Re: Troubleshooting ORA-01555 while using AUM

1. What's your max value of undoblk and 'maxquerylen' ?
2. The usual approach with manually managed undo is to have equi-sized segments large enough not to wrap around during the longest running query.
3. Perhaps the AUM is just not good enough for your environment since it uses segment of different sizes and the probability of the 1555 failure is directly proportional to the smallest segment size. You might be better off using the old-fashined rollback TS with equally sized segments.

Rgds.

VC

"Tech Geek" <Tech_Geek_at_Gawab-IHateSpam.com> wrote in message news:Uly0c.6952$rW6.3654_at_nwrddc03.gnilink.net...
>
> "VC" <boston103_at_hotmail.com> wrote in message
> news:3ux0c.151075$jk2.589509_at_attbi_s53...
> > Pls. see below:
> >
> > "Tech Geek" <Tech_Geek_at_Gawab-IHateSpam.com> wrote in message
> > news:6Qw0c.19711$6K.4704_at_nwrddc02.gnilink.net...
> > >
> > > --- My bad. Please read as zeros, instead of non-zeros.
> > >
> > >
> > > > Could you supply more information about under exactly what
> > circumstances
> > > > the 1555 is happening in your case ?
> > >
> > > This database is being used by end-users for short transactions 24x7
> and
> > > some adhoc batch jobs, which update/insert new records and also does
> some
> > > SELECTS. Mostly, the batch jobs and the daily export job run at time
of
> > the
> > > day, ends up with Ora-1555.
> >
> > I prefer to use StatsPack to figure out how large the undo TS should be
> but
> > let's try and use the V$UNDOSTAT for the same purpose.
> > Let's forget about those two columns (UNXPSTEALCNT/UNXPSTEALCNT) and
take
> > look at 'undoblk' and 'maxquerylen'.
> >
> > Assuming the max value of undoblk is 10000 8K blocks in 10 minutes (0.13
> > MB/sec) and the max value in 'maxquerylen' is 7200 sec, then _each_ of
> your
> > undo segments should at least be 940 MB in order to avoid the 1555
error.
> > For ten segments, it'll amount to 9.4 GB.
> >
> > VC
> >
>
> Don't know, if it helps, I have about 13G. The export completes or fails
> almost after 40 minutes. It's still puzzles me.
>
> Thanks
> TG
>
> >
>
>
Received on Sun Feb 29 2004 - 22:12:14 CST