Re: Filesystem blocksize for archive logs on AIX

From: Nuno Souto <>
Date: Sat, 26 Oct 2013 23:55:08 +1100
Message-ID: <>

I don't specifically know about the archived redo log block size - as far as I know, there is nowhere it can be specifically checked. In my dbs - with Oracle and Aix 7.1 (both P6 and P7) I am using JFS2 and 4K blocksize for the file systems across the board. I am also using 8K block size for DB files (bigfiles) and 4K for all redo logs - with the "_" parameter that enables setting the redo block size.
With this setup, I have not seen any significant I/O waits measured with AWR exceeding 5ms, with >90% of them being <1ms. That's for datafiles, log files and archived logs, including log file sync. More than happy with that - at the moment. We use a VMAX SAN with 4Gbps FC cards for P6 and 8Gbps FC cards for the P7.

Not sure if there is any mechanism to get detail of wait times with less than <1ms granularity via AWR.
And quite frankly, I've got better things to do with my time than pour through 10046 or such traces to find that. I'd rather wait for Oracle to improve the granularity of AWR timings.

For the time being, as long as the vast majority of waits stay <1ms, I'm happy as a bird.

Nuno Souto

On 25/10/2013 10:33 PM, David Roberts wrote:

> snip...
> My personal observation is that file system block size seems
> significant to oracle for the redo logs. On that basis I would tend
> to use the same block size for the archived logs as you have
> suggested.
> As a final caveat, I've used the terms 4K and advanced format hard
> drives interchangeably. I'm not sure if that is correct or if only
> the 512 bios 4K physical SATA drives will be labeled 'Advanced format
> drives'. Again I'm uncertain as to if Oracle would consider 512 bios
> 4K physical SATA drives 512 or 4K drives for the purposes of setting
> up redo logs.
> Dave
> On Thu, Oct 24, 2013 at 6:02 PM, Rich Jesse <
>> wrote:
>> Hey all,
>> I'm configuring new DBs on a new AIX 7.1 box. Instead of
>> cloning our current setup on AIX 5.3, I'm revisiting all options
>> for both flexibility and performance.
>> The docs
>> (
>> and MOS articles like 1507249.1 are great, but one item I can't
>> find anywhere is a recommendation for a filesystem blocksize for
>> the archive logs. 4K for datafiles, 512 for online redos and
>> controlfile, but nothing for archives.
>> 512 would seem to be the likely answer, but it's been too many
>> years since my v7.3 DBA classes and I can't remember how the
>> archiver does its IO (sorry, Joe!). The docs don't seem to have
>> that answer, either.
Received on Sat Oct 26 2013 - 14:55:08 CEST

Original text of this message