ora-1555 not in alert log?

Joel Garry

unread,

Oct 1, 2002, 5:43:08 PM10/1/02

to

I have a client/server odbc application (packaged from a vendor, no
source) which returned an 'ORA-01555: snapshot too old: rollback
segment number 7 with name "RBS08" too small' error. However, that
error is not in the alert log? I see log switches subsequent to that
time. 8.1.7 on hp/ux, a well-tuned production instance that has no
errors besides expected ones from the standby db being down. SMON
left a trace file indicating it cleaned up after problem, taking 70
minutes to do so.

I can understand the ora 1555, as some unusual processing is taking
place as I try do some ETL while others are doing big processing, but
what concerns me is why the error is not in the alert log. Yes, I've
triple-checked the dates and instances, it happened today on the
instance whose alert log I looked at.

jg
--
@home is bogus, substitute @cox.net for email, though I'm a bit slow on
that these days.
"With Oracle9i you'll never see ORA-01555 again!" - New Features presenter

David Sharples

unread,

Oct 1, 2002, 6:38:07 PM10/1/02

to

cos its not a database error, its an 'application' error

"Joel Garry" <joel-...@home.com> wrote in message
news:91884734.02100...@posting.google.com...

Ban Spam

unread,

Oct 1, 2002, 8:50:15 PM10/1/02

to

joel-...@home.com (Joel Garry) wrote in
news:91884734.02100...@posting.google.com:

Yes, you are correct.
Most likely it is because the code is doing "multiple" commits;
such as COMMIT after every 10,000 rows have been inserted;
instead of a single COMMIT at the end.

Richard Foote

unread,

Oct 1, 2002, 8:55:08 PM10/1/02

to

Hi Joel,

As David has said, ORA-1555s are not recorded in the alert log.

One nice thing with 9i is that such errors are now "logged" (or counted) in
the v$undostat view. Which helps with the sizing of the undo tablespace and
the undo_retention parameter.

Cheers

Richard
"David Sharples" <david.s...@ntlworld.com> wrote in message
news:5upm9.193$RI6....@newsfep1-win.server.ntli.net...

Connor McDonald

unread,

Oct 2, 2002, 2:55:47 PM10/2/02

to Joel Garry

You could presumably set an event to dump a trace when it happens

hth
connor
--
==============================
Connor McDonald

http://www.oracledba.co.uk

"Some days you're the pigeon, some days you're the statue..."

Joel Garry

unread,

Oct 2, 2002, 5:10:04 PM10/2/02

to

"David Sharples" <david.s...@ntlworld.com> wrote in message news:<5upm9.193$RI6....@newsfep1-win.server.ntli.net>...

> cos its not a database error, its an 'application' error

[slaps forehead] Man, I totally forgot! I'm so used to my systems not
having this problem. In the past when I've walked into places that have
this problem, they usually had a bunch of other stuff in the alert logs
they didn't know enough to look in, so I must have mentally transferred
1555 there.

I guess this means the application should be responsible for tuning
RBS's! :-)

Turns out the error they didn't tell me about was ORA-04030, which makes
a lot more sense. Let's hear how wonderful tiered applications are
when they give a different error to the user than to the middle tier
log. So now I'm off to check the swap, and look for something in the
patch set about memory leaks so I can convince the boss we really should
be latest patched (he's against patching just because O puts out a set).

jg
--
@home is bogus
"I'm Troy McClure, you may have seen me in the driver training film 'Alice
Through the Windshield Glass'..." - Simpsons, Brother From the Same Planet