ORA-1555 snapshot too old: why?

Chuck wrote:
> Oracle 9.2.0.7
>
> I got an ORA-1555 snapshot too old today on a export where consistent=y
> was set. The export failed at 6 hrs even though undo_retention was set
> to 8 hours. Querying v$undostat shows that during the time when the
> export was running a total of only 150m of undo was written. The undo
> tablespace is 4g. Why would I get an ora-1555 when undo_retention is set
> plenty high enough, and the undoblks written weren't anywhere near
> maxing out the undo tablespace?
>
> TIA

Tom Kyte has this all covered. Try looking for 1555 in
http://asktom.oracle.com

Chuck wrote:
> Oracle 9.2.0.7
>
> I got an ORA-1555 snapshot too old today on a export where consistent=y
> was set. The export failed at 6 hrs even though undo_retention was set
> to 8 hours. Querying v$undostat shows that during the time when the
> export was running a total of only 150m of undo was written. The undo
> tablespace is 4g. Why would I get an ora-1555 when undo_retention is set
> plenty high enough, and the undoblks written weren't anywhere near
> maxing out the undo tablespace?
>
> TIA

Tom Kyte has this all covered. Try looking for 1555 in
http://asktom.oracle.com

hpuxrac wrote:
> Chuck wrote:
>> Oracle 9.2.0.7
>>
>> I got an ORA-1555 snapshot too old today on a export where consistent=y
>> was set. The export failed at 6 hrs even though undo_retention was set
>> to 8 hours. Querying v$undostat shows that during the time when the
>> export was running a total of only 150m of undo was written. The undo
>> tablespace is 4g. Why would I get an ora-1555 when undo_retention is set
>> plenty high enough, and the undoblks written weren't anywhere near
>> maxing out the undo tablespace?
>>
>> TIA
>
> Tom Kyte has this all covered. Try looking for 1555 in
> http://asktom.oracle.com
>

I've already been there and searched for 1555. I didn't find anything
related to ora-1555 that matches this case. All the responses related to
manual rbs management, cases where undo_retention was insufficient, or
cases where the undo tablespace size was insufficient. My case does not
fit any of these.

hpuxrac wrote:
> Chuck wrote:
>> Oracle 9.2.0.7
>>
>> I got an ORA-1555 snapshot too old today on a export where consistent=y
>> was set. The export failed at 6 hrs even though undo_retention was set
>> to 8 hours. Querying v$undostat shows that during the time when the
>> export was running a total of only 150m of undo was written. The undo
>> tablespace is 4g. Why would I get an ora-1555 when undo_retention is set
>> plenty high enough, and the undoblks written weren't anywhere near
>> maxing out the undo tablespace?
>>
>> TIA
>
> Tom Kyte has this all covered. Try looking for 1555 in
> http://asktom.oracle.com
>

I've already been there and searched for 1555. I didn't find anything
related to ora-1555 that matches this case. All the responses related to
manual rbs management, cases where undo_retention was insufficient, or
cases where the undo tablespace size was insufficient. My case does not
fit any of these.

Chuck wrote:
> hpuxrac wrote:
>> Chuck wrote:
>>> Oracle 9.2.0.7
>>>
>>> I got an ORA-1555 snapshot too old today on a export where consistent=y
>>> was set. The export failed at 6 hrs even though undo_retention was set
>>> to 8 hours. Querying v$undostat shows that during the time when the
>>> export was running a total of only 150m of undo was written. The undo
>>> tablespace is 4g. Why would I get an ora-1555 when undo_retention is set
>>> plenty high enough, and the undoblks written weren't anywhere near
>>> maxing out the undo tablespace?
>>>
>>> TIA
>> Tom Kyte has this all covered. Try looking for 1555 in
>> http://asktom.oracle.com
>>
>
> I've already been there and searched for 1555. I didn't find anything
> related to ora-1555 that matches this case. All the responses related to
> manual rbs management, cases where undo_retention was insufficient, or
> cases where the undo tablespace size was insufficient. My case does not
> fit any of these.

I agree. I followed HPUXRAC's links too thinking there was something I
had missed and found nothing like what you are reporting.

I couldn't find an example on metalink so I'd suggest opening an SR.
I'll be interested in what you discover.
--
Daniel A. Morgan
University of Washington
damorgan-at-x.washington.edu
(replace x with u to respond)
Puget Sound Oracle Users Group
www.psoug.org

Chuck wrote:
> hpuxrac wrote:
>> Chuck wrote:
>>> Oracle 9.2.0.7
>>>
>>> I got an ORA-1555 snapshot too old today on a export where consistent=y
>>> was set. The export failed at 6 hrs even though undo_retention was set
>>> to 8 hours. Querying v$undostat shows that during the time when the
>>> export was running a total of only 150m of undo was written. The undo
>>> tablespace is 4g. Why would I get an ora-1555 when undo_retention is set
>>> plenty high enough, and the undoblks written weren't anywhere near
>>> maxing out the undo tablespace?
>>>
>>> TIA
>> Tom Kyte has this all covered. Try looking for 1555 in
>> http://asktom.oracle.com
>>
>
> I've already been there and searched for 1555. I didn't find anything
> related to ora-1555 that matches this case. All the responses related to
> manual rbs management, cases where undo_retention was insufficient, or
> cases where the undo tablespace size was insufficient. My case does not
> fit any of these.

I agree. I followed HPUXRAC's links too thinking there was something I
had missed and found nothing like what you are reporting.

I couldn't find an example on metalink so I'd suggest opening an SR.
I'll be interested in what you discover.
--
Daniel A. Morgan
University of Washington
damorgan-at-x.washington.edu
(replace x with u to respond)
Puget Sound Oracle Users Group
www.psoug.org

Chuck wrote:
> hpuxrac wrote:
> > Chuck wrote:
> >> Oracle 9.2.0.7
> >>
> >> I got an ORA-1555 snapshot too old today on a export where consistent=y
> >> was set. The export failed at 6 hrs even though undo_retention was set
> >> to 8 hours. Querying v$undostat shows that during the time when the
> >> export was running a total of only 150m of undo was written. The undo
> >> tablespace is 4g. Why would I get an ora-1555 when undo_retention is set
> >> plenty high enough, and the undoblks written weren't anywhere near
> >> maxing out the undo tablespace?
> >>
> >> TIA
> >
> > Tom Kyte has this all covered. Try looking for 1555 in
> > http://asktom.oracle.com
> >
>
> I've already been there and searched for 1555. I didn't find anything
> related to ora-1555 that matches this case. All the responses related to
> manual rbs management, cases where undo_retention was insufficient, or
> cases where the undo tablespace size was insufficient. My case does not
> fit any of these.

Any chance you are committing inside a loop?

If not and you can reproduce it at will you can submit a service
request to oracle. I bet they suggest putting 9208 in ( the usual
response right ? ).

Chuck wrote:
> hpuxrac wrote:
> > Chuck wrote:
> >> Oracle 9.2.0.7
> >>
> >> I got an ORA-1555 snapshot too old today on a export where consistent=y
> >> was set. The export failed at 6 hrs even though undo_retention was set
> >> to 8 hours. Querying v$undostat shows that during the time when the
> >> export was running a total of only 150m of undo was written. The undo
> >> tablespace is 4g. Why would I get an ora-1555 when undo_retention is set
> >> plenty high enough, and the undoblks written weren't anywhere near
> >> maxing out the undo tablespace?
> >>
> >> TIA
> >
> > Tom Kyte has this all covered. Try looking for 1555 in
> > http://asktom.oracle.com
> >
>
> I've already been there and searched for 1555. I didn't find anything
> related to ora-1555 that matches this case. All the responses related to
> manual rbs management, cases where undo_retention was insufficient, or
> cases where the undo tablespace size was insufficient. My case does not
> fit any of these.

Any chance you are committing inside a loop?

If not and you can reproduce it at will you can submit a service
request to oracle. I bet they suggest putting 9208 in ( the usual
response right ? ).

By any chance are you using lobs and is that being exported? LOB update
does not use undo instead uses PCTVERSION. I have had ORA-1555 snapshot
too old errors during export and we had a process that was updating the
LOBS.

By any chance are you using lobs and is that being exported? LOB update
does not use undo instead uses PCTVERSION. I have had ORA-1555 snapshot
too old errors during export and we had a process that was updating the
LOBS.