On 19 May 2011 09:05, John R Pierce
<pierce@hogranch.com> wrote:
On 05/18/11 11:39 PM, Szymon Guz wrote:
thanks for the answer. It is not a problem to have 3 oracle instances, in fact there will be hundreds of them probably, but could also be hundreds of Postgres instances :)
to share block storage between cluster nodes requires a cluster-aware file system, as conventional file systems do not expect the disk files to change behind their back. these cluster-aware file systems have a bunch of overhead in maintaining cache coherency between nodes, and such.
as far as I know, postgres can not be run on a read only file system, further its shared_buffer caches couldn't possibly be aware of other nodes modifying blocks that happen to be cached.
Well,
I have no idea how expensive it is, but fortunately that's not my problem :) I was only asked about the possibility of running Postgres that way.
And by the way: how expensive is that?
regards
Szymon