Kenneth Burgener wrote:[...]
I downloaded the 2.2.3 source code, as a previous email mentioned 2.2.3
did not have the Address family error, compiled and ran that code. I no
longer get the Address family not supported problem, but I am still
seeing the very uneven distribution of load balance.
I compiled again with dynscale disabled:
# ./configure --prefix=/usr/pound --disable-dynscale
# make
# make install
Now there was a very clear cut even distribution 40/40/42. I like the
idea of a dynamic distribution, but when you see numbers like 60/20/25
on 3 identical servers, makes you wonder if there is a flaw with the
dynamic distribution logic.
I am now compiling the 2.2.5 to see if I can get better results with that.
Thanks,
Kenneth
|