RE: linux and oracle

Yeah,

I've seen it on 9.2.0.6 IIRC.

The j000 process hasn't released its file descriptor despite the tablespaces were dropped and files removed.
A process-state dump though would have shown the file handles still there.

A workaround instead of shutdown would have been to alter system set job_queue_processes=0, wait until all procs die (or kill 'em if necessary) and then set job_queue_processes back again.

Hmm.. this could make a good entry into my new blog (wink-wink ;)

Tanel.
http://blog.tanelpoder.com

> -----Original Message-----
> From: oracle-l-bounce_at_freelists.org
> [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Johan Eriksson
> Sent: Tuesday, June 19, 2007 20:45
> To: oracle-l_at_freelists.org
> Subject: linux and oracle
>
> Hi
>
> Perhaps more a linux problem than oracle... Database is 10.2
> and the system is a Red hat enterprise 3.4
>
> What I did was deleteing alot of tablespaces (first drop the
> users and then the tablespace) which occupied 12gb of data
> and the files were deleted from the filesystem as expected,
> but the disk usage (with df) showed that the used space was
> still 100% but when examed with "du --max-depth=1" it showed
> that it should have 12gb free.
> When examined with lsof the deleted files where reported as
> opened (with the flag deleted at the end) and owned by the
> processes j000 and j000.

--
http://www.freelists.org/webpage/oracle-l