[ic] Request for comments and documentation - multiple server architecture

Greg Cope greg@rubberplant.freeserve.co.uk
Thu, 21 Dec 2000 11:38:48 +0000

Dan wrote:
> > Database clustering done by http://www.erserver.com (which wont be open
> > source for another few weeks).
> BZZZT.  Correction to self:  PostgreSQL has just CVS commit'ed their
> database replication into the source tree this morning.  Wow.  I love how
> time flies in open source worlds.
> By the way, this discussion might open another can of good-stuff for
> Interchange:
> Application-level database clustering.  Where it becomes easy to specify
> multiple DSN's for a given IC "Database" depending on if it is a write
> action or a read action.  (i.e. all INSERTS use
> masterdatabaseserver.domain.com/, and all SELECT's use
> slave_readonly_databaseserver.domain.com/
> Would anyone like to comment on the feasability of adding that feature?
> (easy,hard,impossible?)

<disclaimer I have nothing to do with Akopia, just playing with MySQL
distributed repklication thing ...>

Well that depends on whom is doing it ;-)

Looking throught the code Vend::DBI does do some checking of where the
statement is a SELECT - so in thoery there is no reason why not as
within these routines that prepare the statments (i.e. define the $sth
from a different dbhandle depending on wherethere replication is being

But what logic would be used i.e if its a select do we use a slave, and
if thats down give up or try another ?  Same goes for the Master - does
it have a backup master ?


Could something like this be put on the Akopia / Interchange "wish list"


> Dan Browning - danb@cyclonecomputers.com
> _______________________________________________
> Interchange-users mailing list
> Interchange-users@lists.akopia.com
> http://lists.akopia.com/mailman/listinfo/interchange-users