Hello Matt,
On 2015-04-28 14:04, Matt . wrote:
It might be quite "simple said" but can't we make 2 hosts sharing the
same DB(cluster) and NFS share for storing messages ?
yes, it's possible, see comments below.
We might have to figure out how to manage sphinx, but that is moslty
what I try to do with HA setups, take out the most important data and
make the software a "dumbnode".
Would that be an idea to start with ?
it may work out by using processing-only nodes, and storing actual data
(stored emails, sql, and sphinx data) on a nuke blast repellent host.
In such a scenario you don't really need 2 piler nodes (provided that
you
have a backup of piler.* files). In fact using a shared database works
only
if an email goes to either node1 or node2.
Anyway I encourage you to test several scenarios before going into
production.
Janos
PS: please don't hit reply-to-all, it's enough to get messages in 1
instance,
so reply to the list only.
