Lars Marowsky-Bree wrote:
On 2006-06-27T15:46:25, Huang Zhen <[EMAIL PROTECTED]> wrote:
Here is the design issues of the quorum server. Please review it.
http://wiki.linux-ha.org/QuorumServerDesign
1) I'd not call that "subcluster" when you mean "site". A subcluster
means something different in the regular terminology.
2) How do you protect against an outage of the quorum server?
The quorum server typically is used as a tie-breaker. So, its
availability is only critical when the previous quorum modules (like
majority voting) declare a tie. It is not a _single_ point of failure,
since in every case where it would affect availability, another failure
has already occurred.
One can use standard techniques (like clustering with STONITH) to ensure
the availability of the quorum server. This would be especially
valuable if the quorum server provides services to multiple clusters.
If one were to stretch one's brain even further (in a way that might not
be profitable), one could imagine a large multi-national company with
hundreds of sites creating a master 5-site quorum server cluster, with
normal straight voting to control the master quorum server.
Since there is no shared storage, setting up machines to do this would
be relatively inexpensive. It might be _crazy_ ;-), but for enterprises
of that size, it would be affordable enough.
[[To go even farther into crazy land... -- one might even imagine a
master/slave quorum server service with state replication. I'm really
not sure if this matters, and we're _certainly_ not planning on
implementing it in the foreseeable future, but it's a thought]].
--
Alan Robertson <[EMAIL PROTECTED]>
"Openness is the foundation and preservative of friendship... Let me
claim from you at all times your undisguised opinions." - William
Wilberforce
_______________________________________________________
Linux-HA-Dev: [email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev
Home Page: http://linux-ha.org/
