Posted by Matthias Rogel on 9. August 2011
looking forward to Tanel Poder’s webinar this evening.
Especially, the announcement to learn something about “ORA-8103s” called my attention.
The reason is, that I was hitting some bug or bugs in 11.2 (sporadically and not reproducible ORA-8103s) last November and working with support on this since then (issue has status “Development working” since some weeks now)
They used to occur several times per day in complex queries (selects in batch jobs).
Usually, the following workaround was helpful: wait for a minute, then issue the same select again.
So, we “patched” the batch jobs they occurred in, set parameter event to
’8103 trace name errorstack forever, level 10′
and uploaded tons (dozens of GBs) of tracefiles to support.
In the meantime, we applied some patches concerning other issues and monitored the number of ORA-08103′s per day via
with days as
select date'2010-11-19' + level as day
connect by date'2010-11-19' + level <= sysdate
count(distinct x.message_text) as count_08103s_per_day
from x$dbgalertext x, days
where x.message_text(+) like '%ORA-08103%'
and trunc(x.originating_timestamp(+)) = days.day
group by days.day
order by days.day
After ugrading to 220.127.116.11.2 via Patch 11724916, this number went strongly down, but we still had issues from time to time with a false ORA-08103.
Today I noticed that since applying Patch 9285259 we did hit no ORA-08103 anymore !
Coincidence ? I don’t know
Update August 28th:
I was wrong, false ORA-08103s have not disappeared yet, we encountered some of them last week, re-activated the SR (which is open unsolved now since last November …).
The reason that I thought it had disappeared was that I didn’t see it mentioned in the alert.log anymore, however, since I had unset event, it was not logged there.