Actually clustering works because multiple repositories *are* writing
to the same database. When one repository writes into it, the others
are notified via the journal, and pick up the changes from the
database and index them in their local indexes.
Alessandro
On Aug 2, 2007, at 9:18 AM, Nick Stolwijk wrote:
I have considered clustering, but I thought it would go horrible
wrong. If multiple repositories in a cluster starts writing to the
same database, I guess bad things happen.
To clearify, multiple applications starts their own repository with
the same database behind it. If you write to one of the
repositories, it gets written into the database. If you cluster the
repositories, the other repositories also wants to write to the
same database.
I could be wrong, but I guess that would go wrong. :)
With regards,
Nick Stolwijk
Stefan Guggisberg wrote:
On 8/2/07, Nick Stolwijk <[EMAIL PROTECTED]> wrote:
I've got a problem with multiple repositories that are reading
from one
database. This was going well, because there was no writing to the
repositories (so read only data). Now, one of the repositories is
going
to write to the database. Because of the ItemStateManager inside
Jackrabbit, the other repositories aren't picking this up. Is
there some
way, to get to the WorkspaceInfo or ItemStateManager instance to
clear
it once in a while?
With regards,
Nick Stolwijk
Ps. I know this is a not supported way of doing it. But for now, we
can't get the customer to run a stand alone repository, that all
applications can connect to.
have you considered using jackrabbit's clustering feature?
cheers
stefan
