Re: how to monitor the progress of inserts

At 03:50 PM 2/1/2005, Mladen Gogala wrote:
> > As Niall demonstrated so beautifully, there are no relational rules when
> > it comes to v$ views (actually the x$ "tables" upon which the views are
> > built).
>
>Yes, but then this statistics is meaningless. What does it mean to have
>1000 inserts without any consequence to the table?

You have to use the v$ views in the context they are meant for. This particular table is meant to help dbms_stats to decide whether the statistics are "stale" and should be refreshed. It is NOT an audit view. If you talk about useless views this is a truly useless view, even if it didn't suffer from the demonstrated shortcoming. What does it mean to have 1000 inserts, even if those inserts actually made into the table?? Statistics could be "stale" if much less than 10% of the content has "changed" and for other tables nothing changes materially in the statistics even if you exchange 100% for new data. I would never base statistics gathering on the content of this view.
PS. there are people who are very knowledgeable about Oracle who consider the data in many of the "cherished" v$ views as useless, meaningless, even misleading.

>I wouldn't generalize just like that. I don't find V$SESSION_LONGOPS very
>useful,
>period. It has never helped me with anything. It most certainly did not
>help me
>to see how much is left of any large job. While V$ tables in general are
>remarkably
>useful, this particular one is useless.

Obviously our mileages vary. I use it frequently to see the progress of full scans and even calculate the estimated finish time. I generally find it more precise than Microsoft's progress bar. And as I said. if the progress is not linear but slows down ( who cares if it speeds up :-) ) then that is a good piece of information as well to go and find the cause for the slowdown.

Regards

Wolfgang Breitling
Centrex Consulting Corporation
www.centrexcc.com

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