[Orca-users] RE: Problem with Multiple Disk Run Percent Graphs sh owing and a quest ion

[Blair Zajac]

Hello,

Orca will remember the different types of data and plots that were
created for a particular host and continue to plot them even when
there's no current data for in
them.  This is due to the data in orca.stat and there's probably no easy
way to
disable this.

If you see incorrect data in plots, it may be due to the columns
changing in the
orcallator output and Orca not picking this up.  Orca will not use any
data from
an input file if the number of columns in a particular line don't match
the first
line, but if the meaning of the columns change but the number do not,
then
you'll get data entered into the wrong RRD files.

Regarding ironing out Orca's bug on making multiple plots, I don't have
the
time or resources to fix this and need some help from somebody to take
this
on in addition to handling when orcallator.se files add/change/delete
the
recorded data.

Also, you said "When ORCA restarts it apparently puts a new "header"
line in
the orca/orcallator/<hostname>/percol-<date> file."  Its actually
orcallator that
does this.

Best,
Blair


-------- Original Message --------
 Subject: RE: [orca-users] Problem with Multiple Disk Run Percent Graphs
          sh owing and a quest ion
    Date: Fri, 20 Jul 2001 09:34:44 -0700
    From: "O'Neill, Sean" <sean.oneill at commerceone.com>
Reply-To: orca-users at yahoogroups.com
      To: "'orca-users at yahoogroups.com'" <orca-users at yahoogroups.com>

K ... I didn't read the comments at the top of orcallator.se v1.28b4.
Apparently giving statisitcs for md devices has been removed and that's
why the headers aren't showing up anymore.But why is ORCA, or RRD in
this case, still generating graphs with MD device names ?I brought this
up because of an idea I have of ORCA showing information related to RAID
strip performance.  In the "Capacity Planning for Internet Servers" book
there is a little section on deriving performance statistics for strip
configurations on page 114.  I think this would be something VERY
interesting to have data on and I think ORCA could be tweaked to handle
it.  Of couse, before this can happen the stability of the disk section
within ORCA needs to be ironed out.

     -----Original Message-----
     From: O'Neill, Sean
     Sent: Friday, July 20, 2001 11:10 AM
     To: 'orca-users at yahoogroups.com'
     Subject: [orca-users] Problem with Multiple Disk Run Percent
     Graphs showing and a quest ion
     First, I'm running the most current ORCA/RRD configuration
     available with SE
     v3.2 on Solaris 8 system.

     Long email but please read ... this is REALLY getting
     frustrating.

     I'm asking this question hoping to get rid of these annoying
     multiple and
     usually incorrect Disk Run Percent graphs.  I've noticed the
     following which
     looking into this:

     - I'm using SDS on my system and for a long while the MD
     devices were being
     graphed.  I stopped ORCA for a while doing something and
     restarted it a
     while later.  The MD devices weren't showing up anymore in the
     graphs.  Only
     the cXtXdX devices were.

     - At midnight, the data files rolled over and the multiple Run
     Percent
     graphs showed up.  This time the MD devices DID show up in one
     of the graphs
     but the data is wrong.

     So I started poking around.

     - When ORCA restarts it apparently puts a new "header" line in
     the
     orca/orcallator/<hostname>/percol-<date> file.  By header I
     mean:

     timestamp locltime DNnsrkcmdit   uptime  usr%  sys%  wio%
     idle%  1runq ....
     <and lots more>

     It also does this when it rolls over to the next day at
     midnight and creates
     a new data file.

     - I noticed that in the file for today that was created last
     night the MD
     devices are NOT showing at all in the headers.  Only the
     cXtXdX devices are.

     - I saw that the "headers" are mirrored in the
     orca/rrd/orcallator/<hostname>/orca.state file used by RRD.
     Same problem
     entries.  I noticed that most entires look this this:

     /usr/local/orca/var/orca/orcallator/chivas/percol-2001-06-22.bz2
     993272102
     995577819 8388687 80022 25799 993272102
     USETHISFILE/usr/local/orca/var/orca/orcallator/chivas/percol-2001-06-21.bz2

     which I think means use the "headers" (USETHISFILE) found for
     this entry.

     So I'm pretty sure the problem isn't with RRD and the graphs.
     Its the data
     being fed to RRD.

     The last time the MD devices were showing up in the data
     correctly was in
     the percol-2001-07-18 data file.  I had stopped and restarted
     ORCA on the
     19th for something and that's when this problem came up.  I
     tried "fixing"
     this by deleting the percol-2001-07-19 data file completely.
     ORCA recreated
     the percol-2001-07-19 data file Did this but the problem was
     still.  Only in
     this fashion:

     - The percol-2001-07-19 and percol-2001-07-20 data files have
     NO references
     to MD devices anywhere in the headers yet a graph (one of
     several) is being
     generated with MD name references in it even though the data
     is crap.

     So why isn't ORCA picking up my MD devices anymore ?  I assume
     this issue
     isn't specific to MD devices.  This is probably happening for
     the other
     devices as well.

     I tried deleting the percol-2001-07-20 data file as well as
     the entry for
     2001-07-20 from the orca.state file.  My MD devices still were
     not picked
     up.

     Any ideas out there ??

     ..................................................
     ....... ..- -. .. -..- .-. ..- .-.. . ... ........
     .-- .. -. -.. --- .-- ... -.. .-. --- --- .-.. ...

     Sean O'Neill
     Senior UNIX Systems Engineer / Project Leader
     Global Services
     Commerce One