PRIMARY and NAT_ROUTER confusion with LVS.CF and PULSE handling.

--- "Brian Gray (BMan)" <[email protected]> wrote:
> Have you read/seen these: 
> http://ha.redhat.com/whitepapers/howto/index.html
> or
http://www.redhat.com/support/resources/howto/piranha/

BMan,

Ok, I read the above "Configuring an LVS Cluster", and
here is one source of confusion - it is probably a
simple thing I just need to be set straight on:

The above doc says to do these steps:
(1) on the primary lvs edit the lvs.cf
(2) copy this lvs.cf to the backup lvs
(3) start/restart the pulse daemon on 
    primary then the secondary.

Here's the initial confusion. Using Piranha_GUI for
setup the "PRIMARY=.." becomes not the PUBLIC address
of PIRANHA/LVS SERVICES usually on the "eth0:1"
virtual interface, but rather "PRIMARY=.." is the host
address for physical interface "eth0" on the public
side!  Again, using Piranha_GUI this is the case - the
"PRIMARY=.." is the public side host interface, not
lvs service interface IP.
Therefore copying the lvs.cf from prime to back will
give the back an incorrect "PRIMARY=.." 
setting!   So how does this confusion affect how
pulse will handle the Piranhs/LVS Public IP address
(versus the host addresses for prime and back).
So shouldn't I edit the lvs.cf on both prime and back
to specify their individual public-side host ip's in
the "PRIMARY=" field (that is what Piranha_GUI does!).
Or am I supposed to set "PRIMARY=.." to the
Piranha/LVS SERVICE IP address which of course is same
on
both prime and back.  Also, I assume I am supposed
to setup eth0:1 (piranha public addr) and eth1:1 (nat
router) to be in place, but not activated at boot
time,
correct? (That is what I do now.)


Additional confusion.  The lvs.cf file has a
"NAT_ROUTER=x.x.x.x eth1:1" setting where 
x.x.x.x is the same (but not simultaneouly activated)
on both the prime and back lvs systems.   
At least that is how I am using it.  Am I mistaken?
Or am I supposed to use two different NAT Router 
IP addresses which remain always active? (Problem is
at the real servers whose apache httpd.conf need to
specify the nat router address.)
So, does pulse also handle failing over of the 
NAT Router IP address also? Or not?  And depending
on if "NAT_ROUTER=.." setting is same or diff on
the prime and back lvs systems, that would mean
more lvs.cf editing not specified in the above 3-step
setup !


Thanks in advance for (please) clarifying
my above statements.

Peter



__________________________________________________
Do You Yahoo!?
Sign up for SBC Yahoo! Dial - First Month Free
http://sbc.yahoo.com