Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
Home -> Community -> Usenet -> c.d.o.server -> Re: HELP: Get ORA-1555 rollback too small
"Howard J. Rogers" <howardjr2000_at_yahoo.com.au> wrote in message news:<F3OG9.88862$g9.249416_at_newsfeeds.bigpond.com>...
> "Ban Spam" <ban-spam_at_operamail.com> wrote in message
> news:Xns92D8438ADF7DSunnySD_at_68.6.19.6...
> > "A. Dischner" <dischner_at_klch.med.uni-muenchen.de> wrote in news:dischner-
> > CE4085.15331002122002_at_hydra1.fw.med.uni-muenchen.de:
> >
> > > Hi,
> > >
> > > with Oracle 8.1.7.0.1 i get ORA-1555 which means Rollback-segment is
> > > too small.
> > >
> > > I have really big rollback segments: Tablespace 4gb, initial 4MB, next
> > > 4MB, optimal 16MB, max 4096, number of RBsegs: 4
> > >
> > > Tablespace usage is always < 1% as seen in Enterprise manager.
> > >
> > > There are no critical locks.
> > >
> > > any ideas? hints?
> > >
> > > Toni
> > >
> >
> > ORA-01555 can also mean that the code is issuing COMMITs
> > too frequently.
>
>
> Not really. If you had an infinitely large rollback segment, then you could
> commit every nanosecond and you still wouldn't loop round the segment and
> start over-writing the earlier rollback.
>
> Of course, given that rollback segments are of finite size, then yes:
> frequent commits give the segment the go-ahead to overwrite the committed
> rollback, and thus give the 1555 a chance to occur. And so it is definitely
> the case that inappropriate commits contribute to the problem. But it still
> comes down, fundamentally to size. After all: what will happen if we don't
> commit so frequently? Then the rollback segment won't be permitted to
> overwrite the earlier rollback, and the segment will begin to grow
> dynamically in response.
>
> Regards
> HJR
What if you have optimal size on a rollback, so potentially it can
grow big, but is not so big initially ? If you commit too often, how
will Oracle know that it should extend rollback ? It will happily
recylce it instead, because it it is allowed to do so by your commit.
Igor Izvekov.
P.S. E-Mail address is fake. Received on Tue Dec 03 2002 - 14:32:12 CST