Problem installing SMC - database issues

Question

Hi Everyone,
I'm installing SMC 361 on Solaris 10.
The install itself seems fine.
However, when i run the guisetup, it fails :
Configure Sun Management Center web server port. Default port is 8080
Press RETURN to force default port.
Enter port you would like to use [ 1100 to 65535 ]: : 8080
Failed to start database.
Failed to start database.
Completing Sun Management Center Server Component setup.
When I look in svcs :
svcs -x sunmcdatabase
svc:/application/management/sunmcdatabase:default (SunMC database service)
State: maintenance since Mon Sep 24 14:03:16 2007
Reason: Start method failed repeatedly, last exited with status 113.
See: http://sun.com/msg/SMF-8000-KS
See: /var/svc/log/application-management-sunmcdatabase:default.log
Impact: This service is not running.
Here's the log /var/svc/log/application-management-sunmcdatabase:default.log :
cat /var/svc/log/application-management-sunmcdatabase:default.log
[ Sep 24 12:42:56 Disabled. ]
[ Sep 24 12:42:56 Rereading configuration. ]
[ Sep 24 12:54:51 Rereading configuration. ]
[ Sep 24 13:13:45 Enabled. ]
[ Sep 24 13:13:45 Executing start method ("/lib/svc/method/es-svc.sh start database") ]
execution of svrmgrl using /tmp/SunMCDBLogFiles/db-start.sql fail
exiting........................
sqlerror code = 01157
sqlerrmsg full = ORA-01157: cannot identify/lock data file 3 - see DBWR trace file
sqlerror code = 01110
sqlerrmsg full = ORA-01110: data file 3: '/var/opt/SUNWsymon/db/oradata/SunMC/temp01.dbf'
[ Sep 24 13:13:53 Method "start" exited with status 95 ]
[ Sep 24 13:49:57 Leaving maintenance because clear requested. ]
[ Sep 24 13:49:57 Enabled. ]
[ Sep 24 13:49:57 Executing start method ("/lib/svc/method/es-svc.sh start database") ]
And yet, if i do a ps -fe :
smcorau 3127 1 0 14:34:40 ? 0:00 /opt/smc/SUNWsymon/oracle/product/8.1.7/bin/tnslsnr smcdblistener -inherit
smcorau 3144 1 0 14:34:40 ? 0:00 ora_dbw0_SunMC
smcorau 3142 1 0 14:34:40 ? 0:00 ora_pmon_SunMC
smcorau 3152 1 0 14:34:40 ? 0:00 ora_reco_SunMC
smcorau 3154 1 0 14:34:40 ? 0:00 ora_s000_SunMC
smcorau 3148 1 0 14:34:40 ? 0:01 ora_ckpt_SunMC
smcorau 3150 1 0 14:34:40 ? 0:00 ora_smon_SunMC
smcorau 3146 1 0 14:34:40 ? 0:00 ora_lgwr_SunMC
So, actually the database is up!
I'm quite puzzled by this. I'm an oracle dba myself, but the SMC has
installed this 817 database in quite a specific way. Debugging why it's
not starting could take quite some time. But what really concerns me is
why this has happened at all on a standard installation on Solaris 10.
Has anyone else seen this particular situation? Any help would be much
appreciated at this point.
regards
Paul Stuart

Answer 1

I am pretty certain it is the existence of the smcorau user and or the smcorag group at the time of the subsequent install. As best as I can tell, subsequent installs end up bumping the userid of smcorau's gid and smcorag and smcorau get out of sync with the filesystem, and then oracle won't start.
I have successfully performed re-installs of SMC without removing everything by removing the smcorau user and the smcorag group. But I have taken to aggressively removing everything if I need to re-install.
NB The install script for the perf mon (i think) database requests a small, medium, large choice, and then asks for a dir on a filesystem with "at lease X MBytes free." That fs must be ufs. Not, that fs can't be nfs. So, you can't use zfs. There are perhaps 2 ways around this:
I know this works: find a spare disk; newfs it; mount it at (say) /var/smcoradb and after smc installs, shut it down and dump/restore the dirs to a zfs filesystem, and then zfs set mountpoint=/var/smcoradb pool/filesystem;
I think this works, but only if installing all of SMC at once: mkdir /var/smcoradb, and then zfs create pool/symondb; zfs set mountpoint=/var/smcoradb/SUNWsymon pool/symondb
Cheers!