Hi!
Using async replication between geographically database servers. I've
some 9.3 and some 9.4 servers, so doesn't matter the version (the only
reason I've not migrated all to 9.4 yet is because I'm short on time).
I've experienced some synchronization failures between Master-Slave
servers (for example, if connection is lost for 4 to 6 hours, the number
of wall keep segments may not suffice).
I've created some bash scripts that will reacreate the master-slave
using rsync, which seems to be farily easy - but really slow.
Depending on server size (130Gb for example) take really long on a 5Mbps
link - in some cases, all night long.
This is the first approach: all databases in one server, and if I loose
the sync between master and slave, all data need to be transfered
between server and slave and is really slow.
The other approach I've tested is to split each database in one cluster.
By using the second approach, I can resynch each server separeted (for
example 15Gb instead 130Gb).
This approach have the problema that it is needed to configure and
maintain another cluster for each database.
Is there a best practice for multi-database systems with async
replication to avoid such "resync" problems?
Thanks for your orientation,
--
Atenciosamente,
Edson Carlos Ericksson Richter