Re: [lvs-users] Connecting directly to realservers in a one-network LVS-

To: " users mailing list." <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Subject: Re: [lvs-users] Connecting directly to realservers in a one-network LVS-NAT
From: Joseph Mack NA3T <jmack@xxxxxxxx>
Date: Wed, 28 Nov 2007 10:41:10 -0800 (PST)
On Mon, 26 Nov 2007, Ben Hollingsworth wrote:

> Apparently, the forwarding rules get first dibs.

Thought about this a little bit. ip_vs running on the 
director will look up its connection table. Unless the 
packet RIP->CIP belongs to a connection that ip_vs knows 
about (specified by among other things CIP:port), then ip_vs 
will ignore the packet, which will then be picked up by the 
forwarding rule. ip_vs doesn't pick up packets from 
connections initiated from the realservers (realservers
can be clients of services which have nothing to do with the 
functioning of the LVS).

In this case, you should be able to connect CIP<->RIP on a 
one-network LVS-NAT (even though the packets will take a 
tortuous path due to redirects being turned off), as ip_vs 
will ignore these packets.

However for me, on what would have been a 2.0 or 2.2 kernel, 
the connection from CIP->RIP hung and the HOWTO contains the 
screen dumps of what happened. I then give an explanation, 
without checking that it is the explanation. This 
explanation is at odds with my statement in the first para.

I'm stumped.


Joseph Mack NA3T EME(B,D), FM05lw North Carolina
jmack (at) wm7d (dot) net - azimuthal equidistant map
generator at
Homepage It's GNU/Linux!

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