[postgis-users] options to scaleout POSTGIS DATABASE

[Greg Williamson]

Pierluigi --


>> 
>> If you can provide some more information about your requirements that might suggest a cleared focus on one tool or another.
>> 
>> HTH,
>> 
>> Greg Williamson
>> 
>
>
>The ideal scenario is 2 identical servers sitting behind 1 ip adress. The user uses this ip as POSTGIS server and couldn't even know if its geodatabase service is server by 1 postgis or two or three (i used to work for a non gis app with oracle RAC and i thought it would be possible to do something similar with postgres+postgis).
>


Not knowing RAC except by reputation, I can't really address this.

>
>I thought that pgpool II could do this for me, but I'm not aware how this could work with POSTGIS.
>I can also use a master server replicated (with postgres streaming replication) on a second server which is used only for querying but i want to know what happens if master dies. Does the replica become a new master trasparently for the clients or do they have to connect to a different ip?
>


There is no automatic switchover as far as I know in the streaming replication -- you have to have identify that there is a failure and intervene; the switch of a server from standby mode to active is quite fast in and of itself. Making outside systems award of the new server is trickier -- perhaps having a VIP which is the connection point to the outside world, then swap what the VIP points to when you failover to the backup server.

>
>The GIS data that i except to manage are 10gb in several tables. I also expect that the user wants to creare in the POSTGIS server new geographic tables.
>

This sounds like the built-in replication may be a better fit for you -- adding new tables to a Londiste replication requires some extra steps, although you could probably write a procedure that would be used to create the new tables in postgres and add those (and their sequences, if any) to Londiste.

Greg W.