Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: Monitoring questionable long running sessions

Re: Monitoring questionable long running sessions

From: Rick Denoire <100.17706_at_germanynet.de>
Date: Thu, 15 Jul 2004 00:04:22 +0200
Message-ID: <o3bbf0d2a6q8u0v2v13sram92knapqoeda@4ax.com>


yong321_at_yahoo.com (Yong Huang) wrote:

>Does the seq# 5625 increment? What does this query give?
>
>select * from dba_extents where file_id = 33 and 213471 between
>block_id and block_id + blocks - 1
>
>It must be an index. Does analyze index <thatindex> validate structure
>throw an error?

Well, I would have to restart the thing and adapt the numbers to do the query.

>It's possible the session waited on this event 83911 seconds ago and
>is no longer waiting. Then the problem may be somewhere else. Can you
>still do a SQL trace on the session (exec
>dbms_system.set_ev(102,5625,10046,12,''))? If you strace the server
>process as VC suggested and it doesn't show anything useful, how about
>ltrace -p <pid>? That's a tool I'm still studying so I may not be able
>to infer anything from it

Would you be surprised that using exactly that command "dbms_system..." did not produce any trace file? Same result trying dbms_system.set_sql_trace_in_session...

After sending tons of files to the Oracle support they concluded that the session was "spinning". I am awaiting a patch or some solution.

The outstanding "feature" of this problem session is that it does nothing at maximum speed. For ever.

Bye
Rick Denoire Received on Wed Jul 14 2004 - 17:04:22 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US