This wait happens when a session wants to access a database block in the buffer cache but it cannot because the buffer is busy. Another session is modifying the block and the contents of the block are in flux during the modification. To guarantee that the reader has a coherent image of the block with either all of the changes or none of the changes, the session modifying the block marks the block header with a flag letting other users know a change is taking place and to wait until the complete change is applied.
The two main cases where this wait can occur are:
Another session is reading the block into the buffer
Another session holds the buffer in an incompatible mode to our request
While the block is being changed, the block is marked as unreadable by others. The changes that are being made should last under a few hundredths of a second. A disk read should be under 20 milliseconds and a block modification should be under one millisecond. Therefore it will take a lot of buffer busy waits to cause a problem.
However, in a problem situation, there is usually a hot block, such as the first block on the free list of a table, with high concurrent inserts. All users will insert into that block at the same time, until it fills up, then users start inserting into the next free block on the list, and so on.
Another example of a problem is of multiple users running full table scans on the same large table at the same time. One user will actually read the block physically off disk, and the other users will wait on Buffer Busy Wait for the physical I/O to complete.
The rest of the information in this section is only valid for this metric when it appears in either the Enterprise Manager Grid Control or the Enterprise Manager Database Control (if applicable).
The following table shows how often the metric's value is collected and compared against the default thresholds. The 'Consecutive Number of Occurrences Preceding Notification' column indicates the consecutive number of times the comparison against thresholds should hold TRUE before an alert is generated.
|
Target Version |
Evaluation and Collection Frequency |
Upload Frequency |
Operator |
Default Warning Threshold |
Default Critical Threshold |
Consecutive Number of Occurrences Preceding Notification |
Alert Text |
|
pre-10g |
Every Minute |
After Every Sample |
> |
20 |
Not Defined |
3 |
%value%%% of service time is spent waiting on the 'buffer busy waits' event. |
(DeltaBufferBusyWaitsTime/DeltaServiceTime)*100 where:
DeltaBufferBusyWaitsTime: difference of 'sum of time waited for sessions of foreground processes on the 'buffer busy waits' event' between sample end and start
DeltaServiceTime: difference of 'sum of time waited for sessions of foreground processes on events not in IdleEvents + sum of 'CPU used when call started' for sessions of foreground processes' between sample end and start
See Idle Events
Look at v$waitstat (or the buffer busy drill down page) and determine the block type with the highest waits.
Block Type and Action:
Undo Header - Use Automatic Undo Management (AUM) or add more RBS segments)
Undo Block - Use AUM (or increase RBS sizes)
Data Block - First determine if it is an I/O problem. The Buffer Busy Waits drill-down page should provide this information. Otherwise, sample from v$session_wait
SELECT p3, count(*) FROM v$session_wait WHERE event='buffer busy wait' ;
If p3 is less than 200 then it is an I/O problem. Either improve I/O performance or change application. Applications running concurrent batch jobs that do full table scans on the same large tables run into this problem.
Free List - Use ASSM (or freelists groups)
Related Topics
About Alerts
About the Metric Detail Page
Editing Thresholds
Understanding Line Charts
Copyright © 1996, 2009, Oracle and/or its affiliates. All rights reserved.
Oracle is a registered trademark of Oracle Corporation and/or its affiliates.
Other names may be trademarks of their respective owners.