|
Hmm, this seems strange to use eth0:0 for real servers. Can anyone else
confirm this for fwmark operations?
When using fwmark, does the hidden lo and all sysctl's still need to be
set?
-jeremy
> I could only get it to work if I used eth0:0 instead of lo:0 for the virtual
> server interface. I posted a question about the reason for usiing lo:0, but
> nobody has yet responded. I'm using the latest patches on 2.2.15 kernel.
> Tom Morris
> CrosStor Software
> Jeremy Hansen wrote:
>
> > So I'm trying to use fwmark and I'm running into problems. I'm using the
> > UltraMonkey examples and I think where I'm having the problem is with the
> > 192.168.0.0/24 network example. My cluster is setup using 10.2.9.0/24.
> >
> > floating ip is 10.2.9.5
> >
> > ld1 and ld2, 10.2.9.2, and 10.2.9.3
> >
> > real servers
> >
> > web1-3, 10.2.9.10-12
> >
> > So I decided to use 10.2.20.0/24 in place of the 192.168.0.0/24 in the
> > example.
> >
> > As far as I can tell I had things exactly as they instruct in the example.
> >
> > On real server lo:0 was config'd for 10.2.20.0
> >
> > ipchains was accepting all, but I did do the ipchains -A input -d
> > 10.2.20.0/24 -m 1 and added the route for 10.2.20.0/24 to gateway via the
> > vip, 10.2.9.5.
> >
> > and ipvsadm used ipvsadm -A -f 1 vip1:80 and ipvsadm -a -f -1 -r
> > webtest1:80, etc.
> >
> > so did all this and it did not work...so the only thing looking fishy to
> > me is my use of 10.2.20.0/24. Any ideas what I could be overlooking?
> >
> > Also is there any real performance advantages to using fwmark? I guess
> > the cool thing is it easily allows me to use other services, right?
> >
> > Thanks
> > -jeremy
> >
>
--
http://www.xxedgexx.com | jeremy@xxxxxxxxxxxx
---------------------------------------------
|
