Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
Home -> Community -> Usenet -> c.d.o.server -> Re: table locking question
I think this was actually a row locking question despite the
subject line, but...
"Ryan" <rgaffuri_at_cox.net> wrote:
> I have seen cases where a small transaction will stay locked and PMON
> does not clear it up. Ive seen several people ask about this on lazyDBA
> as well. So its not unique.
No, it's not unique.
I was hoping there was a command like (cleaned up for mixed company): Alter system fish or cut bait in 60 seconds;
That would give any currently existing transaction 60 seconds to either commit or rollback before it was involuntarily terminated and rolled back.
> <Kenneth Koenraadt> wrote in message
> > When a user gets disconnected, the PMON will clear the server process,
> > rollback its uncommitted transactions and release its locks. As such
> > no problem, but if a large (uncommitted) transaction was taking place
> > at the time of the disconnect, the clean up and thus the release of
> > the locks may require some time.
...
> > That's correct. In fact, thats' the definition of a lock :-).
> > Be glad that PMON keeps the lock until it has recovered fully, else
> > your DB would soon be inconsistent.
This suggests to me that if a process encounters a row locked by a transaction being rolled back by pmon, that it will just sit and wait while pmon slogs through 10,000 blocks that no one cares about to get to this one. Is that the way it works? Or can the server process take the bull by the horns, so to speak, and clean out the offending block for itself.
Xho
-- -------------------- http://NewsReader.Com/ -------------------- Usenet Newsgroup Service New Rate! $9.95/Month 50GBReceived on Mon Feb 03 2003 - 18:07:29 CST