Hit this problem while setting up / testing 4.6 HA: in a setup with 2
nodes you can select Distributed Database on both nodes (that means
that each member will have its copy of data). In this case a mongo
arbiter is needed to be used in primary elections in case of failure.
By definition a mongo arbiter cannot contain data and it will never
become the primary node
(http://www.mongodb.org/display/DOCS/Adding+an+Arbiter). Now, the
issue that I see (and signaled by QA as well) is when you setup
something like:
- node 1 set as Distributed Database & Database Arbiter
- node 2 set as Distributed Database
then after restarting node 1 it will become an arbiter (that is without data)
So even UI allows you to select this config and shows that you have 2
Distributed Databases enabled (one of them being arbiter too) the
reality is that you only have one Distributed Database (with copy of
data) - node 2, and one arbiter - node 1, which is not what is
expected IMO.
Therefore I suggest to change UI and to make Distributed Database
Arbiter and Distributed Database to exclude each other (that is when
one selected the other one to automatically get unselected) and admin
to be warned that in case of 2 nodes with 2 Distributed Databases the
3rd node to act as arbiter is needed.

Thoughts?

Thanks
George
_______________________________________________
sipx-dev mailing list
[email protected]
List Archive: http://list.sipfoundry.org/archive/sipx-dev/

Reply via email to