Jonathan,
OK, what am I missing? All of the PX sessions have to
access the segment header in order to get blocks off
the free lists so can't that result in buffer busy
waits on the segment header unless it was created with
more than one free list group?
-- Anita
--- Jonathan Lewis [EMAIL
The original comment was about PQ
slaves reading data blocks - not
PX slaves running parallel update/inserts.
When PX slaves do parallel inserts the
processing is usually split to avoid
contention - e.g. each slaves gets
one partition of a partitioned object;
or each slave gets a new extent
Raja:
You will know you have freelist contention if you have a significant
buffer busy waits ratio (5%).
Jon Walthour
-Original Message-
Luthra
Sent: Monday, August 20, 2001 11:45 PM
To: Multiple recipients of list ORACLE-L
Hello folks,
How do I come to know that there is a
Bufer busy waits can also be caused by parallel query
servers trying to read the same buffer block at the
same time. So does it always indicate a freelist
problem?
.
--- Jon Walthour [EMAIL PROTECTED] wrote:
Raja:
You will know you have freelist contention if you
have a significant
buffer
You have to learn to listen to your database, grasshoppa'. A good
alternative to the Buddhist and kung fu techniques is the v$waitstat
table. If you see accumulating time for 'extent map' or 'free list',
then you know that you have to rebuild the table with more free lists.
To see the actual
Title: RE: Freelist Contention - PQ slaves
I thought that parallel query slaves were smart enough to divide up the work between them so there was no overlap? What am I missing? Can you elaborate?
-Original Message-
From: Johnson Poovathummoottil [SMTP:[EMAIL PROTECTED]]
Sent: Tuesday
recipients of list ORACLE-L [EMAIL PROTECTED]
Subject: RE: Freelist Contention
Bufer busy waits can also be caused by parallel query
servers trying to read the same buffer block at the
same time. So does it always indicate a freelist
problem?
.
--- Jon Walthour [EMAIL PROTECTED] wrote:
Raja
On Tuesday 21 August 2001 08:21, Johnson Poovathummoottil wrote:
Bufer busy waits can also be caused by parallel query
servers trying to read the same buffer block at the
same time.
Just off the top of my head, and without giving it a lot
of thought, this doesn't sound right.
PQS divvy up
Another thing that comes to mind is incorrectly tuned DBWR, which cannot
keep up.
-Original Message-
From: Deshpande, Kirti [mailto:[EMAIL PROTECTED]]
Sent: Tuesday, August 21, 2001 2:31 PM
To: Multiple recipients of list ORACLE-L
Subject: RE: Freelist Contention - PQ slaves
Not only that, PX slaves do direct reads
anyway, bypassing the buffer cache.
It is possible, though, that the flush
that has to take place before the PX
scan can read back is sufficient to
cause other processes to wait for
write complete waits, and I guess that
you could also get extra rbs
PROTECTED]]
Sent: Tuesday, August 21, 2001 11:21
To: Multiple recipients of list ORACLE-L
Subject:RE: Freelist Contention
Bufer busy waits can also be caused by parallel
query
servers trying to read the same buffer block at
the
same time. So does it always indicate
- Original Message -
To: Multiple recipients of list ORACLE-L [EMAIL PROTECTED]
Sent: Tuesday, August 21, 2001 9:48 AM
On Tuesday 21 August 2001 08:21, Johnson Poovathummoottil wrote:
Bufer busy waits can also be caused by parallel query
servers trying to read the same buffer
by: Subject: Re: Freelist Contention
[EMAIL PROTECTED
Raja, You will only have freelist contention on
conventional insert operations only. Insert operations
from direct loads won't cause freelist contention
because of the way direct inserts are done. You can
check for freelist contention by checking
V$SESSION_WAIT and look for the Event 'Buffer Busy
14 matches
Mail list logo