[Linux-cluster] LVS, not so fun today...

David Brieck Jr. dbrieck at gmail.com
Fri Nov 10 01:47:00 UTC 2006


On 11/9/06, isplist at logicore.net <isplist at logicore.net> wrote:
> Ah, interesting. You have found an error! In the package itself?
> Mine is a default install, nothing custom about it so I have to believe I'm
> not the only one with this problem?
>
> > send = "GET / HTTP/1.0rnrn"
> >
> > It should be (at least one mine it is):
> > send = "GET / HTTP/1.0\r\n\r\n"
>
> Changed as suggested, restarted pulse. Nanny now working properly, no more
> timeouts to real servers.
>
> Nov  9 16:44:56 lb52 nanny[3616]: starting LVS client monitor for
> 192.168.1.150:80
> Nov  9 16:44:56 lb52 lvs[3613]: create_monitor for HTTP/cweb92 running as pid
> 3616
> Nov  9 16:44:56 lb52 nanny[3617]: starting LVS client monitor for
> 192.168.1.150:80
> Nov  9 16:44:56 lb52 lvs[3613]: create_monitor for HTTP/cweb93 running as pid
> 3617
> Nov  9 16:44:56 lb52 nanny[3618]: starting LVS client monitor for
> 192.168.1.150:80
> Nov  9 16:44:56 lb52 lvs[3613]: create_monitor for HTTP/cweb94 running as pid
> 3618
> Nov  9 16:44:56 lb52 nanny[3616]: making 192.168.1.92:80 available
> Nov  9 16:44:56 lb52 nanny[3617]: making 192.168.1.93:80 available
> Nov  9 16:44:56 lb52 nanny[3618]: making 192.168.1.94:80 available
> Nov  9 16:45:01 lb52 pulse[3620]: gratuitous lvs arps finished
>
> However, trying to connect to VIP of 192.168.1.150, nothing shows up.
>
> ifconfig shows all looks good.
>

Alright, so what does the ifconfig look like for one of the real
servers? What method are you using to make sure the real server
responds to the VIP requests?

For me, I have the VIP on the loopback adapter and use an arptable
rule to make sure it doesn't respond to arp requests. You need to make
sure apache is listening on the VIP, not it's actual IP address.




More information about the Linux-cluster mailing list