RE: what's mean of "enqueue hash chains" latch?
- From: Riyaj Shamsudeen <rshamsud@xxxxxxxxxxxx>
- To: Peter.Hitchman@xxxxxxxxxxx, oracle-l@xxxxxxxxxxxxx, oracle.unix@xxxxxxxxx
- Date: Sat, 08 Jan 2005 12:21:24 -0600
Hi eygle
DML locks protects objects from concurrent modification.
Frequently, dml lock allocation latch contention is seen with enqueue
hash chain latches, as dml_locks are implemented through TM enqueue
locks. These resource structures are hanging from enqueue hash chains
serialized by enqueue hash chain latches. So, reducing dml lock
allocation latch contention should resolve enqueue hash chain latch
contention.
As Peter mentioned, these locks are released after each commits.
Since you are using append hint, I am positive that you must commit
after every statement, so your commit frequency is probably high.
Further, if you must have this much contention, then I *guess* this is
probably OLTP kind of application and append and parallel hints might
not be better suited.
Also, append hint will add data after the high water mark and
not sure, how your extents are setup. So if your extents are smaller,
this will lead in to TM locking contention for dictionay objects itself.
If I were you, I would looking to eliminate these append and
parallel hints for inserts. Sorry, I didn't test this *theory* ;-(
Thanks
Riyaj "Re-yas" Shamsudeen
Certified Oracle DBA
-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx
[mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of
Peter.Hitchman@xxxxxxxxxxx
Sent: Saturday, January 08, 2005 11:01 AM
To: oracle-l@xxxxxxxxxxxxx
Subject: RE: what's mean of "enqueue hash chains" latch?
Hi,
From MetaLink:-
"dml lock allocation
This latch protects the list of State Objects (dml locks). Every time a
transaction modifies a table, a DML lock is gotten and released when the
change is committed. The number of State Objects for dml locks is
determined by the init.ora <Parameter:DML_LOCKS>".
So I would be looking at how often you commit.
Also what sort of turn over are you getting on the online redo logs? It
may well be that they are too small. You may want to investigate having
multiple freelists for the table you are inserting into, having many
parallel inserts (assuming you have enabled parallel DML), may well be
causing a lot of contention for the header block at the start of the
segment.
Good luck.
Pete
-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx
[mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of eygle
Sent: 08 January 2005 13:56
To: oracle-l@xxxxxxxxxxxxx
Subject: what's mean of "enqueue hash chains" latch?
hello;
We have a Oracle 8.1.7.0.0 Database on Solaris8.
With a parallel insert , database slow down heavy suddently.
I find lots of latch wait in database,from statspack(with 15 minutes
elapse) list:
Pct Avg
Pct
Get Get Slps NoWait
NoWait
Latch Name Requests Miss /Miss Requests
Miss
----------------------------- -------------- ------ ------ ------------
------
active checkpoint queue latch 260 0.0 0
cache buffers chains 47,228,551 0.0 552
0.0
checkpoint queue latch 3,263 0.0 0
dml lock allocation 46,899,746 12.1 0.0 0
enqueue hash chains 46,899,333 39.2 0.0 0
enqueues 849 0.0 0
job_queue_processes parameter 12 0.0 0
ktm global data 2 0.0 0
library cache 32,739 0.0 0.0 0
library cache load lock 70 0.0 0
list of block allocation 36 0.0 0
and
Latch Sleep breakdown for DB: QCB Instance: qcb Snaps: 150 -151
-> ordered by misses desc
Get Spin
&
Latch Name Requests Misses Sleeps Sleeps
1->4
-------------------------- -------------- ----------- -----------
------------
enqueue hash chains 46,899,333 18,367,985 6,312
18361675/630
8/2/0/0
dml lock allocation 46,899,746 5,678,834 1,228
5677606/1228
/0/0/0
-------------------------------------------------------------
Latch Miss Sources for DB: QCB Instance: qcb Snaps: 150 -151
-> only latches with sleeps are shown
-> ordered by name, sleeps desc
NoWait Waiter
Latch Name Where Misses Sleeps
Sleeps
------------------------ -------------------------- ------- ----------
-------
dml lock allocation ktaiam 0 615
444
dml lock allocation ktaidm 0 604
784
enqueue hash chains ksqrcl 0 3,778
3,453
enqueue hash chains ksqgtl3 0 2,472
2,832
-------------------------------------------------------------
I can not find more info about "enqueue hash chains".
What' s it mean and how to reduce it ?
The SQL of parallel insert is :
insert /*+ append parallel(fc_costgatherresult_m,4) */ into=20
fc_costgatherresult_m
(kjnd,kjqj,pk_dwbm,pk_deptdoc,pk_psndoc,isproduct,pk_productorbalance,ac
countcurrtype,producttype,pk_costitem,originaldwbm,originaldept,calmny,p
k_client,clienttype,pk_credittype)
select /*+ parallel(fc_costcalresult_m,4) */
'2004','11',pk_dwbm,pk_deptdoc,pk_psndoc,isproduct,pk_productorbalance,a
ccountcurrtype,producttype,pk_costitem,originaldwbm,originaldept,sum(cal
mny),pk_client,clienttype,pk_credittype
from fc_costcalresult_m where pk_dwbm =3D originaldwbm
group by
pk_dwbm,pk_deptdoc,pk_psndoc,isproduct,pk_productorbalance,accountcurrty
pe,producttype,pk_costitem,originaldwbm,originaldept,pk_client,clienttyp
e,pk_credittype
All the table is partition table.
Table fc_costcalresult_m with nearly 40G data.
And we have the parameter:
dml_locks =3D 2000
enqueue_resources =3D 2200
I also want to know why "dml lock allocation" Requests is so high?
Any suggestion is appreciate.
----------------------------------------
eygle from China.
my site: http://www.eygle.com
--
//www.freelists.org/webpage/oracle-l
--
//www.freelists.org/webpage/oracle-l
-- Attached file included as plaintext by Ecartis --
-- Desc: Signature
The information transmitted is intended only for the person or entity to
which it is addressed and may contain confidential and/or privileged
material. If the reader of this message is not the intended recipient,
you are hereby notified that your access is unauthorized, and any review,
dissemination, distribution or copying of this message including any
attachments is strictly prohibited. If you are not the intended
recipient, please contact the sender and delete the material from any
computer.
--
//www.freelists.org/webpage/oracle-l
Other related posts:
