|
Hi,
Let me specifying the problem, in details.
I want to serving a website with two machines (A,B), in
loadbalancing-failover enviroment. I have two IP-s, assuming 10.0.2.40 and
10.0.2.39.
The dynamic part of the site will be served by apache+php, listening on
10.0.2.40:80, and the static part of the site (css,images) will be served by
a lighttpd, listening on 10.0.2.39:80
If apache runs on the box A, and lighttpd runs on the box B, both of the
machines are SPOF, so I have to think up something.
I plan to use IPs as VIP addresses, and I want to use 4 more IPs as
realservers. Like this, one box can serve the whole page, and the other box
is not idle.
haresources:
deb1 \
ldirectord::ldirectord.cf \
LVSSyncDaemonSwap::master \
IPaddr::10.0.2.40/24/eth0/10.0.2.255 \
IPaddr::10.0.2.39/24/eth0/10.0.2.255 \
My ldirectord.cf look like:
virtual = 10.0.2.40:80
real = 10.0.2.10:80 gate 10
real = 10.0.2.11:80 gate 10
service = http
scheduler = rr
protocol = tcp
request = "test.html"
receive = "OK"
checktype = negotiate
virtual = 10.0.2.39:80
real = 10.0.2.20:80 gate 10
real = 10.0.2.21:80 gate 10
service = http
scheduler = rr
protocol = tcp
request = "test.html"
receive = "OK"
checktype = negotiate
The IPs 10.0.2.10 and 10.0.2.20 are on the box A, and the IPs 10.0.2.11 and
10.0.2.21 are on the box B.
I need setting up dummy0 and dummy0:0 interfaces on both of the machines
with 10.0.2.39 and 10.0.2.40, thus apache and lighttpd can bind on this IPs.
Now comes the problem: If the heartbeat starts, it drops the dummy0:0
interface with IP 10.0.2.39, but the 10.0.2.40 on the dummy0 remains
untouched.
This reason the lighttpd can't serve request on this IP.
If I call ifdown dummy0:0 and ifup dummy0:0, the system works perfectly. How
can I make it reboot-safe, and failover safe?
This is relevant line of the log:
ResourceManager[2614]: 2009/12/01_19:50:21 info: Running
/etc/ha.d/resource.d/IPaddr 10.0.2.40/24/eth0/10.0.2.255 start
ResourceManager[2614]: 2009/12/01_19:50:21 debug: Starting
/etc/ha.d/resource.d/IPaddr 10.0.2.40/24/eth0/10.0.2.255 start
IPaddr[2890]: 2009/12/01_19:50:22 INFO: Using calculated netmask for
10.0.2.40: 255.255.255.0
IPaddr[2890]: 2009/12/01_19:50:22 INFO: eval ifconfig eth0:1 10.0.2.40
netmask 255.255.255.0 broadcast 10.0.2.255
IPaddr[2890]: 2009/12/01_19:50:22 DEBUG: Sending Gratuitous Arp for
10.0.2.40 on eth0:1 [eth0]
IPaddr[2861]: 2009/12/01_19:50:22 INFO: Success
INFO: Success
ResourceManager[2614]: 2009/12/01_19:50:22 debug:
/etc/ha.d/resource.d/IPaddr 10.0.2.40/24/eth0/10.0.2.255 start done. RC=0
IPaddr[3025]: 2009/12/01_19:50:23 ERROR: 10.0.2.39 is running an interface
(dummy0) instead of the configured one (eth0)
IPaddr[2996]: 2009/12/01_19:50:23 ERROR: Generic error
ResourceManager[2614]: 2009/12/01_19:50:23 info: Running
/etc/ha.d/resource.d/IPaddr 10.0.2.39/24/eth0/10.0.2.255 start
ResourceManager[2614]: 2009/12/01_19:50:23 debug: Starting
/etc/ha.d/resource.d/IPaddr 10.0.2.39/24/eth0/10.0.2.255 start
IPaddr[3105]: 2009/12/01_19:50:24 INFO: Using calculated netmask for
10.0.2.39: 255.255.255.0
IPaddr[3105]: 2009/12/01_19:50:24 ERROR: 10.0.2.39 is running an interface
(dummy0) instead of the configured one (eth0)
IPaddr[3105]: 2009/12/01_19:50:25 INFO: eval ifconfig eth0:2 10.0.2.39
netmask 255.255.255.0 broadcast 10.0.2.255
IPaddr[3105]: 2009/12/01_19:50:25 DEBUG: Sending Gratuitous Arp for
10.0.2.39 on eth0:2 [eth0]
IPaddr[3105]: 2009/12/01_19:50:25 ERROR: 10.0.2.39 is running an interface
(dummy0) instead of the configured one (eth0)
IPaddr[3076]: 2009/12/01_19:50:25 ERROR: Generic error
ERROR: Generic error
--
Üdv,
Ricsi
> -----Original Message-----
> From: lvs-users-bounces@xxxxxxxxxxxxxxxxxxxxxx [mailto:lvs-users-
> bounces@xxxxxxxxxxxxxxxxxxxxxx] On Behalf Of Simon Horman
> Sent: Friday, November 27, 2009 1:16 PM
> To: LinuxVirtualServer.org users mailing list.
> Subject: Re: [lvs-users] more realserver ports question
>
> On Thu, Nov 26, 2009 at 10:29:26AM -0800, James H wrote:
> > Here is a patch to the RedHat Pulse/Piranha system that let's you use
> > NAT port rewriting:
> >
> > http://mobiuscircle.com/2009/06/08/piranhapulse-patch-allow-port-
> rewriting-in-nat-mode/
> >
> > This patch adds the ability for you to set a port in the real server
> > config section in /etc/sysconfig/ha/lvs.conf
> >
> > eg:
> > ...
> > virtual http-shared {
> > port = 80
> > ...
> > server rack104 {
> > address = 192.168.2.104
> > port = 8080 <==== Isn't normally supported by pulse/piranha
> > active = 1
> > weight = 1
> > }
> > ...
> > }
> > ...
>
> What does ipvsadm -Ln look like with this in effect?
> I'm curious to know where the port translation is taking place.
>
>
> _______________________________________________
> Please read the documentation before posting - it's available at:
> http://www.linuxvirtualserver.org/
>
> LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
> Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
> or go to http://lists.graemef.net/mailman/listinfo/lvs-users
_______________________________________________
Please read the documentation before posting - it's available at:
http://www.linuxvirtualserver.org/
LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
or go to http://lists.graemef.net/mailman/listinfo/lvs-users
|
