I did the testing an a different machine as I didn't want to mess around with my main monitoring system. The "testing" machine (17hosts 86 services) has far less checks but the effect is still very recognizable but clearly not as dramatic as it was on the production system (73 hosts 243services). The testing system is a Mac mini 6,1 with a Core i5 Ivy Bridge 2,5Ghz with 16GB RAM running Mac OS X 10.9.3 with Server.app. As the free -m command does not exits on this platform I posted a vm_stat as a substitute.

hope this helps.

Regards

[abrist]

Just to clearly identify the environment and potential causes: Are you using any addons or utils like mk_livestatus, check_mk, thruk, ndo, etc?

[umd]

no nothing like that. Only other service running on that machine is a cacti installation.

[sreinhardt]

I noticed that your graph is saying load percentage. Is that a load check or cpu percentage, because they are quite different. I do believe we have a possible mixture of a bug and systems seeing load increases due to the number of worker processes now. As for the graph, I just wanted to confirm exactly what we were looking at, as it is rather odd to label load as a percentage.

[umd]

Hi

here are the other graphs I can provide. hope this helps.

Henning

[abrist]

The core devs came across a few issues with core that may be partially at fault for some of the load increases:
https://github.com/NagiosEnterprises/na ... a91e548dbc
You may want to try the code right from the master branch and then compare it to your current load results.

[umd]

Hi,

sorry for the late reply. Unfortunately the code did not change the load issues on my 10.9 testing machine.

Henning

[sreinhardt]

How many workers do you have running at a time? One thing to note with core 4 vs core 3, is that while it no longer has to fork the entire process memory, it is able to more actively use the cpu due to having faster executing worker processes. This will somewhat artificially increase load values as more processes are active at the same time, and will definitely increase cpu usage if the amount of checks causes that. Let's get a general idea here:

Code: Select all

ps -ef | grep bin/nag