Enhancement #378

avatar

add more options to curves from RRDTool (was: graphs shows wrong values)

Added by Nikolaus Filus 527 days ago. Updated 91 days ago.

Status:New Start:
Priority:Normal Due date:
Assigned to:- % Done:

0%
Category:Centreon - Graphs
Target version:Centreon-2.3
Resolution:

Description

Hello,
as you can see on the attached graph and the corresponding values, the graph is off by +5 or so. If this should be by design or by misconfiguration, so please explain and document this behavior.

graph.png (20.7 KB) Nikolaus Filus, 02/17/2009 03:41 PM

RMS_RMS_Sensor_Temp1.csv (755 Bytes) Nikolaus Filus, 02/17/2009 03:42 PM

rrdtool_fetch.txt (1.7 KB) Nikolaus Filus, 02/17/2009 04:26 PM

History

Updated by Nikolaus Filus 527 days ago

avatar

I now discovered the values on graph match the scaled raw rrd values (also attached). But this still doesn't explain why rrd values are so different than the raw sensor values stored in centstorage.

Updated by Nikolaus Filus 491 days ago

avatar
I now digged deeper into rrdtool and it's configuration. My problem can be "fixed" by allowing more configuration of rrdtool.
In bin/centstorage all rrd files are created like:
create temperature.rrd -s $interval \
DS:$metric_name:GAUGE:$interval:U:U
RRA:AVERAGE:0.5:1:$my_len_storage_rrd
RRA:AVERAGE:0.5:12:$my_len_storage_rrd

and in www/include/views/graphs/graphODS/generateImages/generateODSImageZoom.php the graphs are created by:
DEF:v".$cpt."=".$RRDdatabase_path.$key.".rrd:".substr($metrics[$key]metric,0 , 19).":AVERAGE

Now I would ask to append an optional ":step=XXX" parameter, so that the average function can be flattened. This is useful when a metric was created with a big interval, which was shortened later. Then the rrd data is still averaging over the old interval.

Updated by Nikolaus Filus 312 days ago

avatar
  • Category deleted (CentStorage)
  • Priority changed from High to Normal
  • Target version changed from Centreon-2.1 to Centreon-2.2

Updated by Nikolaus Filus 181 days ago

avatar
  • Category set to Centreon - Graphs

Updated by Julien Mathis 160 days ago

avatar

Is this bug still open ?

Updated by Nikolaus Filus 160 days ago

avatar

I have to take a look at this again, but I would say "yes". I saw some re-calculation work done on the new centstorage, but for the perl version this should still apply ...

Updated by Julien Mathis 97 days ago

avatar

Hi Niko,

What I have to change on centstorage for fixing your problem ?

Regards

Updated by Nikolaus Filus 97 days ago

avatar

Hi Julien,

the original problem from then was:
I started to collect the values with interval of 5, then I changed to interval of 1. After this change the graphs were still graphed as an average over 5 minutes, although the newest data was already provided at a higher resolution.

Possible solutions:
  • save rrds always at highest resolution as defined in nagios global (default:60s), then graph with check_interval resolution
  • mark RRD for rebuild, when check interval changes (and notify user) [NOTE: due to bug #403 I lost all my data back then!]
  • always append ":step=".$checkinterval parameter to the command line in image generation. Compare: http://oss.oetiker.ch/rrdtool/doc/rrdgraph_data.en.html

Which one (or all) to choose and implement is up to discussion.
In long term we need to introduce more RRA types then just AVG! See enhancement #1555 for a proposition.

Hope that helps to understand the problem.
Regards

Updated by Julien Mathis 97 days ago

avatar

I have developped for client needs a script who check id the RRDTool DB is well configured compared to nagios interval. If it's not the case, the RRD db will be rebuild.

Is it what you want ?

Updated by Julien Mathis 91 days ago

avatar
  • Target version changed from Centreon-2.2 to Centreon-2.3

Also available in: Atom PDF