Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
Home -> Community -> Usenet -> c.d.o.server -> REPOST: Re: Async I/O is NOT supppressed for filesystem datafiles in 8.1.5?
Please show the truss output. Oracle by itself does not prevent kaio calls. But
in Solaris they return with an ENOTSUP exception and Oracle reverts to
pwrite/pread calls. So Oracle tries whatever you tell it in init.ora. If you put
disk_asynch_io = true there it will try kaio calls. If they fail it will go back
to pread/pwrite.
On Solaris it is safe to request kaio on filesystems but it is slow because you add the overhead of exception handling *after each kaio call*.
Regards,
Martin
El Toro wrote:
>
> I've read both Metalink and Sun Blueprint documents that read
> how Oracle 8.1.5 and 8.1.6 will NOT use AIO on filesystem datafiles.
> It runs 'fstat()' to determine if the files are on filesystems, and
> acts accordingly (according to Oracle and Sun). Therefore, turning
> AIO on via 'disk_asynch_io=true' on such databases is futile, based
> on these two sources.
>
> However, I have 8.1.5 running on NFS *filesystems* from NetApps, and
> they are, in fact, using AIO. I've verified this by running a 'truss'
> on the DBWR process, noticing all the LWP processes spawned to perform
> the user-level library AIO after a failed 'kaio()' call (tell-tale
> signs of AIO). The OS is Solaris 2.6 - 2.8, by the way.
>
> So, what's the deal? Does Oracle 8.1.5 and .6 use AIO on filesystems
> or not? When is it that Oracle suppresses AIO on filesystems? I gotta
> save face here in front of my team members to whom I insisted that
> Oracle did away with filesystem AIO on filesystems. Please respond
> quickly, as I have a team meeting about this issue tomorrow.
>
> Thanks in advance.
This message was cancelled from within Mozilla...not Received on Fri Dec 28 2001 - 04:16:51 CST