[one-users] distributed frontend with raft consensus algorithm

[Sharuzzaman Ahmat Raslan]

Hi all,

My last interaction with Opennebula is with the version 3.8

In that version, the frontend can only be running by a single node. To have
a redundancy, it was advised to have the frontend to be configured using HA
tools, such as Red Hat cluster suite. See
http://archives.opennebula.org/documentation:archives:rel4.2:oneha

I believe that kind of setup is not optimal, as you have to bolt in another
subsystem, that could potentially become another head ache to maintain.

As alternative to external clustering system, I would like to highlight to
the developers the availability of Raft Consensus Algorithm. See
http://raftconsensus.github.io/

This algorithm can help to maintain the availability of the frontend, in a
fault-tolerant distributed system. With this algorithm, frontend can be run
on the VM nodes itself, where at any given time, one of the nodes will act
as a frontend for the whole cluster. If that "leader" node goes down, any
other node will step up to become the next leader, and maintaining the
availability of the frontend

This can help simplify deployment, and reduce the requirement to have a
dedicated frontend machine, with another failover machine in standby

I hope this idea could be discussed futher, and implemented on the future
release of Opennebula.

thanks.

-- 
Sharuzzaman Ahmat Raslan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opennebula.org/pipermail/users-opennebula.org/attachments/20140206/acd41127/attachment.htm>