db shutdown ?

Still learning the db, so this may seem like an elementary architecture question but here goes:

When I shutdown normal on the db bdwh I still have processes that are running. Why is it that all process are not shutdown when performing a graceful shutdown of the db? I am sure that most know this but the processes are as follows:

oracle 28478 1 0 14:27:42 ? 0:00 ora_smon_bdwh
oracle 28488 1 0 14:27:42 ? 0:00 ora_snp3_bdwh
dwload 1170 13509 0 14:36:11 pts/9 0:00 grep bdwh
oracle 28480 1 0 14:27:42 ? 0:00 ora_reco_bdwh
oracle 28476 1 0 14:27:41 ? 0:00 ora_ckpt_bdwh
oracle 28474 1 0 14:27:41 ? 0:00 ora_lgwr_bdwh
oracle 28486 1 0 14:27:42 ? 0:00 ora_snp2_bdwh
oracle 28472 1 0 14:27:41 ? 0:00 ora_dbw0_bdwh
oracle 28484 1 0 14:27:42 ? 0:00 ora_snp1_bdwh
oracle 28482 1 0 14:27:42 ? 0:00 ora_snp0_bdwh
oracle 28470 1 0 14:27:41 ? 0:00 ora_pmon_bdwh

If those processes are running, that means your database was not shutdown. If you shutdown normal, Oracle will wait for all existing connections to be disconnected. If you have persistent connections, the database never gets shut down. If you shutdown immediate, Oracle will gracefully terminate existing connections.

Are there any risks to outright deleting these processes once the db is shutdown?

Quote:

---

If those processes are running, that means your database was not shutdown.

---

Those processes will not be running if your database is down. You should not kill these procesess from the OS.

Actually the database is shutdown. When I try and restart I get the following output:

Total System Global
Area 560785840 bytes
Fixed Size 51632 bytes
Variable Size 67100672 bytes
Database Buffers 491520000 bytes
Redo Buffers 2113536 bytes
ORA-01102: cannot mount database in EXCLUSIVE mode
SVRMGR> Disconnected.
SVRMGR> Server Manager complete.

------------
A DOC from metalink states that I must:
(1) shutdown the db
(2) kill all processes active for that SID
(3) Delete lk[SID] file in $ORACLE_HOME/dbs
(4) restart DB

What is the significants of the 'lk' file? Will this process endanger the database?

What is the database version and the platforum? First of all you should never delete the running processes, without the proper execution of shutdown. Is the database in archive log mode?

Sam

db version 8.0.6 on Solaris 7. Yes in archive mode but not a lot of activity on the db - mostly houses data. Maybe my limited experiences in confusing the question.

I shutdown the db, for this purpose, with a shutdown abort, startup restrict, shutdown normal. In some cases the db processes above still exist. What are the issues with killing the processes manually once the db is shutdown.

When you connect does it tell you that you are connected to an idle instance, if so you are shutdown.

Try 'select status from v$instance' to check the status.

when you get a message telling you 'cannot mount exclusive', it usually means that your database and instance are already up.

Shutdown normal will wait for all user session to disconnect before shutting down, is anyone else logged in ?

try 'shutdown immediate', this will shut down dataabase immediately, rolling back any uncommited transactions and kill user sessions.

If you kill any of he mandatory background processes while the database is open it will result in instance failure, the system will automatically recover itself on startup, this is not recommended.

Ok, I am sure that I am confused so let me re-phrase my situation and question:

(1) Shutdown db(bdwh) from server manager:

shutdown abort
startup restrict
shutdown immediate **

** did this using normal and immediate with the same results
Result from startup:

Total System Global
Area 560785840 bytes
Fixed Size 51632 bytes
Variable Size 67100672 bytes
Database Buffers 491520000 bytes
Redo Buffers 2113536 bytes
ORA-01102: cannot mount database in EXCLUSIVE mode
SVRMGR> Disconnected.
SVRMGR> Server Manager complete.

(2) From Solaris7 command line I grep for db process:

ps -ef |grep bdwh

Result from command:

oracle 28478 1 0 14:27:42 ? 0:00 ora_smon_bdwh
oracle 28488 1 0 14:27:42 ? 0:00 ora_snp3_bdwh
oracle 1170 13509 0 14:36:11 pts/9 0:00 grep bdwh
oracle 28480 1 0 14:27:42 ? 0:00 ora_reco_bdwh
oracle 28476 1 0 14:27:41 ? 0:00 ora_ckpt_bdwh
oracle 28474 1 0 14:27:41 ? 0:00 ora_lgwr_bdwh
oracle 28486 1 0 14:27:42 ? 0:00 ora_snp2_bdwh
oracle 28472 1 0 14:27:41 ? 0:00 ora_dbw0_bdwh
oracle 28484 1 0 14:27:42 ? 0:00 ora_snp1_bdwh
oracle 28482 1 0 14:27:42 ? 0:00 ora_snp0_bdwh
oracle 28470 1 0 14:27:41 ? 0:00 ora_pmon_bdwh

Now the database is shutdown.


The db is down and the processes above are still active. For the most part, I only have this problem when I try and run these server manager commands from a k-shell script. Why would it cause this problem? I have set up the environment and connect to internal fine. The simple test script follows:

#!/usr/bin/ksh

. ${HOME}/.profile
ORACLE_SID=${1}
export ORACLE_SID
ORACLE_HOME=/u01/app/oracle/product/8.0.6/
export ORACLE_HOME

# Shutdown and startup database
$ORACLE_HOME/bin/svrmgrl << EOF
connect internal
shutdown abort
startup restrict
shutdown immediate
startup
disconnect
exit
EOF

Any ideas?

Ah ha, now we get to the root problem.

Do these processes exist when you execute these commands from the command line instead of inside a script? I would bet not.

Can you do the following on the terminal window and confirm the results:

Code:

---

svrmgrl
SVRMGR> internal
SVRMGR> shutdown immediate
SVRMGR> host;
$ > ps -ef|grep SID
$ > exit
SVRMGR> startup mount
SVRMGR> host;
$ > ps -ef|grep SID
$ > exit
SVRMGR> alter database open;
SVRMGR > exit

---

Let us know what happen, with the output including any error messages....

Sam

What is not understood is that the INSTANCE is started, but not MOUNTED or OPENED. After the series of events listed the database has NOT been shutdown, so the processes will remain. Do NOT kill them from the operating system; connect via SVRMGRL as INTERNAL and issue a SHUTDOWN ABORT. You SHOULD see:

Oracle instance shut down.

NOW check your process list and you should not see any Oracle processes.

To startup the instance do it in stages at this point (this is NOT necessary for every database startup, but it may prove helpful in this situation). Do the following:

SVRMGR> connect internal
Connected.
SVRMGR> startup nomount
.....
SVRMGR> alter database mount;
Statement processed. <--- the expected result
SVRMGR> alter database open:
Statement processed.

This should help you in figuring out why the database cannot be mounted and opened.

No these processes do not exist when run from command line outside of the script.

Why would the db not shutdown from the script?

In you script

Code:

---

#!/usr/bin/ksh -xv

. ${HOME}/.profile
ORACLE_SID=${1}
export ORACLE_SID
ORACLE_HOME=/u01/app/oracle/product/8.0.6/
export ORACLE_HOME
PATH=$ORACLE_HOME/bin:$PATH;
export PATH
LD_LIBRARY_PATH=$ORACLE_HOME/lib;
export LD_LIBRARY_PATH


# Shutdown and startup database
$ORACLE_HOME/bin/svrmgr << EOF
connect internal
shutdown immediate
EOF
exit

---

Check to see whether any processes are still running. give it some time and check for running processes. If you don't see any, then add the statement "startup" before the "EOF" and see whether it brings back the proceess as normal...

Sam

[Edited by sambavan on 11-07-2001 at 05:28 PM]

I apparently misread your posting of the results of your startup restrict. It appeared that you had shutdown the instance, attempted to perform a startup restrict and left it at that without issuing another shutdown command.

I added PATH, and LD_LIBRARY_PATH. Still get 'ORA-01102: cannot mount database in EXCLUSIVE mode' and processes still running. Any other ideas?

I actually did shutdown and restart after the startup restrict but did not include the output because it was redundant.

Not meaning to drag this issue but, anymore suggestions. Any idea's as to why in am in 'EXCLUSIVE mode' when accessing server manager from script?

What seems to be happening in you case could be this. When you shutdown the database, the processes seems to hold some file system locks which are not getting released. As a result, the processes are awaiting for them to get released. I would now like to see you start and stop the instance explicitly mentioning the pfile= your_path/initSID.ora and see whether this would solve the problem for you....


Sam

I included pfile with the startup command in the script and unfortunately I received the same results: the following error command 'ORA-01102: cannot mount database in EXCLUSIVE mode'.

does your script allow you to just stop and startup the database without any special modes? Also could you try the regular borne shell instead of the korne shell?

Sam

Why not try out all this commands manually first, instead of in script ?
My guess of your problems here:
1. when you Shutdown abort, you may not able to startup in restrict mode. So the database startup, but in normal.
2. because the startup restrice command get an error in getting restric privilege, the connection is dropped.
3. so your third shutdown command failed because you are disconnected.
4. your shutdown command failed, so... all the processes are alive there. The database actually is either startup normal or mounted, or even nomount. Anyway it's running!

Hope you could check and let us know the result.

Really should avoid using shutdown abort!

In addition to avoiding the shutdown abort, try issuing an alter system checkpoint prior to issuing the shutdown IMMEDIATE. Also might consider issuing a switch of the logfile prior to shutdown as well.

Quote:

---

Originally posted by philmer
Really should avoid using shutdown abort!

---

Yes, but often you simply have to use it...