From: "Marcelo A. Ferreira Gomes" <mgomes@trip.com.br>
(the original thread name was "cartinha p/ lista cricket", an improper
subject from a posting of mine. It should have been "Zeroed out graphs"
from the start.
I'm trying to restart this thread, since the problem is still unsolved to me)
On Mon, 3 May 1999 11:44:25 +0200, Tobias Oetiker <oetiker@ee.ethz.ch> writes:
>Hi Jeff Allen,
>
>Today you sent me mail regarding [MRTG-DEV] Re: [cricket-users] Re:...:
>
>*>
>*> Marcelo A. Ferreira Gomes wrote:
>*> > But sometimes, some of the graphs will only show zeroes (a flat
>horizontal
>*> > line covering the x axis). For instance, the hourly and weekly are zeroed
>*> > out and the daily and monthly are ok. Sometime later (after another
>sample
>*>
>*> You are not going nuts. We have seen the same problem here. I have
>*> tried to reproduce it but failed. I am almost 100% certain it's a
>*> problem with RRD Tool, not Cricket. The problem seems to be
>*> time-based, and the options that Cricket sends to rrdgraph do not vary
>*> over time.
>*>
>*> Tobias, any word on this bug? Have you seen it? I'll try again to
>*> characterize the situations where it happens, though it's very hard to
>*> do.
>
>I have not heared from this problem yet ...
>sorry
>tobi
and then
On Wed, 5 May 1999 11:27:01 +0200, Steen Linden <Steen.Linden@ebone.net>
writes:
>Hi Tobi,
>
>Sorry, but I haven't been able to isolate it yet and can't come up with a
>reproducable scenario.
>
>I'm using rrdtool-0.99.25 (also tried rrdtool-0.99.22 with same result),
>cricket-0.63 and perl 5.004_04 on FreeBSD 2.2.6-RELEASE.
>
>The RRDs call for the cricket "hourly graph" look like this:
>
>RRDs::graph /tmp/graph.gif --base 1024 --start -151200
>--vertical-label bits per second --width 500 --height 200
>DEF:ds0=/cricket-data/hqgw/ebone-inc.rrd:ds0:AVERAGE
>DEF:ds1=/cricket-data/hqgw/ebone-inc.rrd:ds1:AVERAGE
>CDEF:sds0=ds0,8,* CDEF:sds1=ds1,8,* AREA:sds0#00cc00:Average bits in
>LINE1:sds1#0000ff:Average bits out VRULE:925423200#ff0000:
>
>Getting rid of the CDEFs makes it work all the time.
>
>--Steen
>
>
>On Wed, 5 May 1999, Tobias Oetiker wrote:
>
>> From: Tobias Oetiker <oetiker@ee.ethz.ch>
>>
>> Hi Steen Linden,
>>
>> Today you sent me mail regarding RE: [MRTG-DEV] Re: [cricket-users] Re:...:
>>
>> *>
>> *> I also believe it has something to do with the RRD scaling. I only see
>> *> this problem on graphs converting bytes/s to bit/s, but then in about 50%
>> *> of the time. A new poll makes the problem go away for a while.
>> *>
>> *> When removing the scaling the problem disappears. Seems to be time
>>related
>> *> or maybe due to something uninitialized.
>>
>> what is this scaling business about ? are you using the latest rrdtool (.25
>> or .26) ?
>>
>> cheers
>> tobi
then again
On Thu, 6 May 1999 14:34:39 +0200, Steen Linden <Steen.Linden@ebone.net>
writes:
>On Thu, 6 May 1999, Tobias Oetiker wrote:
>>
>> in that case I guess the FPE matter is rearing it's ugly head agin
>> can you try to recompile rrdtool and make sure that all components are
>> compiled with -ieee
>>
>
>Hi Tobi,
>
>Changing the CDEF definitions by extending the last CDEF with a null
>operation makes it work. Maybe this rings a bell, maybe it just adds to
>the weirdness.
>
>"CDEF:sds0='ds0,8,*' ".
>"CDEF:sds1='ds1,8,*,0,+' ".
>
>I suspect the bug to be in or around rrd_graph.c:data_calc(), since this
>function is both dependant on the CDEFs and the time, but I haven't been
>able to find it yet.
Well, it seems to solve the problem only for you, Steen, since I didn't see
any difference when I added the ",0,+" to my Defaults file.
BTW, that file is the only place where I have CDEF definitions. I don't
make them explicitly, like you seem to be doing.
Also, since I'm using grapher.cgi almost as it was shipped (with a few
minor modifications that are of no general interest), I am using RRDs, not
RRDp, so that the problem couldn't possibly come from rrd_graph.c. The
piece of software in RRDs equivalent to rrd_graph.c in RRDp is RPN.pm.
I have checked that code myself (RPN.pm) and it looks fine to me.
Does anyone have any clue not mentioned about this problem yet? Does it
show only to me, Steen Linden and Jeff Allen?
Thanks in advance.
-- Marcelo A. Ferreira Gomes <mailto:mgomes@trip.com.br> Analista de sistemas e consultor em Informática (Systems analist and CS consultant) Rio de Janeiro, RJ, Brasil------------------------------------------------------------------------ ONElist: where real people with real interests get connected. http://www.onelist.com Join a new list today!
This archive was generated by hypermail 2b29 : Mon Mar 06 2000 - 19:00:53 PST