"Jack" <Jack_at_none.com> wrote in message
news:4V7Pg.49$u87.30_at_read3.inet.fi...
>
> "Sybrand Bakker" <postbus_at_sybrandb.demon.nl> kirjoitti
> viestissä:pggmg29n2g49ai8j7b92o3bubl38r7onv7_at_4ax.com...
>> On Fri, 15 Sep 2006 22:47:10 GMT, "Jack" <Jack_at_none.com> wrote:
>>
>>>That i/o feels a bit sticky, maybe it is Solaris problem?
>>>Hopefully there is some manageable fix, i do not have contact with SA.
>>>And these kind of things cannot be done without hard evidence.
>>
>> Hard evidence can be obtained by running truss -p on the dbw<n>
>> process.
>> You'll see every lpwrite (IIRC) fail, and it will be followed by an
>> ordinary pwrite.
>>
>> And, IIRC, Solaris 5.8 doesn't have asynchronous I/O on default ufs
>> file systems.
>> You may also need to look in mounting the filesystems with
>> forcedirectio. This will bypass the Solaris filesystem cache. Note:
>> the pagefile is IN the filesystem cache.
>> There are several whitepapers on this topic, obtainable from Sun, or
>> Sunsolve.
>>
>> --
>> Sybrand Bakker, Senior Oracle DBA
> Hi!
>
> Thanks for your info.
> Can you give some links for those pages.
> I am not so familiar with Solaris.
>
> Jack
>
Received on Sun Sep 17 2006 - 11:45:12 CDT
