Bhartendu,
In my humble opinion, you would be well served if you listened to all
the nice people on this list.
Use a local disk subsystems with RAID-type storage, use replication to
have a second "standby" system available if the first one fails.
The path you seem anxious to trod will get very muddy and slippery.
Good Luck.
...john
Bhartendu Maheshwari wrote:
>Dear UC,
>
>You are right for the HA solution but at the same time we are also
>implementing the load balancing solution so we can't have for one Node 2
>different processing entity and database as well. We try to provide
>solution for HA, load balancing both and in that there are 2 different
>processing machine but sharing the common database so that both get the
>latest and synchronized data files.
>
>You are right if the NAS is down then everything get down but the
>probability for the NAS is down is very less and by this we are able to
>provide service for 99% cases and if you are 99% handle cases then you
>are providing good service, isn't?
>
>About the cache to file write :- If the database is writting all the
>stuff to the files after each transaction then both have one
>synchronized set of data file whoever want can acquire the lock and use
>and then unlock it. The MySQL have command "flush tables" to enforce the
>database to write all the cache contents to the files, Is there anything
>similar in postgres? This will definitely degrade the performance of my
>system but its much more fast since I have 2 processing unit.
>
>Anyway if somebody have some other solution for the same please help me.
>One I got have one common postmaster running on one PC and the two nodes
>connect to that server to get the data. Any other please let me know.
>
>regards
>bhartendu
>
>
>