"David Rees" <drees76 at gmail.com> writes:
> On 8/8/07, Christopher Browne <cbbrowne at ca.afilias.info> wrote:
>> Most of this seems fine; fixes consistency between defaults and what is
>> documented.
>>
>> I don't agree with the explanation about the "race condition;" I don't
>> think that represents a race condition at all.  The timeout value is not
>> germane to the things described; if is set higher or lower, that will
>> not lead to data being replicated improperly.
>>
>> The reason for having sync_interval_timeout isn't to avoid a race
>> condition - it is to provide a way to say, with some confidence, "We
>> have replicated this origin up to Time T" even when there have been no
>> recent updates coming in from the application.
>>
>> I have reworded this, and committed it to 1.2 and HEAD...
>>
>> http://lists.slony.info/pipermail/slony1-commit/2007-August/001911.html
>
> Thanks, the updated wording is good. So just to be clear, there is no
> race condition with sequences as (they used to be) described and it
> would be entirely possible to disable sync_interval_timeout and never
> miss data?

There's not a "race condition," per se, with sequences; they
always simply get polled at the end of each SYNC.

There's a change in version 2 where sequence values are held in memory
by the slon so that the values only get updated if they have changed
lately.

You won't miss table updates, regardless.
-- 
select 'cbbrowne' || '@' || 'ca.afilias.info';
<http://dba2.int.libertyrms.com/>
Christopher Browne
(416) 673-4124 (land)