[Orca-users] Re: Disk run % graphs don't seem right...

Fri Feb 25 17:49:09 PST 2000

Once an RRD file is updated with data from a particular time, it cannot
have older data appended to it.  The solution in this case is to delete
both your orcallator.state file and all the RRD files.  I know this isn't
fun, but its the only way.

I've had this problem myself and Orca would use several hundred
megabytes of memory to reload all the percol- or orcallator- files.
That's why the next release of Orca has a new configuration file
parameter that allows users to specify the sorting order for loading
text files and to specify when data loaded from them is flushed.

As for the mailing list functions, I haven't tried, but onelist.com
does not have good support for archiving email.

I'm looking at moving Orca to sourceforge.net but the package
will need to be placed under an open-source license.  Does
anybody have any recommendations on which one?  The
QPL license looks good.

Blair

Greg Winn wrote:

> From: "Greg Winn" <winngre at cscoe.ac.com>
>
> Very true, thought about that after hitting send... I suppose one could
> start coding a modification to include the legend format in the orcallator
> config file and make it graph specific, but after thinking about it that
> doesn't sound useful.  When RRDtool uses the %lf format it has already
> converted the value to a number that will be appended with a 'u' or 'm', so
> I can't just specify the field on the decimals.
>
> Another question, and it may be my lack of experience with RRD again. If I
> have percol files for a time period prior to the last update, but presently
> the RRD is not filled in with data for that time, can I add it back in? For
> example, orca was stopped for a period of time and when it was started again
> it didn't have the path to gunzip, and so it didn't process the previous
> days percol file (it had been compressed and was therefore unreadable), it
> read the current day's percol file and started feeding data in again.  After
> correcting the path, and uncompressing the file, I have tried manipulating
> the orca.state file to try and get the data in but I'm not making any
> progress.
>
> Is it me or does the search function of the mailing list(s) not return
> anything but mailing list categories/groups and descriptions?
>
> Thanks,
> -Greg
>
> -----Original Message-----
> From: Blair Zajac [mailto:blair at akamai.com]
> Sent: Friday, February 25, 2000 4:04 PM
> To: orca-help at onelist.com
> Subject: Re: [orca-help] Disk run % graphs don't seem right...
>
> From: Blair Zajac <blair at akamai.com>
>
> If you change the %lf format than all the plots will be effected, not
> just the disk ones.
>
> Blair
>
> Greg Winn wrote:
>
> > From: "Greg Winn" <winngre at cscoe.ac.com>
> >
> > The scaling makes perfect sense to me, thank you -- although it would be
> > nice to set the number of significant digits I am interested in so that
> the
> > numbers do not look so long -- I am more worried about genuinely large
> > numbers than genuinely small ones!  I guess I can do that by playing with
> > the GPRINT statement (%lf) to specify field width etc?
> >
> > Thanks,
> > -Greg
> >
> > -----Original Message-----
> > From: Blair Zajac [mailto:blair at akamai.com]
> > Sent: Friday, February 25, 2000 3:06 PM
> > To: orca-help at onelist.com
> > Cc: winngre at cscoe.ac.com
> > Subject: Re: [orca-help] Disk run % graphs don't seem right...
> >
> > From: Blair Zajac <blair at akamai.com>
> >
> > The numbers with a 'u' or 'm' after them mean to scale the
> > number by 1e-6 and 1e-3 respectively, so the numbers are
> > actually very small.  They are still a percentage, so 1m is
> > 0.001*100% =1e-5.
> >
> > Is this consistent with the data you would expect?
> >
> > Blair
> >
> > Greg Winn wrote:
> >
> > > From: "Greg Winn" <winngre at cscoe.ac.com>
> > >
> > > Thanks for the reply Paul, the patch worked fine - I can now see all of
> my
> > > metadevices and disks.  The other problem still remains: The legend
> values
> > > (current, average, min, max) on some of my systems show percent values
> (in
> > > that they are numbers less than 100), and on others they seem
> > misrepresented
> > > (perhaps recast? sometimes they have a 'u' or 'm' after the number) and
> > are
> > > set wildly higher than a percentage.  As the graphed data is correct,
> and
> > > the data in the RRD is correct, it seems that RRDtool is producing bad
> > > numbers for the legend on certain graphs -or- part of the
> > > _update_graph_options code is misbehaving.  This is my first
> > implementation
> > > of anything that uses RRDtool, so it could be something I don't
> understand
> > > yet.
> > >
> > > I have obtained and compiled the most recent RRDtool (1.0.13) and
> > confirmed
> > > that the new version's RRDs package and libraries are being used.  I
> built
> > > RRDtool using GCC 2.951, but I am considering using a different compiler
> > to
> > > see if that makes a difference.  Has anyone else seen this kind of
> > behavior?
> > >
> > > Thanks again,
> > > -Greg
> > >
> > > -----Original Message-----
> > > From: Paul Haldane [mailto:Paul.Haldane at newcastle.ac.uk]
> > > Sent: Wednesday, February 23, 2000 4:06 AM
> > > To: orca-help at onelist.com
> > > Cc: Greg Winn
> > > Subject: Re: [orca-help] Disk run % graphs don't seem right...
> > >
> > > From: Paul Haldane <Paul.Haldane at newcastle.ac.uk>
> > >
> > > > Orca-0.25 (out of the box configuration file), Orcallator-1.20.se,
> > > > rrdtool-1.0.10:
> > > ...
> > > > c0t0d0's!).  The rrds are all named for the appropriate disk targets
> for
> > > the
> > > > host, so I'm not sure where this is coming from.  The platform is a
> Sun
> > > > E450, all internal drives (12 + cdrom), Solaris 2.6, using Disksuite
> > 4.2.
> > >
> > > I thought we'd fixed the problem with DiskSuite in the latest
> > > version of orca.  Maybe it hasn't arrived in the released version yet -
> > > see my message at
> > >
> > > http://www.onelist.com/omessages/orca-developers?archive=16
> > >
> > > for a patch.
> > >
> > > Paul
> > > --
> > > Paul Haldane
> > > Computing Service
> > > University of Newcastle
> > >
> > > ------------------------------------------------------------------------
> > > FREE ADVICE FROM REAL PEOPLE!  Xpertsite has thousands of experts who
> > > are willing to answer your questions for FREE.  Go to Xpertsite today
> and
> > > put your mind to rest.
> > > http://click.egroups.com/1/1404/5/_/_/_/951300374/
> > > ------------------------------------------------------------------------
> > >
> > > ------------------------------------------------------------------------
> > > FREE ADVICE FROM REAL PEOPLE!  Xpertsite has thousands of experts who
> > > are willing to answer your questions for FREE.  Go to Xpertsite today
> and
> > > put your mind to rest.
> > > http://click.egroups.com/1/1404/5/_/576712/_/951512623/
> > > ------------------------------------------------------------------------
> >
> > ------------------------------------------------------------------------
> > Learn more with SmartPlanet. It's a new way of learning online.
> > SmartPlanet offers hundreds of courses to take on your time,
> > in your space. Join for FREE today!
> > http://click.egroups.com/1/1704/5/_/576712/_/951513164/
> > ------------------------------------------------------------------------
>
> ------------------------------------------------------------------------
> Tired of missing calls while online? Now you can surf the Internet
> without worrying about missing important calls! CallWave's FREE
> Internet Answering Machine lets you hear who's calling while online.
> http://click.egroups.com/1/1927/5/_/576712/_/951522584/
> ------------------------------------------------------------------------