Re: Lock Fairness

From: Robert Klemme <shortcutter_at_googlemail.com>
Date: Thu, 29 Nov 2007 23:39:49 +0100
Message-ID: <5r8tdmF13g1geU2@mid.individual.net>

On 29.11.2007 19:50, fitzjarrell_at_cox.net wrote:
> On Nov 29, 11:58 am, Robert Klemme <shortcut..._at_googlemail.com> wrote:

>> On 29.11.2007 17:49, fitzjarr..._at_cox.net wrote:
>>
>>> On Nov 29, 10:46 am, "shortcut..._at_googlemail.com"
>>> <shortcut..._at_googlemail.com> wrote:
>>>> while investigating an ORA-02049 (time-out: distributed transaction
>>>> waiting for lock) I was asked which of the TX waiting for a lock will
>>>> obtain the lock after it is released.  I tried Concepts Guide,
>>>> Administrator Guide, searched Oracle docs and Google but could not
>>>> find anything meaningful on Oracle's lock fairness.  I would assume
>>>> that it's simply unspecified.  Does anyone have more specific
>>>> information about this?  In practice it's certainly safe to assume
>>>> nothing about the order TX's are served but I am curious... :-)
>>> Usually the first one in the queue.
>> Thank you, David!  Sounds as if you knew something about exceptional
>> cases where it is not the first one. :-)  I could only speculate here
>> (e.g. a TX waiting for more taken locks at the same time could be left
>> waiting if another one can continue immediately).  Do you have a pointer
>> to further material on the matter?
>>
>> Kind regards
>>
>>         robert

>
> The 'first' one in the queue may no longer exist due to resource
> profile settings/software aborts/etc. Thus the second transaction
> would become the first, which, of course, then makes it the
> transaction obtaining the lock once the existing lock is released.
> This could go on until no transactions are left in the queue (not
> likely, but remotely possible).
>
> The 'usually' refers to the queue at time X.

Ah, ok. So there is no out of order serving under special circumstances.

Thanks again!

        robert

PS: You still did not disclose your sources. :-) Received on Thu Nov 29 2007 - 16:39:49 CST