Cheshire Cat Computing

Hello all,

I apologize if my question has been asked and/or answered before. Current situation is an older mrtg, running on a 32bits linux box. This box was replaced by a 64bits virtual linux box running the latest version of mrtg. Several environment variables have therefore changed. MRTG runs fine, but I want to implement threshold usage for warnings using a perl script. Rrdtool is used for graphs.
When testing out threshold breaches, the values as seen by mrtg make no sense at all. It is unreliable, sometimes using thresholds, sometimes ignoring them completely.
Could this be related to the fact that we're now running on 64bits linux? When I set mrtg to use it's own log format, it works perfectly.

Thanks in advance for any tips or pointers,

Servaas Schrama

On the 64-bit box, are all graphs appearing correctly when in RRDTool mode? If not, it could be an RRDTool issue.

Can you let me know what versions of MRTG and RRDTool (and OS) you are using? This helps identify problems.

In any case, I'd strongly suggets you upgrade MRTG to 2.16.x as there were many changes and fixes before then. RRDTool should be at least 1.2.30 and preferably 1.3.6 or later.

_________________
Steve Shipway
UNIX Systems, ITSS, University of Auckland, NZ
Woe unto them that rise up early in the morning... -- Isaiah 5:11

Hi Steve, all,

We are currently running 2.16.3. I have compared tests on a 32bits machine and the new 64bits (virtual) machine and found no differences yet. However, I have found something else that seems to be key in our case.
If rrdtool is set as logformat, threshold checking is unreliable. If logformat is set to mrtg (actually, just hashed out making it use the default format), threshold checking works perfectly. Must have something to do with the rrd format, as thresholds are only checked after the log has been written.

Is this *known* behavior? I have used mrtg before, but with less emphasis on the threshold functionality, as I believe most people do. I am considering going for nagios implementation on top of mrtg for the handling fo threshold breaches, as this software seems to be (better) tailored for that.
We're looking at some 15k targets to monitor, all snmp, so we need something reliable that will warn if breaches occur.

Thank you for your response - all kinds of input, thoughts and tips are appreciated!

Kindest regards,

Servaas Schrama

This could well be an RRDTool problem -- which version of RRDTool are you running? You should upgrade to a later version if possible (1.3.4 or later)

_________________
Steve Shipway
UNIX Systems, ITSS, University of Auckland, NZ
Woe unto them that rise up early in the morning... -- Isaiah 5:11

Thank you for your input and idea. Right now, I have found that the threshold checking actually works ok on the new platform. However, I am running into another issue.
Some of the routers I talk to have up to 40k targetlines. Running cfgmaker takes a long time, and I am using some template snippet to insert threshold definitions in the configfile. I haven't been able to run this successfully yet, the template simply times out after a while leaving me with defunct snmpget processes.
However, running cfgmaker without the extra threshold definitions seems to work better.
So, cfgmaker seems to do it's job fine. I guess I'd best modify cfgmaker to add the threshold definitions, but that will lead to trouble when new versions of cfgmaker appear.
This is quite a challenge!