[Orca-users] Re: Getting rid of old data and extra disk graphs

[Charles Dennett]

Blair Zajac wrote:
> 
> Charles,
> 
> Charles Dennett wrote:
> >
> > Greetings,
> >
> > Oh, and just to
> > add to the confusion, we've recently started using DiskSuite
> > on some systems so some of the disk run graphs have the disks
> > noted as c0t0d0, and some graphs have that style plus the DiskSutite
> > style of md1 on the same graph.  (I use c0t0d0 and md1 as examples
> > only.  The graphs have all the actual devices on them.)
> >
> > We recently stood up a couple of new machines and DiskSuite was
> > up and running on them before I installed orcallator.  The disk
> > run graphs for those show only the disks in the c0t0d0 style.
> > The DiskSuite style of md1 is not shown.  I assume the machines
> > that have both styles may have been running at the time an older
> > version of orcallator.se which had problems distinguishing
> > DiskSuite from non-Disksuite disks.  Or, perhaps I had orcallator
> > running on the older systems as I was installing and configuring
> > DiskSuite and orcallator and orca got confused.
> 
> Check the output orcallator-XXXX-YY-ZZ files for the md1 data.  If its
> not there, then its an orcallator.se issue, otherwise Orca is not
> picking
> it up.  You may be running an old orcallator.se on that system.

This only happened on one system I'm monitoring.  I dug into the
orcallator files and found the mention of the md* disks from
four days this past Auguest.  That's when I was in the middle
of getting DiskSuite up and running.  It must have gotten confused
when I was using DiskSuite.  I deleted the data files and the 
rrd files for this machine and am running orca to rebuild this
machines files as I type.

Here's another related problem I just found this morning.  We have
a new system we are in the process of standing up.  It had
Solaris 8 installed and then DiskSuite.  After that I installed
orca.  The disk run graphs for this system looked fine.  The
disks were identified in the cxtydz and there was just one
disk run graph for this system.  This morning a second disk run
graph showed up for this syetem.  It had the same disks as the
first graph plus 2 more - sd4 and sd11.  I did some checking
with the person who is working on this system to see what happened.
At the time this graph showed up, they had just added 2 new disks
to the diskpack and run the script to allow the system to see them
(no boot -r required!)  What I did was to stop orcallator, remove
the one orcallator data file that had sd4 and sd11 in it (orcallator
was smart enough to start a new data file when it saw the disk layout
had changed) and restart orcallator.  It created a new data file and
this time the two new disks were properly identified with the
cxtydz notation.  

So, it appears that if you add disks to a system on the fly, 
orcallator can get confused and call the disks with their
sd names rather than their cxtydz names.

--
Charlie Dennett
Eastman Kodak Company, Rochester, NY
Sitemaster http://www.kodak.com