Automatic Scaling

[ChrisAT]

Hi All,

I finally moved from 0.8.8h to 1.2.27 - and things are generally looking good - the only thing that confuses and drives me nuts is that while I do use one single Graphics template for temperature values from various temperature sources (all rely on the sam single data source template) the scaling for the respective graphs are looking pretty different.

A graph that is looking perfectly fine:

Good scaling

Good.JPG (53.48 KiB) Viewed 134 times

and one that isnt looking great:

Messed up scaling

NoGood.JPG (56.06 KiB) Viewed 134 times

How can I sort that out so all the graphs are looking reasonable?

Thanks a lot!

NB: Things are running on a Raspberry Pi, Raspbian Buster. RRD Version is 1.7.1-2

[macan]

Go to templates -> graph -> choose "SNMP - Generic OID Template" -> edit
You can change a lot of things here

[ChrisAT]

Thanks for the advice - while having spent quite some time on testing and trying to find a solution I didnt really succeed though.
I simply can't achive a truly reasonable scaling at little changes of values.
To me it seems if changes are getting close to a value of one the scaling doesn't work properly anymore...

So what have I tired?

- I started from scratch, installing a fresh new RaspberryPi OS "Bookworm" and updated all to the latest version
- Installed Cacti from the RaspberryPi repository with its dependencies and upgraded manually to the latest Cacti Version before running an apt update && apt upgrade again
- RRDTool installed is 1.7.2-4+b8
- Reconfigured Cacti manually
- Tested with SNMP - Generic OID Template as well

As previously the outcome for data sources (its 3 digits integer data, deviding by 10 returns the measured temperature with 1 decimal) is very much the same though:

Here the various options I've tried and the respective results:

--alt-y-grid \
--alt-autoscale \
CDEF:cdefa='a,10,/' \

yG&AS&DIV.jpg (51.76 KiB) Viewed 58 times

--alt-autoscale \
CDEF:cdefa='a,10,/' \

AS&DIV.jpg (55.92 KiB) Viewed 58 times

So by using the CDEF to get the "correct" Temperature values I can decide to either lacking detailed labels/values at the y axis (using -alt-y-grid)

or

having lots showing the same value (without -alt-y-grid)...




--alt-autoscale \
CDEF:None

AS.jpg (69.85 KiB) Viewed 58 times

--alt-y-grid \
--alt-autoscale \
CDEF:None

yG&AS.jpg (49.7 KiB) Viewed 58 times

So without using the CDEF I don't get the "correct" Temperature values but - if the bandwidth of values is sufficient - at least a graph with a reasonable amount of labels at the y-axis (without -alt-y-grid)
If the bandwidth of values is NOT sufficient - like eg having a range that only varies by +/- 0.1 degrees also this approach shows lots of labels at the y-axis with the same value (no example here)

or

again a graph lacking detailed labels/values at the y axis (using -alt-y-grid)


Based on this it looks to me like using the -alt-y-grid option simply doesnt work properly anymore - at least compared to 0.8.8h where I remember having changed things in the code a tiny bit (based on https://forums.cacti.net/viewtopic.php?t=21086 and https://www.grosseosterhues.com/2009/07 ... ned-input/).
(Tried the same now as well, but as expected that didnt change anything compared to the option set at graph templates)

Still not sure how I can tackle this. For now the most reasonable workaround is by omitting the CDEF and live with the "incorrect" values at the y-axis (which isnt bight either as I cant necessarily predict the value bandwidth of the respective data source)...

Any further advice would be highly appreciated...

[TheWitness]

Somehow I suspect RRDtool may be to blame.