I've heard rumors that this will be coming in a future version of zenoss (but 
don't quote me on that).

However, I already have this working on my ZenOSS installation.

If you can't wait and want to see aberrant behaviour working right now, you 
will need to define custom commands to create the graphs and custom commands to 
display the graphs. This is done within a template.

So, as an example: in your template->data source->data point->Create cmd field 
add something like this:-


Code:

RRA:AVERAGE:0.3:1:8640
RRA:AVERAGE:0.3:3:17280
RRA:AVERAGE:0.3:12:17532
RRA:HWPREDICT:1440:0.1:0.0035:288




This will ensure that your RRD file is created with all the requirements to 
enable, Holt-Winter aberrant behaviour detection. The aberrant detection RRAs 
will be filled in automatically when the normal rrdupdate is run by ZenOSS.

 Note the custom create command will not run if the rrd file already exists, so 
either start with a new datasource or manually delete/rename the existing rrd 
file

explanation:-
RRA:AVERAGE:0.3:1:8640    <= standard cycle stored for 30days
RRA:AVERAGE:0.3:3:17280  <= 3*standard cycle stored for 180 days
RRA:AVERAGE:0.3:12:17532  <= 12*standard cycle stored for 2 years
RRA:HWPREDICT:1440:0.1:0.0035:288  <= enable RRAs for aberrant detection, 1 day 
seasonal store for 5 days

See the rrdcreate man page and 
http://cricket.sourceforge.net/aberrant/rrd_hw.htm for info on these commands

To actually see the prediction lines and aberrant detection in your graphs you 
will need to add a custom graph definition in your template -> graph -> custom 
field.

This is the tricky bit as you have to reference the data source in your custom 
command, so this will be different for each graph. Also, you will need to use 
different commands to graph more than one data source.

Here is an example of what I use:-

For a graph with a single data source use the following template (replace 
**ONE** with the exact name of your datasource e.g. ifHCInOctets_ifHCInOctets; 
replace **ONELABEL** with any label for the graph item being graphed)


Code:

DEF:pred=/opt/zenoss/perf${here/rrdPath}/**ONE**.rrd:ds0:HWPREDICT
DEF:dev=/opt/zenoss/perf${here/rrdPath}/**ONE**.rrd:ds0:DEVPREDICT
DEF:fail=/opt/zenoss/perf${here/rrdPath}/**ONE**.rrd:ds0:FAILURES
CDEF:upper=pred,dev,2,*,+
CDEF:lower=pred,dev,2,*,-
TICK:fail#ffffa0aa:1.0
AREA:**ONE**#00cc00:**ONELABEL**
VDEF:**ONE**_last=**ONE**,LAST
GPRINT:**ONE**_last:cur\:%0.2lf%s
VDEF:**ONE**_average=**ONE**,AVERAGE
GPRINT:**ONE**_average:avg\:%0.2lf%s
VDEF:**ONE**_maximum=**ONE**,MAXIMUM
GPRINT:**ONE**_maximum:max\:%0.2lf%s\l
LINE1:upper#ff4444:Baseline Upper 97.5th Percentile
LINE1:lower#ff8844:Baseline Lower 2.5th Percentile




Once you have done this, sit back and wait 3*seasonal periods (3 days in my 
example) before you see the extra lines representing the baseline, as rrd 
cannot predict anything until it has enough data. And when your done you should 
see something like this:-

[img]
http://cricket.sourceforge.net/aberrant/aberrant.png
[/img]

The only thing I can't work out is how to alert whenever aberrant behaviour is 
detected. However, having used this for a while now I actually find it useful 
as post-analysis thing and don't think it would be that useful for alerting 
unless you are running a super-stable super-predictable network(!).

=====
Surfraz
Open Systems Intelligence
http://www.go-osi.com

------------------------
Surfraz Ahmed




-------------------- m2f --------------------

Read this topic online here:
http://community.zenoss.com/forums/viewtopic.php?p=10340#10340

-------------------- m2f --------------------



_______________________________________________
zenoss-users mailing list
[email protected]
http://lists.zenoss.org/mailman/listinfo/zenoss-users

Reply via email to