Re: Script to overcome ORA-54 during DDL

Notes in-line.

Regards

Jonathan Lewis

http://www.jlcomp.demon.co.uk

http://www.jlcomp.demon.co.uk/faq/ind_faq.html The Co-operative Oracle Users' FAQ

http://www.jlcomp.demon.co.uk/seminar.html Optimising Oracle Seminar - schedule updated May 1st

• Original Message ----- From: "Powell, Mark D" <mark.powell_at_eds.com> To: <oracle-l_at_freelists.org> Sent: Friday, June 25, 2004 8:33 PM Subject: RE: Script to overcome ORA-54 during DDL

Mark, UL locks once granted are held until released via dbms_lock.release or the session terminates and therefore would cover multiple commits actions if that is what the holding process does.

>> Release_on_commit is an option which, bizarrely
>> defaults to the wrong value (i.e. false) for UL
>> locks. Given Mark's comment about DDL
committing, though, the code would have to choose to make the lock 'hold on commit' and release it explicitly after the DDL was over.

UL locks are also a relatively expensive resource according to the Oracle documentation and requiring every session to take a share level UL lock on a table and then release it for OLTP type work would be a lot of overhead relative to the work being done.

>> I think the 'relatively expensive' comes from Oracle's
>> vague warning that is supposed to stop people from
>> taking out UL enqueues at the rate they could take
>> out row-locks. A UL enqueue is no more expensive
>> than a TM enqueue because it's exactly the same
>> type of thing.- so all you would be doing is taking
>> two "TM locks" to update a table.

IMHO -- Mark D Powell --

-----Original Message-----
From: oracle-l-bounce_at_freelists.org
[mailto:oracle-l-bounce_at_freelists.org]On Behalf Of Bobak, Mark Sent: Friday, June 25, 2004 3:21 PM
To: oracle-l_at_freelists.org
Subject: RE: Script to overcome ORA-54 during DDL

<Caveat: untested concept>

Seems to me, if you have control over the code base, and you know these = types of
things will be necessary, you could implement something via DBMS_LOCK. Any time a DML executes, it could first take a user defined (UL) enqueue in shared mode. It does it's DML, commit/rollback, and the UL is = released
atomically w/ the DML. Now, you come along and need to do mainenance. So, you exec dbms_lock on the same UL enqueue, but in exclusive mode, and if any DML is executing, you'll queue up and wait on that enqueue. When all pending DML commits/rolls back, you'll get your exclusive UL lock, and you can do the DDL. While this is going on, all DML will queue up, till the exclusive lock clears. When you're done w/ the DDL, release the exclusive lock, all DML continues where it left off.

Hmm...just thought of something. DDL implies commit before and after the change that it's making. Would that first commit release the lock on the UL enqueue? Hmmm, maybe that's not such a good idea.

I'll have to test it out.....

-Mark

-----Original Message-----
From: oracle-l-bounce_at_freelists.org
[mailto:oracle-l-bounce_at_freelists.org]On Behalf Of Jeremiah Wilton Sent: Friday, June 25, 2004 2:56 PM
To: oracle-l_at_freelists.org
Subject: Script to overcome ORA-54 during DDL

Sometimes when trying to perform DDL on really hot objects (heavy read/DML), I get:

ORA-00054: resource busy and acquire with NOWAIT specified.

I guess this is because you need to obtain a library cache lock in order to change the object definition. Since there is no enqueue mechanism for this resource, you can't just 'lock table ... in exclusive mode;' All that gives you is a DML lock.

One way to avoid this is to write a PL/SQL routine that uses DBMS_SQL and spins trying to run the DDL, stopping only when it succeeds. This seems to work most of the time.

Does anyone have a script for doing the above that they would like to share?

Please don't tell me to just use dbms_redefinition. That is unnecessarily complex when the above technique can be used instead.

Thanks

--
Jeremiah Wilton
http://www.speakeasy.net/~jwilton
----------------------------------------------------------------
Please see the official ORACLE-L FAQ: http://www.orafaq.com
----------------------------------------------------------------
To unsubscribe send email to:  oracle-l-request_at_freelists.org
put 'unsubscribe' in the subject line.
--
Archives are at http://www.freelists.org/archives/oracle-l/
FAQ is at http://www.freelists.org/help/fom-serve/cache/1.html
-----------------------------------------------------------------
----------------------------------------------------------------
Please see the official ORACLE-L FAQ: http://www.orafaq.com
----------------------------------------------------------------
To unsubscribe send email to:  oracle-l-request_at_freelists.org
put 'unsubscribe' in the subject line.
--
Archives are at http://www.freelists.org/archives/oracle-l/
FAQ is at http://www.freelists.org/help/fom-serve/cache/1.html
-----------------------------------------------------------------
----------------------------------------------------------------
Please see the official ORACLE-L FAQ: http://www.orafaq.com
----------------------------------------------------------------
To unsubscribe send email to:  oracle-l-request_at_freelists.org
put 'unsubscribe' in the subject line.
--
Archives are at http://www.freelists.org/archives/oracle-l/
FAQ is at http://www.freelists.org/help/fom-serve/cache/1.html
-----------------------------------------------------------------


----------------------------------------------------------------
Please see the official ORACLE-L FAQ: http://www.orafaq.com
----------------------------------------------------------------
To unsubscribe send email to:  oracle-l-request_at_freelists.org
put 'unsubscribe' in the subject line.
--
Archives are at http://www.freelists.org/archives/oracle-l/
FAQ is at http://www.freelists.org/help/fom-serve/cache/1.html
-----------------------------------------------------------------