[CentOS] Load Balancer (heartbeat, ldirectord, ipvsadm) + Apache Problem

Wed Feb 17 09:49:52 UTC 2010
Peter Hinse <loco at d0pefish.de>

Am 17.02.2010 09:29, schrieb Roberto HT:

> I'm having this problem with load balancer.
> I already setup the load balancer and two servers behind it with this
> kind of topology
> 
> -----------------------------------------------------                   
> request -> DIRECTOR -- Server 1
>                                    \
>                                      Server 2
> -----------------------------------------------------
> 
> DIRECTOR's (RHEL 5) Virtual IP = 192.168.1.1
> Server 1's (RHEL 4.4) Virtual IP = 192.168.1.2
> Server 2's (CentOS 5.3) Virtual IP = 192.168.1.3
> 
> The problem is when I try to hit the 192.168.1.1, it sometimes give me a
> HTTP 200 and sometimes it will timed out. I hit it using lynx -dump. It
> should return "Hello world!" if success.
> 
> But if we hit it directly to 192.168.1.2:80 <http://192.168.1.2:80> and
> 192.168.1.3:80 <http://192.168.1.3:80>, it will return HTTP 200 which
> means no error and no timeout. Since the Server 1 is the same as Server
> 2 in configuration, are there somethings I missed here?
> 
> Another information is that I set the director using round robin
> algorithm, so everytime a successful hit returned from server 1, the
> next one is surely heading to Server 2 right? Based on the Apache's
> access log, I also knew that the timed out occured when it is the Server
> 2's turn to handle the request.
> 
> Aside from the same httpd.conf and the same httpd version are there any
> things in the system level, kernel level that should be the same between
> those two servers?

Could you please post the heartbeat/ldirectord configs from /etc/ha.d/?

Regards,

	Peter