Re: Are all_max_peak percentile calculations wrong? - bug report pending

On Sat, 10 Nov 2012 17:59:19 +0100
Gandalf <[email protected]> wrote:

> It would be a good move to open a bug report to help tracking this issue
> (http://bugs.cacti.net).
> In case you have an example (e.g. set of rrd files and a graph or the like) this
> would help reproducing this issue and verifying the fix.

Thanks, I intended to do so anyway if nothing results from my 
posting here. I tried the list and forum first as I thought the 
Cacti project prefers discussion before opening a bug report if 
the reporter is not sure whether there is an error. I will open
a bug report shortly.

I do not have example RRD files ready. The two methods do have
different results for rare resp. special values only. I found 
it difficult to verify the correctness of the calulations by 
looking at an RRDTool/Cacti graph. I did use the following 
definitions to print line and number for the percentiles:
Item # 10 	HRULE: |95:bits:0:max:2| 	
Item # 11 	COMMENT: |95:bits:6:max:2| mbit p95( max(traffic_in,traffic_out) )
Item # 12 	HRULE: |95:bits:0:all_max_peak:2|
Item # 13 	COMMENT: |95:bits:6:all_max_peak:2| mbit max( p95(traffic_in),p95(traffic_out) )
Using this you can see wether the calculations differ or not
for a given graph.

I will include the last paragraph into the bug report.


Regards
Florian


-- 
  Florian Roscher                          private: [email protected]
                                            Debian: [email protected] 
PGP Key / ID: 1024D/B4071A65
Fingerprint : F9AB 00C1 3E3A 8125 DD3F  DF1C DF79 A374 B407 1A65

------------------------------------------------------------------------------
Keep yourself connected to Go Parallel: 
TUNE You got it built. Now make it sing. Tune shows you how.
http://goparallel.sourceforge.net