Michael Dykman wrote on Fri, Sep 21, 2007 at 01:37:57PM -0400:
> There is a setting in your my.cnf which specifies the threshold at
> which temporary tables will be put to disk instead of being held in
> RAM. This has to be a dynamic decision as the system has to consider
> available RAM and the s
There is a setting in your my.cnf which specifies the threshold at
which temporary tables will be put to disk instead of being held in
RAM. This has to be a dynamic decision as the system has to consider
available RAM and the size of any given temporary table.. under
normal circumstances, the my.
Baron Schwartz wrote on Thu, Sep 20, 2007 at 07:46:44PM -0400:
> Michael explained it well, but just to say it a different way, the
> temporary table is created as an intermediate step in the table, and is
> implicit, not explicit. So it's not "sent" to the slave -- the query is
> sent to the s
Michael explained it well, but just to say it a different way, the
temporary table is created as an intermediate step in the table, and is
implicit, not explicit. So it's not "sent" to the slave -- the query is
sent to the slave, and if the query optimizer makes the same decisions
on the slave
Temporary tables only exist for the length of time that the connection
that created them remains connected and are only visible to that
connection. There is no reason to replicate these to a slave at all,
as no client connecting to that slave would ever be able to see them.
- michael dykman
On
