[FreeNX-kNX] FreeNX and LVS (Linux virtual Server - cluster)

[Henk Poley]

Op maandag 11 juli 2005 00:48, schreef Daniel Schwager:
> Architecture: We have to install 3 LNX-applicationserver in the backend
> (joined together via LVS - Linuxvirtual Server = Node A). The LVS
> loadbalance between the 3 servers. Because we want to use the
> nx-permanent-session feature, the nxserver have to run outside the
> LVS-cluster (we want to run it on the redundant LVS-node).
>
> <snip>
>
> Now my question: Is it possible to use the
> nx-protokoll between the thinclient  and the node A before the
> applicationcluster
> _AND_ start the user-applications via nx-protocoll between the node A
> and the applicationserver VIP
> _AND_ use nx-multimedia and nx-samba-ressources between thinclient and VIP
> _AND_ suspend/resume sessions (means the nx-session between Node A and
> the VIP must
> keep alive because a new session will link to another applicationserver
> (app1,2,3) via LVS.
>
> Maybe somebody can help me to track down this question.

From the LVS Howto:

"The association between the client and the realserver will last for only the 
life of the tcp connection (or udp exchange). For the next tcp connection, 
the director will choose a new realserver (which may or may not be the same 
as the first realserver). Thus a web browser connecting to an LVS serving a 
webpage consisting of several hits (images, html page), may get each hit from 
a separate realserver."

NX runs over just one SSH connection. This will basicly mean NX will work for 
one connection, but on a resume the client will most probably hit another 
server. If I'm correct, the servers in an LVS cluster still have different 
personalities (it's not a single shared system image with shared RAM, like 
for example beowulf). So hitting another server will create a new NX session 
on this server, it won't know of the pre-existing NX session since it's not 
available on this server.

Googling for "LVS SSH" brings up that there's something named "LVS-NAT" that 
will redirect certain ports to one single server. Maybe that mechanism is 
strong enough to point different groups of clients to a different single 
server (based on MAC address for example). But that would be a crude way of 
load balancing.

I should note that LVS seems to be built as a system to load balance HTTP 
servers. HTTP has relatively shortlived connections and session data is 
mostly stored in seperate databases accessable for all the servers in the 
cluster.

	Henk Poley <><