Please help ! Need to graph custom SNMP OID

[gandalf]

Thank you both so much for your suggestions. Reinhard you were right I did have to manually make the graph (i thought making the graph template would do this but my understanding of graph templates was off a bit...). Made a graph and did a test call to the dialup shelf to see if graphing works and sure enough it did but it did not give me a "1" like I expected to represent 1 call. Instead it is showing on the left side "0.0 through 3.0 m" and on the legend it showed the current value as "3.32m" but snmpget is returning a value of 1 to represent 1 call.

Sorry to be a pain guys but again thank you for your assistance I used MRTG many years ago and also Concord eHealth ($$$$$), and I think Cacti is the by far a much better application.


Thank you again!!


-Rainman

The data source type is GAUGE, then? If not, please change the rrd file using rrdtool tune
Reinhard

[gandalf]

This is what I did (in pictures):

That's not correct. But to make suggestions I need to know your poller interval: is it 1 minute or 5 minutes?
Then, the "exact number" trick is "nice" but not correct. You may do some to come closer. But with large numbers, you may miss the value returned by snmpget by a few "points".
This is due to normalization of rrdtool. This happens, if rrdtool updates occur "off poller interval", e.g. by 5 sec off. Assuming a 300 sec poller interval, rrdtool then will divide by 305 and multiply by 300 to calculate the value that will be store in the file. There's no chance to switch this off. See more discussion of this basic rrdtool feature in the rrdtool-users mailing list on documentation.
Reinhard

[IlPini]

Edit: clarified role of the poller (my previous sentence was wrong).

Edit: clarified per-minute (calculations were wrong); added per-hour.

Made a graph and did a test call to the dialup shelf to see if graphing works and sure enough it did but it did not give me a "1" like I expected to represent 1 call. Instead it is showing on the left side "0.0 through 3.0 m" and on the legend it showed the current value as "3.32m" but snmpget is returning a value of 1 to represent 1 call.

Mathematically correct. You just have to get "in the mental groove", so to speak.

You make a call. The value goes up by 1.

The poller polls, and Cacti stores the value which will go into the calculations for the graph.

The value changed by 1 modem during one polling interval of approximately 300 seconds. 1 modem in 300 seconds is 0.00333... modems/second, or 3.33... millimodems/second, which is approximately the "3.32m" you got.

The "m" indicates "milli-", the way the "M" in MB indicates "mega-" in megabytes.

If you make 2 phone calls during one single polling interval, you should get approximately 6.67m in your graph, which is 0.00666... modems/second.

If you somehow manage to make more than 300 calls in five minutes, the graph will go higher than 1 (one). You need 3000+ calls in five minutes to go higher than 10 (ten). And so on. I guess it'll take a lot of friends to test this.

So, basically you're graphing "average calls per second".

Note that you can easily graph avg. calls/minute by multiplying your raw data by 60. That's what I did when graphing printer page output, which is traditionally measured in pages/minute.
Multiply by 3600 and you get avg. calls/hour.
But please also note that while that's acceptable practice for printer speed, that could be an unacceptable distortion of meaning for your data (even one single call in a whole week would get translated to 12 average calls per hour in that single five-minute interval).

Whether or not this mathematical representation suits your needs depends on you and your boss, I guess. Once I got the hang of it, I found myself at ease with it.

[gandalf]

That's a fairly well done description. Usually, I do not comment correct user responses. But this one really deserves a big fat "WELL DONE".
Reinhard

[Nybble]

{Background: I'm an intern, and we are attempting to do some graphing with our printers. Just started using Cacti a few days ago, and our senior Sys Admin has more pressing tasks}

I'm attempting a similar project, where I am graphing the total number of pages a printer has printed in its lifetime, and the delta in 1 (or 5? Still new) minute intervals. This is being accomplished with GAUGE and COUNTER respectively.

I am getting the same problem with the data being pages/second. However, I am not aware if I can manipulate the data as Ilpini refered to. I am grabbing the value by OID, but do not see anywhere that I can change this data. It works just fine in GPRINT comments below the graph, but the actual graph itself has (relatively) nonsense numbers. Is there a way to massage these numbers or change the axis to show the correct numbers?

Random thought, does the Step value have anything to do with it? If I increased/decreased it would I get more accurate data? (I have no problem with polling the printers a little bit more, I do not grab nearly as much data from them)

[gandalf]

I recommend reading some rrdtool stuff concerning STEP. You ARE quite right. The means to change numbers is called CDEF
Reinhard