Re: v$undostat question
Date: Wed, 13 Feb 2008 08:57:29 -0800 (PST)
Message-ID: <442da920-7bca-4e0a-99ba-1c126373671c@e6g2000prf.googlegroups.com>
On Feb 13, 10:00 am, Chuck <skilover_nos..._at_bluebottle.com> wrote:
> Oracle 9.2.0.8
>
> How is it possible for maxquerylen to go from 1 to 9000 in a single 10
> minute interval? Is this a bug? Can't find anything about it on
> metalink. TIA
>
> SELECT begin_time,
> end_time,
> maxquerylen
> FROM v$undostat
> WHERE begin_time BETWEEN
> to_date('10-FEB-08 19:19:19', 'dd-mon-yy hh24:mi:ss')
> AND
> to_date('10-FEB-08 19:29:19', 'dd-mon-yy hh24:mi:ss')
> ORDER BY begin_time;
>
> BEGIN_TIME END_TIME MAXQUERYLEN
> ------------------ ------------------ -----------
> 10-FEB-08 19:19:19 10-FEB-08 19:29:19 1
> 10-FEB-08 19:29:19 10-FEB-08 19:39:19 9387
>
> 2 rows selected
Interesting. I have no clue and do not have the time to attempt to follow up the one idea I have. Namely if the database was just restarted about 19:39 on Feb 10 if the reported query run time could include the prior run time for a transaction that now was being rolled back after the restart or after a transaction failover.
By any chance was the database re-started around the time of the first entry?
Is TAF is use on this system? (If so, was there a failover)
I bet the answer is no to both but those were the only ideas I had. That would leave a 9.2.0.8 bug of some kind.
Just speculating -- Mark D Powell -- Received on Wed Feb 13 2008 - 10:57:29 CST