2009/7/23 Net Tree Inc. <nettreeinc@gmail.com>:
> Hello,
> I am having a concept question about how the Read/Write process transaction
> works within the DB network structure.
> Assume I have an environment of one master DB with three slaves running
> Slony-I replication server.
> As you all know Slony is a Single master to multiple slaves replication
> system supporting cascading and slave promotion. Master DB can do both read
> and write, but the slaves can only do read and for reporting purpose.
>
> Here is the question of concept that I don't understand, if you have answer
> or provide me a link to referral is appreciated.
>
> Q: For application Developer write codes to accessing the DB server to
> getting/modify datas. Since there is only one server that can handle both
> Read/Write which is the Master and slave is only to retrive data.
> How should the developer know which DB server should they connecting
> to for either to read or to write? If developer is using the same fixed
> connection strings to a fix IP to Master server for Read/Write and another
> connection string to a fix IP to (for example) my SLAVE1 server to retrive
> datas.
> How and what would happen if my SLAVE1 failed or due to maintaince that I
> need to replace it?
>
> Also what would happen for Master DB if issues such as failover or switch
> over that I need to replacing the master server due to failure or
> for maintaince, or to promote one of the slave to master?
>
> I just need answer that tells me how PostgreSQL and Slony would react if
> such issues happens. Answer can simply just to provide me a web link for me
> to read further. Thanks!
In addition to what Christopher said, you might want to look at
http://linux-ha.net/
--
Michael Wood <esiotrot@gmail.com>