On 6/24/2005 12:13 PM, Christopher Browne wrote:

> Victoria Parsons wrote:
> 
>>Thanks for that info. Another question:
>>
>>If lots of database changes happen in "very quick" succession on the origin node, so that they are all replicated during the same SYNC, are they guaranteed to be replicated out in the true order they happened?
>>  
>>
> Yes, they surely are.
> 
> There is an "actionseq" column on sl_log_1 which is used to indicate the
> order in which changes are applied.

There could be race conditions where the after trigger doing the 
replication log collection is effectively fired in a different order 
than the row updates really happened with respect to concurrent 
transactions. But since concurrent transactions would block each other 
in the case of conflicting updates, the resulting replication action 
order is guaranteed to be in an agreeable order that leads to the same 
end result. And since all changes of one SYNC are applied in one 
transaction, nobody can see intermediate states, so nobody can actually 
tell that things got applied in a slightly different order.


Jan

-- 
#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me.                                  #
#================================================== JanWieck at Yahoo.com #