LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: Question about LVS-DR setup

To: "'''lvs-users@xxxxxxxxxxxxxxxxxxxxxx' ' '" <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Subject: Re: Question about LVS-DR setup
From: djo@xxxxxxxxxxxxxxxxxxxxxx
Date: Tue, 18 Jun 2002 00:38:10 -0700
Either on of these topologies below will work.  It is not clear to me yet
what is gained by the two NIC solution over the one NIC one.

#               ________
#              |        |
#              | client |
#              |________|
#                  | 
#                  |
#                  |
#                  |       __________
#                  |      |          |
#                  |------| director |
#                  |      |__________|
#                  | 
#                  | 
#                  |  
#                  |
#                  |
#         -------------------
#         |                 | 
#         |                 |
#   ______________    ______________
#  |              |  |              | 
#  | realserver1  |  | realserver2  |
#  |______________|  |______________|

------------------------------ or ----------------------------------------
#               ________
#              |        |
#              | client |
#              |________|
#                  |
#                  |
#              __________
#             |          |
#             | director |
#             |__________|
#                  |  
#                  |
#                  |
#         -------------------
#         |                 | 
#         |                 |
#   ______________    ______________
#  |              |  |              | 
#  | realserver1  |  | realserver2  |
#  |______________|  |______________|

In all the LVS setup documents I keep reading they keep refering to the
client i.p. or CIP as 192.168.1.254   This is the first thing that throws
me.  In the real world the client would have some address like 206.23.124.48
I will have one "live" IP such as 66.121.124.117  The examples are confusing
because they seem to imply that the client is on the same physical segment
as the director.
The next question is: is the VIP supposed to be the "live" address I have,
i.e. 66.121.124.117?  If this is so, why in the examples do they use what
looks as an IP address in the range reserved in RFCs for internal private
subnet addresses?  
To get this straight, assuming I am correct, then if the VIP is suppposed to
be a "live" address I have available at my colo or from my ISP, then for
example in a one-nic setup I would have something like this in lvs_nat.conf:
VIP=eth0:117 66.121.124.117 255.255.255.248 66.121.124.119
DIP=eth0 192.168.1.10 255.255.255.0 192.168.1.255

Again, from the examples it is so confusing because the CIP and VIP both
appear to be an address on a private internal subnet and if that is the
case nothing is mentioned about how this is happening.  I mean, for example,
what if I throw a box outside of the internal subnet which is a
router/gateway running iptables/SNAT?  

So anyhow, this is where I am stuck because I have setup lvs_nat (I've
abandoned lvs_dr for reasons you mentioned: I don't want the realservers to
be using live IPs) yet nothing is able to reach the 66.121.124.117 VIP
eth0:117.  Interestingly enough though my realservers are able to ping out
through 192.168.1.10 as their gateway so something at least is working.


<Prev in Thread] Current Thread [Next in Thread>