|
Hi Serge
On Thu, 2008-12-11 at 17:34 +0100, Serge Fonville wrote:
> Based on everything I read about LVS and especially ldirectord it seems to
> me that the load balancer would become the SPOF.
Just to clarify and expand a bit on what Joe and Malcolm have said, I
need to step back a bit from the detail of your setup.
There are three parts to a High Availability setup:
1. The load balancing component (in this case, always LVS)
2. The health check component
3. The failover component
There are a number of ways of providing components (2) and (3), and the
most common ones give you a configuration-based method of managing
component (1).
As you mention, you're using heartbeat and ldirectord.
ldirectord provides both the LVS management component and the health
check component in one easily configurable package. Heartbeat then
provides the failover part *and* can also do things to ldirectord,
depending on your configuration. Taken together, heartbeat can detect
the failure of (for example) your current live load balancer and switch
the failover load balancer in to place without you being involved, and
then notify you that it's done this.
There is an alternative, which Malcolm mentioned, called keepalived
(http://www.keepalived.org/). This uses the VRRP protocol to provide IP
failover, manages the LVS for you and provides health checking of your
"real" servers, all in one package.
Configured correctly, even with a streamlined setup with only two
servers, it is possible using either of these methods to remove the SPOF
associated with a load balancing device.
HTH
Graeme
_______________________________________________
Please read the documentation before posting - it's available at:
http://www.linuxvirtualserver.org/
LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
or go to http://lists.graemef.net/mailman/listinfo/lvs-users
|
