Last week, we at ETV implemented a postgreSQL backend for election
updates at Gujarat. There were twelve input clients, and three output clients
which generated realtime on air graphics. Clients were all Windows 2000
(anybody know of any good character generators ala CG2000 from matrox running
There were two Linux machines running postgresql, in an almost failsafe
environment. The secondary database would pull the most recent data every
minute via a cron job from the primary and update itself. In case the primary
db failed, it would shoot the primary (if it was still alive), and take over
by a simple ip change (I know I know, there are better ways). The clients
with uncommited transactions would rollback, and the data remains online with
a worst case scenario of 30 seconds downtime. When the first machine came
back online in such a scenario, it would begin functioning as the
What would make postgres production ready is realtime replication, and
guess that would be available in ver 7.3 (any ETA for this?).
Downsides : The developers were from the Windows world, and made
incorrect assumptions (ex, setup and teardown of tables is fast in MS
sql2000, and quite slow in postgres; and since they were unaware of this fact
they were creating temporary tables all over the place for complex queries).
So, the only problem was their unfamiliarity with the tools....
in a related note, is anybody aware of MS-Access like DB's for Linux
ease of use) ?
We will find a way, or we will make one - Hannibal