So far (see below) I've found that the 3 xmls in slech's zip file (Page 9 of this thread), work well - at least as well as my original's, though you DO still need to associate the data items AFTER creating the graphs. I've attached this zip to this post. {Edit: I have new versions with a little tweak to them almost ready to upload}
- Pink variant
- Multi_CPU_Example.PNG (54.52 KiB) Viewed 73063 times
Quick tutorial on how to use/customise:
1. Import the template in cacti
-- Goto the console pane
-- Click on Import Templates
-- Choose the file (template) to import
-- Click Import
2. Create Graphs
-- Goto console pane
-- Goto Devices
-- Choose your device (host)
-- Click on Create *Create Graphs for this Host
-- Choose from the drop down in the "Graph Template Section" the relevant graph - "HOST MIB - MULT CPU utilization on X processor box" - Where X is the number of CPUs in your device
-- Click Create
-- on the next screen you will asked the index-value/type ... just click on save and its not over yet !!
3. Get the Graphs working
-- Goto Console
-- Goto Graph Management
-- locate the graph you just created, click on it
-- in "Supplemental Graph Template Data” select the cpu you wish to graph on the multi-cpu-graph, for each data source item, from the relevant dropdown. This is most likely HOSTNAME - CPU Utilization - CPU0 (cpu), HOSTNAME - CPU Utilization - CPU1 (cpu) and so on ...
-- (optional - : you can now delete the initial graphs that you created in step 2 but when asked be sure to keep the data-sources intact because the multi-cpu-graph needs these data sources.)
4. A little bit of customisation
-- This bit just changes the colour of the Average (or MEAN) CPU utilization from yellow to PINK (as above), but you can choose any colour/opacity combination
-- Goto Graph Templates on the Console Tab
-- Click on the Multi-CPU Template you want to change
-- Click on Item #1
-- Choose the "Color" [sic] from the dropdown (I chose FF00FF)
-- Choose the "Opacity/Alpha Channel" from the dropdown (I chose 30%)
-- Click Save
Some of the above has been borrowed from this tutorial: http://www.goldfisch.at/knowwiki/howtos ... u_graphing, though I do take exception to point 6 - as far as I can work out the average IS calculated correctly my way. If you use the ALL_DATA_SOURCES_NODUPS special data source, you end up with a higher average CPU ute than you would expect - as it also includes ITSELF in the calculation - unless TPTB have fixed it since I wrote the original CDEF. Either way the original way does work, and IS correct (IMO).
More to come... Maybe... if I can pretty these things up (ponders if you do more than 8 should/could you do some going down as well as some going up ?). I may have worked out a way to shortcut some of the above...
All, I've just spent an hour trawling through this thread, and realised that, to a Cacti Newbie, or just someone wanting to do this for the first time, its a complete nightmare.... Everything above is new/subject to change
Thus this edit...
I'm at a new site, brand new CactiEZ install (Cacti 0.8.8a), with no Multi-CPU graphs. So I tried using Aggregate to create Multi-CPU, well it works, but its not pretty, and just not as good as my earlier efforts, so... I shall look at the templates we've all come up with get 3 (or 4) which actually work, maybe pretty them up a bit, and put them back here, and probably over at the repository - may start a new thread here too...
This MAY take a while, as I'll be doing it in my lunch hours...
(I've decided against using WMI unless I REALLY can't avoid it as I just kept coming up against performance problems when using it before, and most, if not all, of what I need is now exposed via SNMP).
OK - first thing - this template by jzondor: cacti_graph_template_multi_cpu_utilization_on_4_processor_box_v2a.xml only has stuff to do with Cisco 3000 in it - does anybody know why this is?
----------------------------------------------------------------------------------
Original Post Below this line
----------------------------------------------------------------------------------
Right, like many people I have many servers with multiple CPUs.
What I want to see is:
a graph which shows :
1) the average utilization over all CPUs (maybe as a light area in the "background")
2) the utilization of each CPU (perhaps as dark lines in the foreground).
That's it really...
Trying to work this out at the moment, worked out how to do part 2 on a case by case basis (as long as the individual data sources exist) - but now want to turn it into a template (using the convert to template function works but takes the data sources I used for the graph with it - instead of using the query...)
So how do you graph the average over 2 data sources? Create a new source?
Argon0
