Patch Install/ Reboot

[Nobody's Hero]

I installed a patch depot of 100 patches this weekend. An autoreboot=true was used, however I didn't shutdown my Oracle databases prior to installing the patch. I just assumed that everything would shutdown OK. DBA's claim that it didn't go down clean. I know I should have dropped the init level prior to installing the patches. Is there anywhere I can look to see how everything shutdown after the patch finished loading?

[Pete Randall · Accepted solution]

Patch reboots are of the quick and dirty variety. They do not do a normal shutdown but a "reboot -rq". Installation directions tell you that your system should be in a quiesced state, i.e. no databases, no users.


Pete

[Cheryl Griffin]

swinstall is known to run reboot instead of shutdown which would have used the kill scripts.

If you examine /etc/rc.log.old you should not see entries for the shutdown process. During a "normal" shutdown this is the log file that would be updated as each kill script completes.

[Cheryl Griffin]

There is an SR for this
8606223534 "Request for SD-UX to allow customized messaging to users" which says:
"The default messages indicate that the system is about to be rebooted when kernel related filesets are installed or removed, but these messages do not indicate that the system will not be performing a shutdown(1M), but rather a reboot(1M), which does not execute the standard /sbin/init.d scripts to gracefully shutdown applications."

Some users have changed /var/adm/sw/defaults to reference shutdown vs. reboot but this is not recommended as it can cause other problems.

It's simply best to remember to shutdown databases prior to installing patches, when the patches cause a reboot.

[S.K. Chan]

What you should have done is either make sure the oracle shutdown script gets run (manually) before you install these patches, or disable the oracle startup and shutdown -r the system and install these patches afterwards. When the system comes up eventually, you can always start up oracle later. The /etc/rc.log file (or any other older file) may give you the indication if the script was ever run.

[Tim Sanko]

I just hope that your DBAs didn't have to do a recovery.

Here is my patch cheklist.

1. MAKE AN IGNITE TAPE
make_recovery -AV

some might prefer make_tape_recovery.

2. check the tape is readable

dd if=/dev/rmt/0m of=/dev/null -bs 512 (see if tape is intact.)

3. shutdown Oracle database

4. swreg -l depot .


5. swinstall -f
/mountPoint/depotname

(This step may require a reboot).

6. swreg -l depot -u


7. rm /mountPoint/depotname

8. restart Oracle database.

Sometimes a procedure followed like a habit is a very good thing...

Tim

[Steven E. Protter]

make_tape_recovery

#!/usr/bin/sh
/opt/ignite/bin/make_tape_recovery -Av -x inc_entire=vg00


That will do the boot disk.

Ignite isn't really good at doing open databases anyway. If you want to backup oracle and what not, chagne it to add a -x inc_entire=vg01 Make sure the database is down when you do that.

SEP