RE: locking issues

Kevin,

I know you -- I know you are a developer not a DUHveloper and I know Joe -- he's a very smart DBA

Okay, the next step is to reproduce the problem, but do it via two sqlplus sessions, NOT via the system, one doing the insert, one doing the update.. I would then run through the following tests:

1. start the insert first, then the update, different rows but the same partition
2. start the update first, then the insert, different rows but the same partition
3. repeat 1 above, but now make sure the rows are in different partitions
4. repeat 2 above, again, make sure they are in different partitions

I'd also drop and recreate the index on the FK...

this won't solve it, but if you can get the problem down to a very very specific set of circumstances, you may be able to pinpoint what's happening... and I would call support anyway, even if it is a deadlock, that's what they are there for.

If you cannot reproduce it via sqlplus and those two SQL statements, there is another statement somewhere in that screen that is causing the deadlock....

>From: "Toepke, Kevin M" <ktoepke_at_cms.cendant.com>
>Reply-To: ORACLE-L_at_fatcity.com
>To: Multiple recipients of list ORACLE-L <ORACLE-L_at_fatcity.com>
>Subject: RE: locking issues
>Date: Tue, 08 Aug 2000 04:54:24 -0800
>
>All:
>
>I am the developer working with Joe on resolving the Duhveloper's problems
>(I didn't write the code, I'm just charged with fixing it. Ugh!)
>
>First of all, Steve Adams suggested we look at the PCTFREE on the tables
>involved. All have a PCTFREE of 10.
>
>First, Rachel's questions.
> >
> > you mean the fact that the trace file specifically states
> > that it is an
> > application problem doesn't hold any weight with the developers?
>
>Not with the Duhveloper that wrote the code. He still maintains it can't be
>his problem?!?!? He's turned the problem over to us stating to his manager
>that it is a DBA issue.
>
> >
> > do the users log in as separate accounts or does everyone log
> > in as the same
> > userid?
>
>Separate accounts.
>
> >
> > can you reproduce it with only one or two users on the
> > system, each of whom
> > is tracking what they are doing?
>
>It has been reproduced with exactly 2 users on the system each accessing
>the
>screen for a different Invoice.
>
> >
> > you are deadlocking on an insert statement???? are there
> > triggers on the
> > table that reference another table that could be locked?
>
>Deadlocking is occurring on and INSERT statement and an UPDATE statement
>(both on the same table). There are no triggers on the table. There is a
>single-column FK on the table. And it is properly indexed. The table that
>the deadlocking is occurring on is range partitioned by its Primary Key
>column.
>
>Kevin
>
> > >From: "Joseph S. Testa" <teci_at_oracle-dba.com>
> > >Reply-To: ORACLE-L_at_fatcity.com
> > >To: Multiple recipients of list ORACLE-L <ORACLE-L_at_fatcity.com>
> > >Subject: locking issues
> > >Date: Mon, 07 Aug 2000 10:54:45 -0800
> > >
> > >I'm hitting up against locking(including deadlock) issues.
> > Here is what
> > >we've done so far:
> > >
> > >using utllockt, dba_waiters, etc all of the normal delivered
> > stuff has
> > >been used.
> > >
> > >I'm sure there is a way to see what the values of bind
> > variables are, i
> > >think its an event, but i dont remember.
> > >
> > >According to the duhvelopers, its an oracle issue not an app issue(i
> > >disagree totally), short of me going thru their code line by
> > line(and i
> > >want to hurt someone since it has goto statements in a
> > stored proc), is
> > >there anything else i can do to determine the sql(on deadlock i get a
> > >trace file so i can see what is happening).
> > >
> > >At first it we thought it had to do with a bitmap index but
> > we converted
> > >it to a normal index and it didnt go away.
> > >
> > >here is part of the deadlock trace file(we're already ran
> > the event for
> > >60, to get more info).
> > >
> > >It appears that there is a share lock issue, anyone willing to take a
> > >stab at this?
> > >
> > >
> > >DEADLOCK DETECTED
> > >Current SQL statement for this session:
> > >INSERT INTO ATTEMPT_STATE (
> > >ATTEMPT_STATE_OID,ATTEMPT_OID,ACTUAL_DATE,PLAN_DATE,
> > >SEQUENCE_NUMBER,NAME ) VALUES ( :b1,:b2,:b3, NULL ,:b4,:b5 )
> > >----- PL/SQL Call Stack -----
> > > object line object
> > > handle number name
> > >82feecd4 296 package body FTC.ATTEMPT_STATE_CHANGE_PKG
> > >82ff75f0 1 anonymous block
> > >The following deadlock is not an ORACLE error. It is a
> > >deadlock due to user error in the design of an application
> > >or from issuing incorrect ad-hoc SQL. The following
> > >information may aid in determining the deadlock:
> > >Deadlock graph:
> > > ---------Blocker(s)--------
> > >---------Waiter(s)---------
> > >Resource Name process session holds waits process session
> > >holds waits
> > >TX-0002001d-0003050c 12 20 X 13
> > >18 S
> > >TX-00050022-000302db 13 18 X 12
> > >20 S
> > >session 20: DID 0001-000C-00000002 session 18: DID
> > >0001-000D-00000002
> > >session 18: DID 0001-000D-00000002 session 20: DID
> > >0001-000C-00000002
> > >Rows waited on:
> > >Session 18: no row
> > >Session 20: no row
> > >
> > >
> > >thanks, joe
> > >
> > >
> > >--
> > >Author: Joseph S. Testa
> > > INET: teci_at_oracle-dba.com
> > >
> > >Fat City Network Services -- (858) 538-5051 FAX: (858) 538-5051
> > >San Diego, California -- Public Internet access /
> > Mailing Lists
> > >--------------------------------------------------------------------
> > >To REMOVE yourself from this mailing list, send an E-Mail message
> > >to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in
> > >the message BODY, include a line containing: UNSUB ORACLE-L
> > >(or the name of mailing list you want to be removed from). You may
> > >also send the HELP command for other information (like subscribing).
> >
> > ______________________________________________________________
> > __________
> > Get Your Private, Free E-mail from MSN Hotmail at
>http://www.hotmail.com
>
>--
>Author: Rachel Carmichael
> INET: carmichr_at_hotmail.com
>
>Fat City Network Services -- (858) 538-5051 FAX: (858) 538-5051
>San Diego, California -- Public Internet access / Mailing Lists
>--------------------------------------------------------------------
>To REMOVE yourself from this mailing list, send an E-Mail message
>to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in
>the message BODY, include a line containing: UNSUB ORACLE-L
>(or the name of mailing list you want to be removed from). You may
>also send the HELP command for other information (like subscribing).
>--
>Author: Toepke, Kevin M
> INET: ktoepke_at_cms.cendant.com
>
>Fat City Network Services -- (858) 538-5051 FAX: (858) 538-5051
>San Diego, California -- Public Internet access / Mailing Lists
>--------------------------------------------------------------------
>To REMOVE yourself from this mailing list, send an E-Mail message
>to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in
>the message BODY, include a line containing: UNSUB ORACLE-L
>(or the name of mailing list you want to be removed from). You may
>also send the HELP command for other information (like subscribing).
Received on Tue Aug 08 2000 - 09:08:38 CDT