Re: ST-enqueue problem

First thing of course is to move off 8.0.1. (I didn't know there was one). Even in 8.0.3 I found lots of problems with OPS.

The ST enqueue has to be held for a
space transaction - i.e. manipulation
of the uet$, fet$ tables.

Since you are rebuilding indexes
across all nodes all nodes are going
to need to manipulate some blocks in
fet$, uet$.

Even though each node is working
against a single tablespace,
perhaps you still have some blocks
in the two tables (or clusters holding
the two tables) which share the same
DLM lock element - so you could have
pinging occuring which slows down the
acquisition and release of the ST lock.

--

Jonathan Lewis
Yet another Oracle-related web site: http://www.jlcomp.demon.co.uk

Cho, Hyon-Hyok wrote in message <80r8tj$6tf$2_at_news2.kornet.net>...
>Hi !!
>I am running oracle8.0.1 OPS on IBM SP Machine...
>On my system, st-enqueue problem is very serious.
>so I'm very careful whenver I create any objects.
>I can bear it, too small initial or next size can cause
>st-enqueue contention.
>
>But I can't understand why st-enqueue arose when
>I rebuild index!!!!(alter index ~ rebuild ;)
>My index-rebuild script performe rebulding across the
>whole nodes, and each node performe rebulding
>on its own tablespaces.
>But when I start rebulding, I found that every command
>running across all nodes got a contentition problem.
>
>Is there any parameter that can avoid st-enqueue
>contention problem?
Received on Fri Nov 26 1999 - 14:24:34 CST