[lvs-users] MySQL Load Balancer/Cluster Connection Refused on Virtual IP

[Graeme Fowler]

On Fri, 2008-05-30 at 11:38 +0100, Paul Bowden wrote:
> I'm not sure if you wanted me to run tcpdump -qn host client-1 on the
> load balancers as well, but I ran it on lb1, lb2, sql1 and sql2 and
> here are the results:

...and a good thing you did, because:

> lb1 (which currently has the vip) is the only machine that mentions
> anything to do with my attempted telnet:
> 
> lb1:
<snip>
> 11:34:59.362084 IP 192.168.1.177.3312 > 192.168.1.243.3306: tcp 0
> 11:34:59.362104 IP 192.168.1.243 > 192.168.1.177: ICMP 192.168.1.243
> tcp port 3306 unreachable, length 56
> 11:35:02.352094 IP 192.168.1.177.3312 > 192.168.1.243.3306: tcp 0
> 11:35:02.352108 IP 192.168.1.243 > 192.168.1.177: ICMP 192.168.1.243
> tcp port 3306 unreachable, length 56

There's the interesting bit - the connection is rejected. If I now take
a bit more time to look back at your previous message :) then I see:

> IP Virtual Server version 1.2.1 (size=4096)
> Prot LocalAddress:Port Scheduler Flags
>  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
> TCP  192.168.1.243:3306 wrr
>  -> 192.168.1.246:3306           Route   0      0          0
>  -> 192.168.1.247:3306           Route   0      0          0

Aha! Both realservers have a weight of zero - the VIP, therefore, can't
handle any (new) requests. You have "quiescent=yes" which means if the
server is determined to be down, the weight is set to 0.

You need to determine why the servers look to be down. Can you connect
to the MySQL port on the realservers from the director using the same
connection details you have configured in ldirectord.cf?