|
Hi Raph
On Tue, 2008-05-06 at 15:16 +0100, Raphael Gangneux wrote:
> I apologise in advance if this seems obvious to you, but this is my first
> attempt at installing/configuring LVS.
My advice is: keep it simple, and build from that.
Get yourself one director and one realserver. Make sure that you can do
the ipvsadm stuff *by hand* on the director before moving towards
keepalived, ldirectord, heartbeat, mon, ultramonkey et al to produce a
fully load-balanced+HA solution.
Remember the following general points:
1. The realservers can't be clients of the LVS.
2. The director can't be a client of the LVS.
3. Any clients *should* - unless you have a very specific need otherwise
- be on a different IP subnet to your realservers, ideally on a
different LAN segment.
The first two points are very specific, and complex to work around.
The third one is a generalisation applicable to LVS-NAT and can be
worked around by forcing them to only talk via the director.
How have you got on so far?
Graeme
|
