i/o slave wai in 10.2.0.5 Solaris
- From: amonte <ax.mount@xxxxxxxxx>
- To: Oracle-L Group <oracle-l@xxxxxxxxxxxxx>
- Date: Sun, 17 Oct 2010 15:36:22 +0200
Hi
Recently we upgraded a solaris rac database from 10.2.0.4 to 10.2.0.5, after
that we noticed all scatter reads comes with i/o slave wait as well.
For example
select /*+ full(a) */ count(*)
from
SVV.CAP_SIG a where rownum < 500001
call count cpu elapsed disk query current
rows
------- ------ -------- ---------- ---------- ---------- ----------
----------
Parse 1 0.00 0.00 0 0
0 0
Execute 1 0.01 0.00 0 0
0 0
Fetch 2 0.60 2.01 10833 10847
0 1
------- ------ -------- ---------- ---------- ---------- ----------
----------
total 4 0.61 2.02 10833 10847
0 1
Event waited on Times Max. Wait Total
Waited
---------------------------------------- Waited ----------
------------
SQL*Net message to client 3 0.00
0.00
SQL*Net message from client 3 21.58
21.59
library cache pin 1 0.00
0.00
row cache lock 4 0.00
0.00
rdbms ipc reply 1 0.00
0.00
db file sequential read 1 0.01
0.01
* i/o slave wait 677 0.03
1.10
db file scattered read 677 0.03
1.16*
gc cr multi block request 527 0.00
0.39
gc current grant 2-way 10 0.00
0.00
select /*+ full(a) */ *
from
admksk.qos_jpo a where rownum < 500001
call count cpu elapsed disk query current
rows
------- ------ -------- ---------- ---------- ---------- ----------
----------
Parse 1 0.00 0.00 0 0
0 0
Execute 1 0.00 0.00 0 0
0 0
Fetch 33335 9.24 21.93 54240 83659 0
500000
------- ------ -------- ---------- ---------- ---------- ----------
----------
total 33337 9.24 21.94 54240 83659 0
500000
Event waited on Times Max. Wait Total
Waited
---------------------------------------- Waited ----------
------------
SQL*Net message to client 33335 0.00
0.00
gc cr grant 2-way 1 0.00
0.00
db file sequential read 4 0.00
0.00
gc cr multi block request 2731 0.00
1.96
* i/o slave wait 3390 0.04
11.25
db file scattered read 3390 0.04
11.56*
SQL*Net message from client 33335 0.00
3.90
SQL*Net more data to client 186934 0.00
3.11
There is a bug introduced with 10.2.0.5, Bug 9772888 but seems to me is not
a performance bug, rather it generates more trace than usual (at least in
Solaris since AIO is dynamic). Anyone faced this issue :-?
Or this bug actually slows down the queries?
Thanks
Alex
Other related posts:
