I too had this problem in 8.1.6 Never found an answer to it...., upgrade
to 8.1.7 and it went away. May be a bug ????
--
John Jones
Senior Oracle DBA
Duke University, OIT
john.jones_at_duke.edu
Chris Kordish <chris.kordish_at_Sun.COM> wrote in message
news:3AF8D664.1B5FC894_at_Sun.COM...
> 1) is it unix permissions ?
>
> 2) are you running this as the oracle DBA user?
>
> 3) maybe if you are using c-shell
> and you have
>
> set noclobber
>
> and you 'alter .. to trace' and it writes it
> once then if you try to do it again
> the 'set noclobber' wont let it fly
>
> are you toggling between CSH and SH or KSH ?
>
> 4) if you're solaris saavy then
> in another window you can truss -f -o truss.out
> on your svrmgrl session and then see why the
> write() doens't occur ... usually a failed syscall
> gives the real answer ...but reading it can be a pain
> sometimes
>
> 5) if oracle writes out controlfile to trace
> using buffered io then you may need to kill your
> oracle session ( exit it ) to close the file
> descriptor representing the trace file in the unix filesystem
>
>
> good luck
> chris
>
>
> Andrew Au wrote:
> >
> > Has anyone seen this problem before?
> > I am running Oracle 8.1.6 on Solaris. When I issue
> > Alter database backup controlfile to trace
> >
> > it is supposed to generate a trace file in the udump directory. But
> > sometimes it does not create anything in the udump. It just executes the
> > command but no trace file is generated. I checked the udump directory as
> > specified in:
> >
> > show parameters dump
> >
> > so I am sure I am looking at the right directory. Sometimes it works and
> > sometimes it does not. This is driving me crazy.
> >
> > Does anyone have any ideas?
> >
> > Please Email me as well as posting to this group.
Received on Mon May 14 2001 - 08:38:45 CDT
