[pratheep]

Hi,

One of my production database shutdown process hangs every planned shutdown. (Both shutdown immediate and abort options).

Database version & OS : 10.2.0.4 on SuSe Linux 10 SP2, All the latest patches, cpus installed.

Some of the active database server process are preventing the database from shutdown. I tried to kill the process from OS root user also. But the process are not closing. When I check ps command the process are in ‘D’ state (uninterruptable sleep). For the particular pid, strace command also stuck. As I verified in v$session before shutdown immediate there is no active sessions or jobs in database. But above mentioned “D” state pid’s are exist in v$process.

Please suggest me, how to find the history of the process.? like when it is hanged/which session id/which sql statement caused that..?

Thank you all,

Pratheep

[Saurabh Sood]

Hi Pratheep,

Processes with status “D” cannot be killed at all. They are in “D” status because they are sleeping and waiting for a resource like “Kernel lock” etc. which they will never get.

If you are seeing these ospid in v$process they will not allow “shutdown immediate” or “shutdown abort”.

The only way to remove such processes is to “reboot” server.

These processes cannot be traced down.

Check If you can apply Dead Connection detection (DCD) to avoid such situations in future.

Cheers!!!

[Amit Bansal]

If ‘Shutdown Normal’ hangs

Symptom:

‘Shutdown normal’ has been issued to the Oracle database, but Oracle will not shutdown.

Possible Causes and Remedies:

There are two likely causes:

1. ‘Shutdown Normal’ will only shutdown Oracle when there are no remaining connections to Oracle. It is likely therefore that an application still has an active session.

Possible Remedies:

* close all open sessions

* or issue a ‘shutdown immediate’.

2. If using Oracle Parallel Server issuing ‘Shutdown normal’ on more than one node at a time can cause the system to hang. This is a recognised Oracle problem (Oracle 8.0.5).

Possible remedies:

* Only issue ‘Shutdown normal’ on one node at a time, and only issue it on the next node when the previous node has completed its shutting down.

* Alternately, use ‘shutdown immediate’ instead.

[Author]

Posts