1

I have have set up MySQL multi-master replication with HAProxy as a load balancer following Alex Williams' tutorial found here; http://www.alexwilliams.ca/blog/2009/08/10/using-haproxy-for-mysql-failover-and-redundancy/

I am using a status checker found here; http://sysbible.org/2008/12/04/having-haproxy-check-mysql-status-through-a-xinetd-script/

as well as a replication checker found here; http://forge.mysql.com/tools/tool.php?id=6

The Mysql status checker works fine when I have the following line in my HAProxy config file: 

server db02 192.168.15.119:3306 check port 9200 inter 1s weight 1 rise 1 fall 1

But when I try to test the check replication script with this line:

server db02 192.168.15.119:3306 check port 9201 inter 1s weight 1 rise 1 fall 1

and try to connect through the proxy I get a failure:

ERROR 2013 (HY000): Lost connection to MySQL server at 'reading initial communication  packet', system error: 0

When I telnet from the HAProxy machine to that port it works fine:

# telnet 192.168.15.119 9201

Trying 192.168.15.119...
Connected to 192.168.15.119.
Escape character is '^]'.
Replication slave is running
Connection closed by foreign host.

Is there any reason that that would not work while I can still check it through telnet? Is this a limitation of HAProxy? I would assume not as it seems that this is what others have used for their setups. Is it a configuration or permissions issue? I am confused as to how one of them would work and let me connect fine, but the other will not.

Joe Gibson
  • 45
  • 8
  • What do you want to do with the replication script checker? Please show us the content of `haproxy.cfg`, `/etc/xinetd.d/`? – quanta Jul 12 '12 at 03:15
  • I wanted to check if replication was working every time a connection was made through the proxy. I already gave the pertinent info from haproxy.cfg, and I know the xinetd service is running properly because I can telnet to the port and it runs the script. It is only when telling HAProxy to check the port for the replication checker that it fails. In any event, I just put the script to run every minute in crontab, which is suitable for my situation, although not exactly an answer to the question. – Joe Gibson Jul 12 '12 at 13:46
  • I need the full `haproxy.cfg` or at least MySQL `backend` section. – quanta Jul 12 '12 at 14:43

1 Answers1

1

Your replication script run by xinetd should return HTTP 200 or 503, so HAProxy can detect the health of MySQL replication.

Arie K
  • 1,583
  • 5
  • 18
  • 27