Rich,
thank you very much for your clarification and advice.
If we get the chance to do the testing, I'll let you know.
Have a nice Christmas time!
Martin
--
Martin Klier | Performing Databases GmbH
Managing Partner | Senior DB Consultant
Oracle ACE
martin.klier@xxxxxxxxxxxxxxxxx | http://www.performing-databases.com ;
Von: "Rich J" <rjoralist3@xxxxxxxxxxxxxxxxxxxxx>
An: oracle-l@xxxxxxxxxxxxx
Gesendet: Mittwoch, 21. Dezember 2016 21:11:49
Betreff: Re: Oracle 11.2 on AIX 7.1 JFS2 mount options
On 2016/12/21 13:46, Martin Klier - Performing Databases GmbH wrote:
I am calling out for Oracle-on-AIX knowledge/experience bearers, and have a
question whether setting the "cio" mount option for datafiles, redo logs and
control files for DB version 11.2.0.4 on AIX 7.1 with JFS2 or not.
We are using filesystemio_options=setall for this customer setup.
...
Simple question:
I have massively concurrent IO on datafiles (usual read/write activity) and
redo
logs (streams capture) with many (so far unavoidable) "db file sequential
read"
and "log file sync" waits.
Should I activate JFS2 mount option -cio for my data files and redo log file
systems?
Short answer: use "filesystemio_options=setall" in Oracle 11.2 and do not use
the "cio" mount option on JFS2 mountpoints in AIX 7.x.
You can verify that CIO is being used by tracing sessions on a test Oracle
instance. A google and MOS 1478924.1 can help you with that. Please test! When
you run into issues in the future, you can prove that it's not a lack of CIO
that is the problem.
Also see the "AIX Concurrent I/O for Oracle archive logs" thread on this list
from 2010.
"log file sync" waits do not necessarily come from slow disk resulting from
suboptimal IO setup. I have very low latency writes, but have log file sync
waits due to over-committing, for (a poor) example.
GL!
Rich
