On Mon, 27 Nov 2006, Doug Curtis wrote:
So I guess what it boils down to is while the connections
per ip address are evenly distributed, the connections per
hmm. This is a feature, not a bug :-)
In LVS each VIP is scheduled separately. The imbalance
presumably is usually a persistence problem (which you can't
do a whole lot about), ie some clients stay connected for a
long time and some don't. I would expect over a large enough
number of clients (100? 1000?) the connections would even
One way of handling it would be to schedule all connections
under one fwmark, then you only have one service.
I guess the other thing that may throw a wrench into things is that I am
using keepalive in apache.
use the standard value. for https you don't disconnect
Joseph Mack NA3T EME(B,D), FM05lw North Carolina
jmack (at) wm7d (dot) net - azimuthal equidistant map
generator at http://www.wm7d.net/azproj.shtml
Homepage http://www.austintek.com/ It's GNU/Linux!