Nagios Support Forum

I think I've worked out what is happening here. When I select GraphByService from the web page, I then select my service to graph and it graphs all hosts which match a particular volume structure, /, /tmp.data, /home/username, and /srv.data.

About ten servers match this structure and their graphs are displayed. At the top right of the page is a + symbol which opens out a selection list where i can select data sets however I can't clear the pre-loaded sets. For the servers not displaying graphs, their disk structures differ. How can I alter the displayed data sets?

I think the real problem here is that the data sets measured are dynamic and change for each server, the graph by server list just seems to look at the first server and assume that all graphs for that service will have the same data sets.

The out message from server1 for the service (custom check disk) is

DISK OK - '/'=45% '/tmp'=1% '/srv'=59% '/home/myname'=9% '/srv/backup'=52%

perfdata is '/'=45% '/tmp'=1% '/srv'=59% '/home/myname'=9% '/srv/backup'=52%

The output from the second server has the same disk structure as the first and therefore the same datasets and consequently the same rrd files for the service check. The third server outputs this perfdata:

'/'=35% '/tmp'=1%

The graph list for this service (Graphs By Service) complains that the third server is missing 3 rrd file names which correspond with the datasets not included on the third server. Only this page is affected.

This link looks like the problem. https://sourceforge.net/p/nagiosgraph/d ... /4d42186b/

Can you try and duplicate the service check with a slightly different name and see if that works.

Also, take a look at this link and see if creating a Service Type would work.
https://sourceforge.net/p/nagiosgraph/g ... EADME#l734

From the same document, this might be needed to make this work. Currently, my rrd files follow the second pattern, in different files.

993: Some people prefer to put all data from a single service into a single RRD
file, even if the data have different units. For example, for the PING
service their RRD files look something like this:

PING___ping.rrd (losspct, losswarn, losscrit, rta, rtawarn, rtacrit)

Others prefer a separate file for each data source:

PING___losspct.rrd (losspct)
PING___losswarn.rrd (losswarn)
PING___losscrit.rrd (losscrit)
PING___rta.rrd (rta)
PING___rtawarn.rrd (rtawarn)
PING___rtacrit.rrd (rtacrit)

As Tom mentioned - what happens if you duplicate the service check, but with a different name?

rkennedy wrote:As Tom mentioned - what happens if you duplicate the service check, but with a different name?

It will have to wait until Monday and I am reluctant to play too much as it in in a puppet system which controls 30 servers. I will bet that any service check with dynamic label names will have the same issue as described in a link I posted above.

A possible solution would be to store all the data for this check in a single rrd file but I can't find out how to do this.

yes this is a limitation of nagiosgraph it assumes that a service with a specific name is the same for all hosts essentially
nagiosgraph has not been updated in 4 years


so I think two easy options immediately come to mind

1) sounds like you have two different partition / mount point structures, so perhaps you can get away with two different service names
2) dont use nagiosgraph, use perhaps pnp4nagios instead


id suggest option 1 would be the best way to do it as you already have nagiosgraph operating etc..

nozlaf wrote:so I think two easy options immediately come to mind

1) sounds like you have two different partition / mount point structures, so perhaps you can get away with two different service names
2) dont use nagiosgraph, use perhaps pnp4nagios instead


id suggest option 1 would be the best way to do it as you already have nagiosgraph operating etc..

I tend to agree. I generally recommend one service per mountpoint anyway, as this gives you more granular control over notifications.

We have about 3 dozen servers with many different mount points. It would not be optimal or at least easily maintainable to have fixed mount point checks for each server. I'm wondering if putting all the perfdata into a single RRD file with a different template would help. Then at least each server has the same RRD file structure.

markhorrocks wrote:I'm wondering if putting all the perfdata into a single RRD file with a different template would help. Then at least each server has the same RRD file structure.

not sure im out of ideas without more information like the actual perfdata logs which box293 suggested or the plugin which you made i'm not sure how I can help further good luck hope you get a solution

nozlaf wrote:more information like the actual perfdata logs which box293 suggested or the plugin which you made

Agreed, the perfdata logs hold a lot of answers.