On 9/6/2016 2:10 PM, Ashish Chauhan wrote:
...Cannot afford downtime on db servers...
thats an impossible target in general unless you expect computer systems to run perfectly forever without ever needing hardware or software maintenance. there are strategies to /minimize/ downtime, such as having active/standby failover, but the simple act of failing over requires a short downtime where your applications are forcibly disconnected from the formerly active database and have to reconnect to the newly activated master, restarting any transactions that were in progress at the time of interruption. then there's the issues of building fully redundant network infrastructure, as well as redundant pooling. In my experience, the more pieces you add to a system in an attempt to make it more robust, the more fragile and delicate it becomes.
--
john r pierce, recycling bits in santa cruz