We have 2 SRs currently open for this same issue, and
ever since we set the param to FALSE, we haven't seen
the issue repeat (so, yes it's mutex related). We saw
a case where 1 session blocks another, we have also
seen this happening with parallel process, where
parent and child get locked on each other.

5737552.993 - SESSION WAITING VERY LONG ON 'LATCH:
LIBRARY CACHE'
5703409.993 - SESSION HANGS DUE TO WAITS ASSOCIATED
WITH ENQ: PS - CONTENTION

HTH,
Deepak

--- Charles Schultz <sacrophyte@(protected):

> 10.2 turns on mutexes by default. The short-term
> workaround is to turn off
> mutexes (_kks_use_mutex_pin = FALSE). There are
> several related bugs,
> notably:
> 5184776 10.2.0.2.0 RDBMS 10.2.0.2.0 PRODID-5
> PORTID-23
> Abstract: HIGH 'CURSOR: PIN S WAIT ON X'
>
> We had one database that blocked on some dictionary
> objects, prohibiting all
> logins (even as sysdba). Unfortunately, since we
> were not able to reproduce,
> we had to close the case as Oracle wanted a
> hanganalyze. If you can get more
> information from various mutex related views, I
> strongly advise you open a
> case with Oracle.
>


__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
--
http://www.freelists.org/webpage/oracle-l