|
On 24 Feb 2000, Doug Bagley wrote:
> To me, their approach is very appealing. It looks like each node in
> the cluster runs a cluster management daemon, which allows you to
> easily synchronise the cluster configuration, and it also provides HA
> by electing a new master Director if the master dies.
This approach has been mentioned in this group by both Horms and Julian
and is mentioned without any deatils in the HOWTO from Dec 99 (grep for
"failover" and in section "10.2.3 LVS for failover of director"). I don't
know where the idea first came from (it sounds like a standard part of
distributed computing from what you say) and I didnt know it was in the
Turbo product. I like the idea of being able to promote any realserver to
a director rather than preassigning 2 machines to be the failover pair.
An approach to automatic director promotion using gated infront of the LVS
has been implemented (someone in England I believe), but I didn't get
enough details to understand how it works. A reply from Lars indicated
that he understood what was going on.
No-one is working putting any of this into LVS but I think we should have
it.
Joe
--
Joseph Mack mack@xxxxxxxxxxx
|
