|
Hi,
Im having a bit of problems getting the LVS-dr technique to work.
I have a load balancer:
eth0: (10.0.0.171)
eth0:1 (10.0.0.170) <- virtual cluster address
eth1 (192.168.12.5)
And one realserver:
eth0: (10.0.0.172)
eth1 (192.168.12.5)
The servers are connected with two seperate hubs. I can ping both
servers trough both ip adresses.
But now im a but puzzled on how to get things up and running:
I used lvsadm to add a redirect to the realserver on the load balancer
for port 22 (ssh)
i was able to connect to the cluster ip (170) before i added the
redirect and i was getting the console of the load balancer.
Now since i have the redirect i get a timeout (because the realserver is
not responding to the 170 address), i can connec to both the other
adresses (171, 172)
Now i understand that i need to add the 170 addres (noarp) to the
realserver for it to respond to those IPs.
The ssh servers are setup to listen to all ip.s
As soon as i try to add 10.0.0.170 to any of the Interfaces on the
realserver my other address stops responding (10.0.0.172)
I tried adding it to LO, to eth0, and to eth1, none of them work.
If i add -arp to the ifconfig command, the entire interface uses noarp,
not just the alias.
Now i see all kind of noarp patches for the realserver's their kernels,
i use a 2.6 kernel, and since i can't find a patch for a 2.6 kernel i
did not use any
Do i need a patch to make this noarp thing work as it should?
I am using the document created by jeremy heslop, (even though my
realservers are linux servers, and not win2k servers)
thank you loads!
greets
Jan
|
