[rrd-users] RRD GAUGE problem

Sat May 18 09:18:00 CEST 2013

Thanks for your efforts, I found culprit:

I created these rrd files in directory, for which there was cron which
builds graphs, to make me easier see changes.
I found that if  I create database file elsewhere, all was just as in your
case. I dug further and found out that there is "historically" one more
script, which periodically tries to get data from old place, and if there's
no data just putting "0" into all rrd.
I turned off this old script, and now I have perfectly line with given
value, no matter how un-periodically this happen.

Thanks again for you patience :)

2013/5/17 Donovan Baarda <abo at minkirri.apana.org.au>

> With the following script I get the following expected graph. If you don't
> get that graph, then something is wrong with your system.
>
> 
>  test.sh<https://docs.google.com/file/d/0B1T6TeYwG0lodWE4S0I5NGk0SE0/edit?usp=drive_web>
> 
>  testing2.png<https://docs.google.com/file/d/0B1T6TeYwG0lodTUxVWg0alVSSDQ/edit?usp=drive_web>
> 
> Without seeing the exact script you are running, it's hard to know if
> there is some subtle thing that you are doing that is causing the problem.
>
>
>
> On 16 May 2013 22:16, Alexander Yerenkow <yerenkow at gmail.com> wrote:
>
>>
>> Donovan!
>> With your script I got all values set to 5, as I can  see both in dump
>> and graph.
>> 1. But from your email I didn't got clear info that you tried to run
>> script with delays and "N:", is that correct?
>>
>> 2. If I simply put in `for` cycle "sleep 63" and change update not to use
>> calculated time, but instead N:5, I got these data seen in dump:
>>
>> <!-- 2013-05-16 23:11:00 EEST / 1368735060 --> <row><v>NaN</v></row> │
>> <!-- 2013-05-16 23:12:00 EEST / 1368735120 -->
>> <row><v>1.7920090000e+00</v></row> │
>> <!-- 2013-05-16 23:13:00 EEST / 1368735180 -->
>> <row><v>3.5915094167e+00</v></row> │
>> <!-- 2013-05-16 23:14:00 EEST / 1368735240 -->
>> <row><v>3.8415422500e+00</v></row> │
>>
>> Here's my for cycle (for clarification):
>>
>> for ((t=$start; t<=$start+60*63; t=t+63 )); do
>> echo $t
>> rrdtool update testing3.rrd N:5;
>> sleep 63
>> done
>>
>> I see some weird bug when N used, that's at least.
>> Can anyone confirm present or absent of such behavior?
>> Thanks.
>>
>>
>>
>>
>> 2013/5/16 Donovan Baarda <abo at minkirri.apana.org.au>
>>
>>> I don't get that result at all... I'm using RRDtool 1.4.7 on a Debian
>>> system.
>>>
>>> Are you sure you are using GAUGE and not something like ABSOLUTE?
>>>
>>> Some minor funnies in what you are doing is using step "-s 60" when
>>> creating your rrd, and -S 1 when generating your graph. This means your
>>> data is only resolution 60, and you are asking it to give you a 1 second
>>> resolution graph. However, that should not cause what you are seeing.
>>>
>>> In your graph cmd you have "CDEF:realcnt=cnt,1,*
>>>  LINE:realcnt#00A000:Cnt", which means you have an unnecessary multiply by
>>> 1 and you could just use; "LINE:cnt#00A000:Cnt".
>>>
>>> Try running the attached script and having a look at the testing2.png
>>> output. I definitely don't get any strange stepping. If you do, then I'd
>>> suspect something strange on your platform.
>>>
>>>
>>>
>>> On 16 May 2013 18:40, Alexander Yerenkow <yerenkow at gmail.com> wrote:
>>>
>>>> This is my command.
>>>>
>>>> rrdtool create -s 60 testing2.rrd DS:cnt:GAUGE:120:U:U
>>>> RRA:AVERAGE:0.5:1:1440 RRA:AVERAGE:0.5:60:720 RRA:AVERAGE:0.5:1440:365
>>>>
>>>> graph:
>>>>
>>>> rrdtool graph testing2.png -s -1200 -S 1 -a PNG -w 1000 -h 250 -v
>>>> "testing" DEF:cnt= testing2.rrd:cnt:AVERAGE CDEF:realcnt=cnt,1,*
>>>>  LINE:realcnt#00A000:Cnt
>>>>
>>>> Even if I got problem in graph, how do I get not "5" in dump:
>>>>
>>>> <!-- 2013-05-16 16:49:00 EEST / 1368712140 -->
>>>> <row><v>4.8531933333e+00</v></row>
>>>> <!-- 2013-05-16 16:50:00 EEST / 1368712200 -->
>>>> <row><v>0.0000000000e+00</v></row>
>>>> <!-- 2013-05-16 16:51:00 EEST / 1368712260 -->
>>>> <row><v>1.9154833333e-01</v></row>
>>>> <!-- 2013-05-16 16:52:00 EEST / 1368712320 -->
>>>> <row><v>4.4032950000e-01</v></row>
>>>> <!-- 2013-05-16 16:53:00 EEST / 1368712380 -->
>>>> <row><v>6.8845091667e-01</v></row>
>>>> <!-- 2013-05-16 16:54:00 EEST / 1368712440 -->
>>>> <row><v>9.4017525000e-01</v></row>
>>>>
>>>> Could anyone *please* run this test?
>>>> create testing db, and gather there data with my script (with real
>>>> delays one), and not via dumping into file a lot of data in just moment...
>>>>
>>>> I'll need or confirmation that it's somehow   my only problem, or we'll
>>>> get know that problem elsewhere.
>>>> Thanks.
>>>>
>>>>
>>>>
>>>> 2013/5/16 Donovan Baarda <abo at minkirri.apana.org.au>
>>>>
>>>>> What does your rrdtool create and graph lines look like? I suspect
>>>>> that's where the error lies.
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> On 16 May 2013 16:01, Alexander Yerenkow <yerenkow at gmail.com> wrote:
>>>>>
>>>>>> Thanks for hint, but this is working only when invoking at moment of
>>>>>> time.
>>>>>> E.g. If I run script
>>>>>>
>>>>>> tt=`jot - 1 30`
>>>>>> for i in $tt
>>>>>> do
>>>>>> t=`date +"%s"`
>>>>>> rrdtool update testing.rrd $t:5; rrdtool update testing2.rrd N:5;
>>>>>> sleep 63;
>>>>>> done
>>>>>>
>>>>>> I still  see same seesaw graph (I changed from 61 to 63 to make
>>>>>> problem more visible).
>>>>>> The script was run on fresh created two files, according your
>>>>>> command.
>>>>>> The problem  (if there is problem) lies somewhere in some rounding
>>>>>> time, as it seems appear in both files, when I specify N, and time.
>>>>>> Here's my fresh graphs:
>>>>>> https://javaz.org/wp-content/testing.rrd-avg.png
>>>>>> https://javaz.org/wp-content/testing2.rrd-avg.png
>>>>>>
>>>>>> Could you confirm that if you run my script (with real half-hour
>>>>>> datamining) you will have same good-looking graph?
>>>>>>
>>>>>> BTW, I'm using rrdtool-1.4.7.
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> 2013/5/16 Donovan Baarda
>>>>>>
>>>>>>> G'day Alexander,
>>>>>>>
>>>>>>> You need to read http://oss.oetiker.ch/rrdtool/doc/rrdcreate.en.html,
>>>>>>> in particular the bit towards the end about "The Heartbeat and the Step".
>>>>>>> In particular, heartbeat on the DSS is the *maximum" time between samples,
>>>>>>> otherwise they are thrown away and recorded as unknown. You can think of
>>>>>>> step/heartbeat as the "xff" of the input data. In your case I'm guessing
>>>>>>> using step=60 and heartbeat=120 is probably closest to what you want.
>>>>>>>
>>>>>>> I looked at your seesaw-graphs blog entry and the graph you have
>>>>>>> looks like a kind of artefact of using a low heartbeat (resulting in very
>>>>>>> few valid inputs) combined with a low xff or maybe graphing options to
>>>>>>> interpolate the missing values. When i do the following;
>>>>>>>
>>>>>>> $ rrdtool create -s 60 testing.rrd DS:cnt:GAUGE:120:U:U
>>>>>>> RRA:AVERAGE:0.5:1:1440 RRA:AVERAGE:0.5:60:720 RRA:AVERAGE:0.5:1440:365
>>>>>>> $ t=`date +"%s"`
>>>>>>> $ for ((i=$t; i<$t+60*61; i=$i+61)); do rrdtool update testing.rrd
>>>>>>> $i:5; done
>>>>>>> $ rrdtool graph testing.png -s $t -S 60
>>>>>>> DEF:cnt=testing.rrd:cnt:AVERAGE LINE1:cnt#ff0000:cnt
>>>>>>>
>>>>>>> I get the following graph;
>>>>>>>
>>>>>>> <http://minkirri.apana.org.au/~abo/testing.png>
>>>>>>>
>>>>>>> Which is probably what you expect.
>>>>>>>
>>>>>>>
>>>>>>> On 15 May 2013 11:03, Alexander Yerenkow <yerenkow at gmail.com> wrote:
>>>>>>>
>>>>>>>> Hello list!
>>>>>>>> I'm having problem with GAUGE and non-regular updates of values, I
>>>>>>>> described my problem fully here:
>>>>>>>>
>>>>>>>>
>>>>>>>> https://javaz.org/2013/05/rrdtool-strange-rounding-produces-seesaw-graphs/
>>>>>>>>
>>>>>>>> I'd appreciate if you point me to right place where I find why I
>>>>>>>> can't get what I need.
>>>>>>>>
>>>>>>>> I already tried to tune heartbeat value, but to no avail.
>>>>>>>>
>>>>>>>>
>>>>>>>> --
>>>>>>>> Regards,
>>>>>>>> Alexander Yerenkow
>>>>>>>>
>>>>>>>> _______________________________________________
>>>>>>>> rrd-users mailing list
>>>>>>>> rrd-users at lists.oetiker.ch
>>>>>>>> https://lists.oetiker.ch/cgi-bin/listinfo/rrd-users
>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> --
>>>>>>> Donovan Baarda
>>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> --
>>>>>> Regards,
>>>>>> Alexander Yerenkow
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> Donovan Baarda <abo at minkirri.apana.org.au>
>>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> Regards,
>>>> Alexander Yerenkow
>>>>
>>>
>>>
>>>
>>> --
>>> Donovan Baarda <abo at minkirri.apana.org.au>
>>>
>>
>>
>>
>> --
>> Regards,
>> Alexander Yerenkow
>>
>
>
>
> --
> Donovan Baarda <abo at minkirri.apana.org.au>
>

-- 
Regards,
Alexander Yerenkow
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.oetiker.ch/pipermail/rrd-users/attachments/20130518/dafb6b3f/attachment-0001.htm 