|
Hi!
I just setup my 3 node network (1 lvs, 2 real servers), and have what seems
to be all of the networking componens working correctly: NAT to communicate
through the LVS to the real servers, and the real servers can see nodes on
the public network.
>From my mx on a cable modem, I can ftp and telnet into the LVS machine, but
when I try to connect to the LVS on port 80 via http, I always get
connection refused.
I do have web servers configured on the real servers, they were working just
great prior to the move to LVS (they do virtual hosting). When I connect to
the web servers on the real server mx, it responds correctly.
Here's my network config:
lvs1.mydomain.com
eth0 216.74.128.10/24
eth1 10.75.1.1/24
eth1:0 10.75.1.254/24 (NAT router)
web1.mydomain.com
eth0 10.75.1.10/24
web2.mydomain.com
eth0 10.75.1.11/24
Here's the configuration file:
nat-router = 10.75.1.254 eth1:0
primary = 216.74.128.10
service = lvs
network = nat
virtual www.myserver.com {
address = 216.74.128.10 eth0
port = 80
active = 1
protocol = tcp
scheduler = wlc
load_monitor = ruptime
server web1.mydomain.com {
address = 10.75.1.10
active = 1
weight = 100
}
server web2.mydomain.com {
address = 10.75.1.11
active = 1
weight = 100
}
}
all servers are running rwhod and I can do an ruptime with good results.
The scheduler is wlc with each web server having a weight of 100.
Two things that I suspect are the web1/web2 not going on eth1:0, should they
bind to the router alias? and also the virtual server's IP -- should it have
one on its own? I want to be able to host a plethora of virtual hosts on
the same IP and have the web server figure out the application routing.
I checked the messages log to see if IPCHAINS was denying the connection
from the LVS node to the real server, but no good ... all I see there are
messages from nanny and it's sending broadcast pings (lame, wish I could
shut off those messages). Maybe that has to do with it? I dunno, but my
firewall configuration doesn't want to allow the broadcast message on
10.75.1.255:513 from nanny. Maybe the LVS doesn't think the system is
running? But when I run the ipvsadm tool, I see the two real servers listed
as "Masq" under the LVS node.
Just to clarify, the message from nanny is "bad load average returned:
lvs1.mydomain.com" and the ipchains message is "REJECT eth0 PROTO=17
216.74.128.10:513 216.74.128.255:513 L=112 S=0x00 I=27039 F=0x0000 T=64
(#5)"
Any help that you guys can provide is greatly appreciated. At this point, I
don't even know where else to go except install a web server in lieu of the
lvs and see if I can get that configured to work.
BTW. I am running RedHat 6.2 with kernel "2.2.14-12". I dunno what my LVS
version is, tho ... it's whatever came with the RH 6.2 distribution (1.1-2
according to the ipvsadm RPM's name).
Thanks.
-- Jake
=============================
Jacob W Anderson
JAVATOPIA.COM
javadude@xxxxxxxxx
=============================
"Never try to be better than someone else, but never cease to try to be the
best you can be." -- Coach John Wooden
|
