[SunHELP] RE: Solaris 9 and SDS/SVM

[sunhelp at sunhelp.org]

I'll second the motion for Veritas being out of the question on root
volumes.
The information I'm giving you is courtesy of Reggie Beavers, Simon Burr,
and
Eugene Schmidt's post to me on similar questions.  His solution was to have
a script that monitors metastat output for problems and upon detection to
notify you so that you can address the problem before a reboot or you can
delete metadb's for the failing disk (metadb -d cXtYdZsN) and then be able
to continue the boot process.  I would this proactive measure would be a
good
starting point for remote disks as well.

Here is Reggie's script.  I haven't actually tested it at this point.

#!/bin/sh
#
# metachk.sh
#
# look for non-Okay states
#
# rbeavers 8-28-02

# if recently booted then exit
if [ `uptime|grep -c min` -ne 0 ]; then
        exit 0
fi

BASENAME=`basename $0`
MAILTO=reggiebeavers at fstha.com

LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/lib
export LD_LIBRARY_PATH


/usr/ucb/echo -n "Not Okay: " > /tmp/$BASENAME.msg

# check non-ctd devices
#
/usr/sbin/metastat -p|grep -v "c.*t.*d"|while read md
rest; do
   if [ `/usr/sbin/metastat $md|grep State:|grep -vc
Okay` -ne 0 ]; then
      if [ ! -f /tmp/$BASENAME.$md.notified ]; then
         touch /tmp/$BASENAME.$md.notified
         /usr/ucb/echo -n " $md" >> /tmp/$BASENAME.msg
      fi
   else
      rm -f /tmp/$BASENAME.$md.notified
   fi
done

# check ctd devices
#
/usr/sbin/metastat -p|grep "c.*t.*d"|while read md
rest; do
  /usr/sbin/metastat $md|grep "c.*t.*d"|while read ctd
sb db state rest; do
   if [ ! "$state" = Okay ]; then
      if [ ! -f /tmp/$BASENAME.$ctd.notified ]; then
         touch  /tmp/$BASENAME.$ctd.notified
         /usr/ucb/echo -n " $ctd" >>
/tmp/$BASENAME.msg
          
      fi
   else
      rm -f /tmp/$BASENAME.$ctd.notified
   fi
  done
done

MSG="`cat /tmp/$BASENAME.msg`"
if [ ! "$MSG" = "Not Okay: " ]; then
   # notify
   #
   /usr/sbin/metastat|/usr/ucb/mail -s "$BASENAME:
$MSG" $MAILTO
   logger -p user.error -t METASTAT $BASENAME: $MSG
fi


HTH
also a good link to look at http://docs-pdf.sun.com/816-4519/816-4519.pdf


-----Original Message-----
From: Rick von Richter [mailto:rickv at mwh.com]
Sent: Monday, August 18, 2003 5:40 PM
To: Sun Managers Mailing List
Subject: Solaris 9 and SDS/SVM


OK I have read the archives but no answer for my question.
Our standard install to this point has been a mirror of the OS disks 
using SDS.  The key point being that if you lost a disk you would still 
have 50% of your metadbs and it would still boot off the remaining disk.

On Solaris 9, SDS chokes.  So we tried using the newer SVM but found 
that its metadb requirement is 50% + 1 viable metadbs.  So upon reboot 
with one bad disk the system will boot into single user mode.  This is 
not too much of a problem with local systems, HOWEVER, a lot of our 
systems are remote X1s or V100s with two disks.  So if this happens we 
are dead in the water.
Is there a way around this?  Or, maybe is there some other (small) 
software that can mirror the OS disks?  BTW, Veritas is out of the question.

TIAWS
-- 

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 Rick von Richter  IS Production Support Manager    Voice: 858-831-2222
 rickv at mwh.com     Maintenance Warehouse/Home Depot   Fax: 858-831-2221
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
   The box says: Win98, WinNT or BETTER. That's why I installed Linux.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
_______________________________________________
sunmanagers mailing list
sunmanagers at sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers