Skip to Main Content

Oracle Database Discussions

Announcement

For appeals, questions and feedback about Oracle Forums, please email oracle-forums-moderators_us@oracle.com. Technical questions should be asked in the appropriate category. Thank you!

Question on flash back waits

User_OCZ1TDec 27 2019 — edited Jan 8 2020

Hi we are using version 12.1.2.0 of Oracle Exadata. And we are seeing high flash back related waits(flashback buf free by RVWR) during peak hours in our database and the DMLs are suffering. We want to see if we can reduce this by some way as we need to have the flashback turned on as per business agreement.

Below is the specific sections of AWR showing "flashback buf free by RVWR" as the second top most "foreground wait event". As per the support team we got the recommendation stating , " We have ~27 log switches per hour which is high and it should be ~4 per hour. And the issue is related to the I/O bottleneck, in order to reduce the write I/O try to increase REDO size and have less REDO switched. The scope of REDO increase is to reduce the log switched and give bandwidth to flashback writer between log switches." ..

Now i am not able to understand how the minimization of redo switch will help in minimizing the "flashback buf free by RVWR" wait. As because RVWR writes the flash back buffer to the flash back log which is off-course on same disk as redolog file but both(redolog and flash back log) writes are performed by two different processes and if we will increase the redolog file size to reduce the log switch then also the amount of log write which has to happen will remain same. So need experts thoughts on, how it will reduce the I/O and help us here? I am not able to relate the flash back log writer i.e RVWR to the redolog writes here. And is there any possible negative side of having the size increased for the redolog file?

Below is the specific section of the AWR for peak 2hrs snap:-

Snap IdSnap TimeSessionsCursors/SessionInstances
Begin Snap:27502601-Nov-19 05:45:21321  40.72
End Snap:27503401-Nov-19 07:45:24373  42.02
Elapsed:   120.05 (mins)
DB Time:   5,088.47 (mins)

Load Profile

Per SecondPer TransactionPer ExecPer Call
DB Time(s):  42.4  0.1  0.01  0.00
DB CPU(s):  7.8  0.0  0.00  0.00
Background CPU(s):  1.4  0.0  0.00  0.00
Redo size (bytes):  28,606,192.3  71,380.8
Logical read (blocks):  509,949.3  1,272.5
Block changes:  92,065.5  229.7
Physical read (blocks):  27,320.1  68.2
Physical write (blocks):  5,663.6  14.1
Read IO requests:  11,344.7  28.3
Write IO requests:  2,839.6  7.1
Read IO (MB):  213.4  0.5
Write IO (MB):  44.3  0.1
IM scan rows:  0.0  0.0
Session Logical Read IM:
Global Cache blocks received:  526.7  1.3
Global Cache blocks served:  149.1  0.4
User calls:  13,970.3  34.9
Parses (SQL):  1,986.7  5.0
Hard parses (SQL):  0.1  0.0
SQL Work Area (MB):  769.4  1.9
Logons:  1.1  0.0
Executes (SQL):  4,895.4  12.2
Rollbacks:  0.0  0.0
Transactions:  400.8

Top 10 Foreground Events by Total Wait Time

EventWaitsTotal Wait Time (sec)Wait Avg(ms)% DB timeWait Class
cell single block physical read79,110,64292.2K1.1730.2User I/O
flashback buf free by RVWR107,07968.8K642.5322.5Configuration
DB CPU 56.4K 18.5
buffer busy waits5,757,49047.2K8.1915.4Concurrency
enq: TX - index contention1,137,92628.4K24.969.3Concurrency
log file sync2,907,8935129.61.761.7Commit
flashback log switch5,4454318.5793.111.4Other
cell multiblock physical read1,111,6271827.71.64.6User I/O
read by other session270,4341341.44.96.4User I/O
gc cr block busy1,163,8321262.71.08.4Cluster

Wait Classes by Total Wait Time

Wait ClassWaitsTotal Wait Time (sec)Avg Wait (ms)% DB timeAvg Active Sessions
User I/O81,063,59597,6561.2032.013.6
System I/O8,366,95792,66011.0730.312.9
Concurrency7,531,61275,79410.0624.810.5
Configuration160,30171,145443.8223.39.9
DB CPU 56,359 18.57.8
Other18,947,71713,3410.704.41.9
Commit2,908,0915,1301.761.70.7
Cluster6,421,8152,0140.31.70.3
Application10,04969469.11.20.1
Network87,363,494640.00.00.0

Host CPU

CPUsCoresSocketsLoad Average BeginLoad Average End%User%System%WIO%Idle
  32  16  2  3.72  3.54  25.7  4.6  0.0  68.5

Instance CPU

%Total CPU%Busy CPU%DB time waiting for CPU (Resource Manager)
  29.1  92.3  0.0

IO Profile

Read+Write Per SecondRead per SecondWrite Per Second
Total Requests:  15,067.0  11,511.0  3,556.0
Database Requests:  14,184.3  11,344.7  2,839.6
Optimized Requests:  11,383.6  11,189.4  194.2
Redo Requests:  656.6  43.2  613.4
Total (MB):  465.3  271.6  193.8
Database (MB):  257.7  213.4  44.3
Optimized Total (MB):  223.9  209.7  14.2
Redo (MB):  99.8  43.1  56.7
Database (blocks):  32,983.7  27,320.1  5,663.6
Via Buffer Cache (blocks):  28,613.7  23,101.4  5,512.3
Direct (blocks):  4,370.0  4,218.7  151.3

Memory Statistics

BeginEnd
Host Mem (MB):  257,973.7  257,973.7
SGA use (MB):  45,056.0  45,056.0
PGA use (MB):  6,105.5  6,094.5
% Host Mem used for SGA+PGA:  19.83  19.83

Cache Sizes

BeginEnd
Buffer Cache:  35,200M  35,200MStd Block Size:  8K
Shared Pool Size:  8,192M  8,192MLog Buffer:  254,664K
In-Memory Area:  0M  0M

Shared Pool Statistics

BeginEnd
Memory Usage %:  82.48  80.86
% SQL with executions>1:  99.37  99.12
% Memory for SQL w/exec>1:  98.66  97.72

Background Wait Events

  • ordered by wait time desc, waits desc (idle events last)
  • Only events with Total Wait Time (s) >= .001 are shown
  • %Timeouts: value of 0 indicates value was < .5%. Value of null is truly 0
EventWaits%Time -outsTotal Wait Time (s)Avg wait (ms)Waits /txn% bg time
db file parallel write3,033,926060,39919.911.0552.75
Log archive I/O90,057012,305136.630.0310.75
flashback log file write467,82608,19317.510.167.16
log file sequential read215,53508,17537.930.077.14
ASM file metadata operation71,40903,05142.720.022.66
KSV master wait41,501113,02572.890.012.64
log file parallel write2,199,36502,0330.920.761.78
control file parallel write8,591080693.820.000.70
flashback log file sync72251495685.190.000.43
enq: RO - fast object reuse4,633042992.500.000.37
flashback buf free by RVWR79537420528.760.000.37
reliable message28,772034612.040.010.30
flashback log file read19,831033216.750.010.29
control file sequential read596,42503260.550.210.28
LGWR wait for redo copy2,874,79802090.071.000.18
Disk file operations I/O137,32701591.160.050.14
cell single block physical read33,90101173.460.010.10
enq: CF - contention2,5288111143.880.000.10
change tracking file synchronous read2,773011039.610.000.10
db file single write4970104209.750.000.09

Instance Activity Stats - Thread Activity

  • Statistics identified by '(derived)' come from sources other than SYSSTAT
StatisticTotalper Hour
log switches (derived)5426.99

Key Instance Activity Stats

  • Ordered by statistic name
StatisticTotalper Secondper Trans
db block changes663,162,53292,065.48229.73
execute count35,262,5334,895.4212.22
gc cr block receive time137,10619.030.05
gc cr blocks received2,021,649280.660.70
gc current block receive time25,8203.580.01
gc current blocks received1,772,386246.060.61
logons cumulative8,2331.140.00
opened cursors cumulative15,501,2442,152.015.37
parse count (total)14,310,3651,986.684.96
parse time elapsed24,7873.440.01
physical reads196,791,19727,320.1168.17
physical writes40,795,7865,663.5914.13
redo size206,055,036,97228,606,192.2571,380.75
session cursor cache hits2,274,911315.820.79
session logical reads3,673,247,167509,949.261,272.47
user calls100,630,07213,970.2634.86
user commits2,886,692400.751.00
user rollbacks110.000.00
workarea executions - onepass00.000.00
workarea executions - optimal3,477,519482.781.20
Comments
Post Details
Added on Dec 27 2019
22 comments
3,786 views