|
I am being asked to deliver 100% availability between two data centers, one
in Atlanta and one in Knoxville.
I have decided that the LVS project is the way to go, but I'm really not too
sure on how to get there as of yet.
Here's the general gist of the configuration. I have two multi-function
servers, one in Atlanta, GA and one in Knoxville, TN.
These two servers have several primary tasks.
First, they are my nameservers.
Second, they are my Apache proxy servers. I must use two different
apache proxies for each of my primary hosts because the Weblogic apache
module (mod_wl_ssl) does not play nice with mod_proxy and mod_rewrite. I
need mod_proxy and mod_rewrite for my Zope Virtual Hosts.
Thirdly, they are my directors for my LVS configuration. Each of these
two servers will run four LVS servers.
One for Apache (so that I can add more apache virtual hosts and/or
move the apache "realserver" off of the nameserver).
One for my Zope/Apache workhorse server farm.
One for my Bea Servers.
One for my mail exchangers.
Quesion #1: can LVS be used effectively to produce failover from one
data center to another? (If any of my critical servers go down in Atlanta,
such as BEA, can I failover to Knoxville?) If so, what would be a good
place to start on reading up on how to pull this off?
Question #2: When prototyping the LVS_DR setup:
a) my production servers are all Linux machines, will I need to
patch the kernel for the ARP problem on these machines if the director is
using 3 real NICs (eth0 DMZ, eth1 LVS, eth2 internal)?
b) if yes to above, then what patch should I be applying to the
realserver kernels? What is the patch command to apply it? (i.e. patch $0 <
some_patch.diff?)
c) is LVS_DR the way to go, or should I use NAT for my style of
configuration?
Thanks in advance.
-Matt
|
