RE: Deadlock problem
- From: "Bobak, Mark" <Mark.Bobak@xxxxxxxxxxxxxxx>
- To: "Jared Still" <jkstill@xxxxxxxxx>
- Date: Mon, 8 May 2006 14:08:20 -0400
Huh?
Sure, missing indexes on FKs can cause deadlocks. Well, ok, not having
an index on a FK column on the child table does not inherently cause a
deadlock, but the absence of the index certainly increases the
likelihood of an enqueue wait (and possibly a deadlock) duing concurrent
DML.
-Mark
--
Mark J. Bobak
Senior Oracle Architect
ProQuest Information & Learning
For a successful technology, reality must take precedence over public
relations, for Nature cannot be fooled. --Richard P. Feynman, 1918-1988
________________________________
From: Jared Still [mailto:jkstill@xxxxxxxxx]
Sent: Monday, May 08, 2006 2:06 PM
To: Bobak, Mark
Cc: Thomas.Mercadante@xxxxxxxxxxxxxxxxx; alever@xxxxxxxxx; Oracle
Freelists.org
Subject: Re: Deadlock problem
... which also do not cause deadocks.
On 5/8/06, oracle-l-bounce@xxxxxxxxxxxxx < oracle-l-bounce@xxxxxxxxxxxxx
<mailto:oracle-l-bounce@xxxxxxxxxxxxx> > wrote:
If it were FKs that are missing indexes, waits would be on a TM
enqueue,
not a TX enqueue.
-Mark
--
Mark J. Bobak
Senior Oracle Architect
ProQuest Information & Learning
For a successful technology, reality must take precedence over
public
relations, for Nature cannot be fooled. --Richard P. Feynman,
1918-1988
-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx
[mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Mercadante,
Thomas F
(LABOR)
Sent: Monday, May 08, 2006 1:51 PM
To: alever@xxxxxxxxx; Oracle Freelists.org
Subject: RE: Deadlock problem
Allesandro,
There were two theories concerning where a commit statement
should be
issued from.
The first theory I heard from Oracle when PL/SQL first came out
was that
packages should not have commit statements in them - that the
application should issue the commit when all of the pieces of
work were
completed. It was thought that the application would better
know when a
commit should be issued.
The other theory was to put all of the work in the PL/SQL
packages and
let it control everything and either report back success
(commit) or
failure (rollback) to the application.
Today, either way works just fine in my view.
As for your problem, dig a little deeper. Most deadlocks that
I've seen
are caused by foreign key constraints and missing indexes. So
look at
the tables involved and look for the table being updated being
referenced by another table via a FK. Simply adding indexes to
the
foreign key columns solves this problem.
And remember - this is an application problem. Somebody might
have to
fix some code!
Good Luck.
Tom
-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx
[mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Alessandro
Vercelli
Sent: Monday, May 08, 2006 1:34 PM
To: Oracle Freelists.org
Subject: Deadlock problem
Hi all,
I'm trying to solve an ora-4020 (deadlock) issue; the trace file
(sorry
if messed) shows:
ksqded1: deadlock detected via did
DEADLOCK DETECTED
Current SQL statement for this session:
Update <TABLE> set <FIELD1>='<VALUE>' where <FIELD2> like
'<STRING>%'
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-00040015-0000305b 13 11 X 10
14
X
TX-0007000b-0000309e 10 14 X 13
11
X
session 11: DID 0001-000D-00000001 session 14: DID
0001-000A-00000001
session 14: DID 0001-000A-00000001 session 11: DID
0001-000D-00000001
Rows waited on:
Session 14: obj - rowid = 0000147E - AAABR+AAKAAAAJIAAH Session
11: obj
- rowid = 0000147E - AAABR+AAKAAAzEeAAH
The trace file shows clearly that session 11 and 14 are blocking
each
other.
Note that <STRING> can be very long, but it's almost certain
that this
is not causing the problem.
Database version is 8.0.5 on Solaris 8 sparc.
So, I'm looking at the piece of source containing the affected
code (I'm
not the developer neither a skilled one) and I have seen
something
strange, that is a sql package containing many procedures with
insert/update statements and none of these insert/update was
followed by
a commit; I asked the developer for this matter and she said
that a
commit would prevent a possible rollback of database
transaction.
Now, my questions are:
1. Is it correct an insert/update without a commit into a sql
package?
If yes, when are the inserted/updated data commited?
2. Would this the possible cause of the deadclock, as the table
indexes
could be locked by a large number of records inserted/updated?
3. Is this the correct way to get the choice of performing a
rollback?
Thanks for you help,
Alessandro
--
//www.freelists.org/webpage/oracle-l
--
//www.freelists.org/webpage/oracle-l
--
//www.freelists.org/webpage/oracle-l
--
Jared Still
Certifiable Oracle DBA and Part Time Perl Evangelist
Other related posts:
