On 11/08/2010 03:49 PM, Johan De Meersman wrote:
> On Mon, Nov 8, 2010 at 8:47 AM, Mats Kindahl <mats.kindahl@stripped
> <mailto:mats.kindahl@stripped>> wrote:
> Also, I wonder why you want to lock the other table?
> The obvious answer to that would be multimaster writes, I guess.
But then you can't use a (traditional) dual-master setup because that
will block the changes on one master from getting to the table on the
other master. Hence the reason for me asking.
It is always possible to have the application send the changes to both
the master, but I wonder if this is what the original poster intended.
> And where that is concerned, it is worth noting that the latest
> versions of 5.1 actually have two-phase commit, locally known as
> (semi-) synchronous replication.
Are we talking about the same thing? Do you mean 2PC with a slave as a
cohort? Semi-synchronous replication is not the same as a 2PC in this
sense since the transaction is committed to the master before sending
the transaction to the slave.
Also, this is distributed with 5.5, but not with 5.1.
Just my few cents,