Nagios Support Forum

Hi all.

I have some graphcs for a cisco router and several of the graphs are showing blanks in the middle of a traffic spike. How can I fix it?

Anything unusual in "/usr/local/nagios/var/npcd.log" or "/usr/local/nagios/var/perfdata.log"?

Everything looked pretty normal in npcd.log. perfdata.log showed a lot of timing out though so I increased the parallel threads to 10 and wait time to 30 in npcd.cfg. Anything else I should look at?

cwscribner wrote:Everything looked pretty normal in npcd.log. perfdata.log showed a lot of timing out though so I increased the parallel threads to 10 and wait time to 30 in npcd.cfg. Anything else I should look at?

If you are getting timeouts make sure that the server isn't too overloaded...

Also, make sure your restart npcd for the changes to take affect.

The server is definitely overloaded and a new 16 core CPU is en route. So basically, the overload could be what's causing the blanks?

Well, the overloading is likely causing the timeouts that you described, which would lead to the perfdata not getting correctly put in the rrd file.

by default the npcd has a little bit of a safeguard against this by checking the load on the server with the following line npcd.cfg

A few questions...

Is fixing the load problem the only way to fix this problem?

How much of a safeguard is that line? (The load is regularly in the 8-12 range and is often in the 15-20 range)

Will the npcd.cfg changes make a difference in lightening the load so the data can properly be processed?

cwscribner wrote:A few questions...

Is fixing the load problem the only way to fix this problem?

How much of a safeguard is that line? (The load is regularly in the 8-12 range and is often in the 15-20 range)

Will the npcd.cfg changes make a difference in lightening the load so the data can properly be processed?

Honestly, I doubt that will lighten the load, I actually think if actually may help prevent the hard spikes if you set
sleep_time = 10
npcd_max_threads = 5

Also I would also check the /usr/local/nagios/etc/pnp/npcd.cfg and the /usr/local/nagios/etc/pnp/process_perfdata.cfg files and make sure logging is set to 0.

Already decreased the logging verbosity. I figured that would help things along.

What do the sleep_time and npcd_max_threads parameters do in terms of system metrics and graphing? Will increasing these increase the server load? Will increasing them provide better or more accurate graphing? Are there implications if you do the inverse?

the amont of work to be done is the same, but if you increase them, it tries to do more work at the same time, and could cause timeouts if the load gets too high.

If you decrease them, it will help prevent a large spike in load, but causes a more even load without the spikes