LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: LVS DR trouble reaching back the client

To: "LinuxVirtualServer.org users mailing list." <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Subject: Re: LVS DR trouble reaching back the client
From: Graham Purcocks <grahamp@xxxxxxxxxxxxx>
Date: Wed, 16 Feb 2005 15:23:31 +0000
Isn't the problem that the VIP is required on all the Real servers as well for DR? So they all have to be on the same network. What you have only works for NAT or TUN.

Maybe someone else understands some other magic.

Graham

Harsha wrote:
Hi all

thanx in advance for helping(urgent help needed)

we are a group of 2 students working on LVS impelemtation as a part of final 
year project.

We have succesfully implemented the LVS NAT a month ago and since then we have been struggling to implement LVS TUN and LVS DR.

Can some one please help us with LVS DR ???

This is real long letter please have some patience i have made sure i have not 
missed out on even the smallest of things.

OK here is the test bed

All Systems are Redhat 9 Based Running with a updated Kernal vs 2.4.20 (support 
for LVS)
All Packages running have been installed with packages downloaded from the 
Ultra monkey web Site

heartbeat, ldiretord, ipvsadm running succesfully.

1 Director
2 Real Servers (web server running on it)
1 Client

Client
=======
eth0 CIP=192.168.3.2

Director
========
eth0 VIP=192.168.3.146 (Virtual IP in the subnet of the client)
eth1 DIP=192.168.1.146 (Virtual IP in the subnet of the Real Servers)

Realservers
============

Real server 1

eth0 192.168.1.145 (Real server on which the web server is running)
eth1 192.168.3.145 (for the response to reach the client)
lo:0 192.168.3.146


Real server 2

eth0 192.168.1.147 (Real server on which the web server is running)
eth1 192.168.3.147 (for the response to reach the client)

lo:0 192.168.3.146

Topology
========

                     _________
| Client | |_________|
  _LAN__________|  eth0 192.168.3.2 (CIP)
| | | |
 |                          |  eth0 192.168.3.146 (VIP)
 |                   ____|____
| |Director | | |_________| | | eth1 192.168.1.146 (DIP) | |
 |                          |
 |                          |
 |                 _____|_________________
 |                 |  eth0 192.168.1.145         |  eth0 192.168.1.147
| ____|_____ ___ _____ | | Real | | Real | | | Server 1 | | Server 2 | | ---------------- ----------------
 |                  | eth1 192.168.3.145          |  eth1 192.168.3.145
 |                  |                                       |
 |                  |______________________|
 |                                     |
 |                                     |
 |_______LAN__________ |


The eth0 interfaces of Director & Real servers are connected through a hub i.e 192.168.1.145 192.168.1.146 192.168.1.147

and all the eth1 interfaces (except client) are connected to the internal 
network to which the client is connected (the client has only one interface)

No routers present !!!!!!!!!!!


Things we have done


1) Configured the lDirector with DR configuration
2) Configured the web servers on both real servers
3) Enabled IP fwd on both the real servers

IP forwarding is not needed on the Linux Director ???.

Problem
=======

when a Request is sent from the client to the VIP. The error "Page cannot be displayed " is encountered

Director#

when we type the command 'ipvsadm -L --stats'

Based on the output we can see that the director is receiving packets and load 
balancing is also being done.

The connections are distributed between the real servers, we can see the received Bytes BUT the SENT BYTES column is always zero no matter how many conections are recieved

Assumptions
===========

Based on the output we are assuming that the realservers are recieving the packets
But we are not sure whether the packets are being processed by the Real servers 
or there is a problem with the request reaching back the client.

we are not sure what GATEWAYS to be specified ???

ARP PRoblem
===========

we have created a lo:0 VIP interface on the real servers so that the packets 
appear local and are processed.

we have PROBABLY hidded the lo:0 interfaces on both the real servers

we have actually created the config files for lo:0
and then used the commands

echo "1" > /proc/sys/net/ipv4/conf/all/hidden
echo "1" > /proc/sys/net/ipv4/conf/lo/hidden

which should have solved the problem.
but we are not sure whether the arp patch is set up or not

Does the kernal i have is already patched with the noarp patch ????

i haved tried typing 'lsmod' the output had nothing called noarp module

i have tried 'insmod noarp'
which said module not found

HELP US
=======

We would be grateful if someone could help us because we are in a very troubled state as we need to complete the project in a week's time We want to know
1) what GATEWAYS to be specified on which machines ???
2) Clarity on the noarp mod and hidding the lo:0 interfaces
3) how to route back the request to the client

PS: we would be even more grateful if u can mail back us with a solution at any 
one of the following addresses

harsha040@xxxxxxxxx
pradeepnair83@xxxxxxxxx

Yahoo! India Matrimony: Find your life partneronline.
_______________________________________________
LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
or go to http://www.in-addr.de/mailman/listinfo/lvs-users




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