Re: PostgreSQL in Shared Disk Failover mode on FreeBSD+CARP+RAIDZ

Στις Tuesday 21 December 2010 03:23:25 ο/η snoop(at)email(dot)it έγραψε:
> Hi everybody,
> I'm trying to figure out a way to setup a PostgreSQL HA cluster solution.
>
> I've done some research and I've an idea of different possible solutions,
> but I'd honestly like to implement it using CARP in a "Shared Disk Failover"
> fashion.

This reminds me of the serialization stack on shared disks (DASD) in the MVS IBM oper system.
It takes a lot of work to do that on OS level. Its smth beyond multiprocessing and high availability.
Think about it. In FreeBSD (or any of-the-shelf Unix), there is no inherent way to implicitly lock files.
Even on the same machine, if two users/processes modify the same file, the one who saves (closes)
the file last, generally wins. In order to do that, you will have to use explicit locking done by the application.
The OS by itself does not do that by default.
Things get tougher in a networked environment since file serialization should be applied at network-level.
This combination of data sharing (the idea of decoupling the concept of redundant hardware
from the concept of redundant disks) with the characteristics of parallel computing was realized in
IBM's Parallel Sysplex technology: http://en.wikipedia.org/wiki/IBM_Parallel_Sysplex
On top of it, someone could have CICS, DB2,etc... almost any available application.
It would be interesting to know if there is some concept close to it in the open source-Unix world.

--
Achilleas Mantzios