|
Hello,
I have set up the following configuration:
serverA: Director + RealServer0 (LocalNode)
serverB: Realserver1
The client connects to VIP:nnnn and gets one
HTML page: either page from serverA (pageA)
or page from serverB (pageB). The expected
maximum load is 10 clients/sec.
Each page contains a few links:
pageA contains links pointing to the DIP
pageB contains links pointing to the RIP
After this page, should the client click on
any link, it gets bound to either DIP or RIP.
The client is not going through LVS anymore.
I monitor serverB health, and in case of
failure, I remove it from the LVS tables.
This means the whole load will be supported
by serverA.
Recently, I was asked to NOT deliver full
load on serverA, but rather send a
"Service not available" page for all those
clients which were meant to get pageB.
The problem is: if serverB is really dead,
I can only store the "Service not available"
page on the Director; but it's already a
LocalNode.
Any ideas along the LVS/iptables lines?
I am proficient with C and wouldn't mind
to use a modified LVS, but I have little
experience with networking.
Thank you,
Daniel Borca
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
|
