On 8/21/08, Arjen Lentz <[EMAIL PROTECTED]> wrote: > > Hi Jeremy > > On 22/08/2008, at 9:50 AM, Jeremy Zawodny wrote: > >> On Thu, Aug 21, 2008 at 4:33 PM, Arjen Lentz <[EMAIL PROTECTED]> >> wrote: >> On 22/08/2008, at 1:59 AM, Jeremy Zawodny wrote: >> Does it make sense to explicitly record the database(s) and/or table(s) >> involved in a statement. I don't know how common it is to >> filter replication based on one or both of those, but I tend to encounter >> it WAY more than I'd expect to... >> >> >> It'd be great if a slave could give a master a ruleset for this, so it >> doesn't have to drag across everything before filtering. >> So, I like the idea of explicitly recording the dbs/tables involved in a >> query (and not just the current default db), but more for the benefit of the >> master being able to filter before sending it to an individual slave, rather >> than having all the extra info go across the wire too. >> >> That'd be useful. Or per-db binlogs, which is a different can of worms... >> > > > Ye I don't think that flies. > Multi-table updates and deletes already make filtering difficult, but > storing per db would complicate that even further.
I'd buy that. Isn't this one of the problems Falcon is suffering from with respect to > replication, making it necessary to go only row-based repl? Beats me. But if it's big and hairy, we can drop it. I just wanted to make sure it is at least discussed. Jeremy
_______________________________________________ Mailing list: https://launchpad.net/~drizzle-discuss Post to : [email protected] Unsubscribe : https://launchpad.net/~drizzle-discuss More help : https://help.launchpad.net/ListHelp
